In a combination with a wallet size of 1 billion it should never be able to run out of money avoiding false-positives of some users who just wanted to test a strategy without actually checking how the stake_amount-variable should be used in combination with the strategy-function custom_stake_amount.
reason: some strategies demand a custom_stake_amount of 1$ demanding a very large wallet-size (which already was set previously)
Others start with 100% of a slot size and subdivide the base-orders and safety-orders down to finish at 100% of a slot-size and use unlimited stake_amount.
Edited docs to reflect that change.
In a combination with a wallet size of 1 billion it should never be able to run out of money avoiding false-positives of some users who just wanted to test a strategy without actually checking how the stake_amount-variable should be used in combination with the strategy-function custom_stake_amount
reason: some strategies demand a custom_stake_amount of 1$ demanding a very large wallet-size (which already was set previously)
Others start with 100% of a slot size and subdivide the base-orders and safety-orders down to finish at 100% of a slot-size and use unlimited stake_amount.
Edited docs to reflect that change too
There are some trade- and candle-related fields that are always available to output on the indicator-list so have updated the docs to include the most commonly used ones.
- moved doc from utils.md to lookahead-analysis.md and modified it (unfinished)
- added a check to automatically edit the config['backtest_cache'] to be 'none'
- adjusted test_lookahead_helper_export_to_csv to catch the new catching of errors
- adjusted test_lookahead_helper_text_table_lookahead_analysis_instances to catch the new catching of errors
- changed lookahead_analysis.start result-reporting to show that not enough trades were caught including x of y
moved doc from utils.md to lookahead-analysis.md and modified it (unfinished)
added a check to automatically edit the config['backtest_cache'] to be 'none'
Looking at has_bias should be enough to statisfy the test.
The tests could be extended with thecking the buy/sell signals and the dataframe itself -
but this should be sufficient for now.
There was a seeding error in SB3 after the gymnasium update, the stable baselines team has patched and fixed the issue, but the reset function has to be aligned.
switched from args to config (args still work)
renamed exportfilename to lookahead_analysis_exportfilename so if users decide to put something into it then it won't compete with other configurations
- optimized pairs for entry_varholder and exit_varholder to only check a single pair instead of all pairs.
- bias-check of freqai strategies now possible
- added condition to not crash when compared_df is empty (meaning no differences have been found)
open ended timeranges now work
if a file fails then it will not report as non-bias, but report in the table as error and the csv file will not have it listed.
removed args_common_optimize for strategy-updater
backtest_lookahead_bias_checker:
added args and cli-options for minimum and target trade amounts
fixed code according to best-practice coding requests of matthias (CamelCase etc)
@@ -136,7 +136,7 @@ class MyAwesomeStrategy(IStrategy):
### Dynamic parameters
Parameters can also be defined dynamically, but must be available to the instance once the * [`bot_start()` callback](strategy-callbacks.md#bot-start) has been called.
Parameters can also be defined dynamically, but must be available to the instance once the [`bot_start()` callback](strategy-callbacks.md#bot-start) has been called.
@@ -304,7 +305,7 @@ A backtesting result will look like that:
| Sharpe | 2.97 |
| Calmar | 6.29 |
| Profit factor | 1.11 |
| Expectancy | -0.15 |
| Expectancy (Ratio) | -0.15 (-0.05) |
| Avg. stake amount | 0.001 BTC |
| Total trade volume | 0.429 BTC |
| | |
@@ -323,6 +324,7 @@ A backtesting result will look like that:
| Days win/draw/lose | 12 / 82 / 25 |
| Avg. Duration Winners | 4:23:00 |
| Avg. Duration Loser | 6:55:00 |
| Max Consecutive Wins / Loss | 3 / 4 |
| Rejected Entry signals | 3089 |
| Entry/Exit Timeouts | 0 / 0 |
| Canceled Trade Entries | 34 |
@@ -408,7 +410,7 @@ It contains some useful key metrics about performance of your strategy on backte
| Sharpe | 2.97 |
| Calmar | 6.29 |
| Profit factor | 1.11 |
| Expectancy | -0.15 |
| Expectancy (Ratio) | -0.15 (-0.05) |
| Avg. stake amount | 0.001 BTC |
| Total trade volume | 0.429 BTC |
| | |
@@ -427,6 +429,7 @@ It contains some useful key metrics about performance of your strategy on backte
| Days win/draw/lose | 12 / 82 / 25 |
| Avg. Duration Winners | 4:23:00 |
| Avg. Duration Loser | 6:55:00 |
| Max Consecutive Wins / Loss | 3 / 4 |
| Rejected Entry signals | 3089 |
| Entry/Exit Timeouts | 0 / 0 |
| Canceled Trade Entries | 34 |
@@ -466,6 +469,7 @@ It contains some useful key metrics about performance of your strategy on backte
-`Best day` / `Worst day`: Best and worst day based on daily profit.
-`Days win/draw/lose`: Winning / Losing days (draws are usually days without closed trade).
-`Avg. Duration Winners` / `Avg. Duration Loser`: Average durations for winning and losing trades.
-`Max Consecutive Wins / Loss`: Maximum consecutive wins/losses in a row.
-`Rejected Entry signals`: Trade entry signals that could not be acted upon due to `max_open_trades` being reached.
-`Entry/Exit Timeouts`: Entry/exit orders which did not fill (only applicable if custom pricing is used).
-`Canceled Trade Entries`: Number of trades that have been canceled by user request via `adjust_entry_price`.
@@ -533,6 +537,7 @@ Since backtesting lacks some detailed information about what happens within a ca
- ROI
- exits are compared to high - but the ROI value is used (e.g. ROI = 2%, high=5% - so the exit will be at 2%)
- exits are never "below the candle", so a ROI of 2% may result in a exit at 2.4% if low was at 2.4% profit
- ROI entries which came into effect on the triggering candle (e.g. `120: 0.02` for 1h candles, from `60: 0.05`) will use the candle's open as exit rate
- Force-exits caused by `<N>=-1` ROI entries use low as exit value, unless N falls on the candle open (e.g. `120: -1` for 1h candles)
- Stoploss exits happen exactly at stoploss price, even if low was lower, but the loss will be `2 * fees` higher than the stoploss price
- Stoploss is evaluated before ROI within one candle. So you can often see more trades with the `stoploss` exit reason comparing to the results obtained with the same strategy in the Dry Run/Live Trade modes
@@ -7,7 +7,7 @@ This page provides you some basic concepts on how Freqtrade works and operates.
* **Strategy**: Your trading strategy, telling the bot what to do.
* **Trade**: Open position.
* **Open Order**: Order which is currently placed on the exchange, and is not yet complete.
* **Pair**: Tradable pair, usually in the format of Base/Quote (e.g. XRP/USDT).
* **Pair**: Tradable pair, usually in the format of Base/Quote (e.g. `XRP/USDT` for spot, `XRP/USDT:USDT` for futures).
* **Timeframe**: Candle length to use (e.g. `"5m"`, `"1h"`, ...).
* **Indicators**: Technical indicators (SMA, EMA, RSI, ...).
* **Limit order**: Limit orders which execute at the defined limit price or better.
@@ -20,6 +20,20 @@ This page provides you some basic concepts on how Freqtrade works and operates.
All profit calculations of Freqtrade include fees. For Backtesting / Hyperopt / Dry-run modes, the exchange default fee is used (lowest tier on the exchange). For live operations, fees are used as applied by the exchange (this includes BNB rebates etc.).
## Pair naming
Freqtrade follows the [ccxt naming convention](https://docs.ccxt.com/#/README?id=consistency-of-base-and-quote-currencies) for currencies.
Using the wrong naming convention in the wrong market will usually result in the bot not recognizing the pair, usually resulting in errors like "this pair is not available".
### Spot pair naming
For spot pairs, naming will be `base/quote` (e.g. `ETH/USDT`).
### Futures pair naming
For futures pairs, naming will be `base/quote:settle` (e.g. `ETH/USDT:USDT`).
## Bot execution logic
Starting freqtrade in dry-run or live mode (using `freqtrade trade`) will start the bot and start the bot iteration loop.
This page explains the different parameters of the bot and how to run it.
!!! Note
If you've used `setup.sh`, don't forget to activate your virtual environment (`source .env/bin/activate`) before running freqtrade commands.
If you've used `setup.sh`, don't forget to activate your virtual environment (`source .venv/bin/activate`) before running freqtrade commands.
!!! Warning "Up-to-date clock"
The clock on the system running the bot must be accurate, synchronized to a NTP server frequently enough to avoid problems with communication to the exchanges.
@@ -138,7 +138,7 @@ Mandatory parameters are marked as **Required**, which means that they are requi
| `stake_currency` | **Required.** Crypto-currency used for trading. <br> **Datatype:** String
| `stake_amount` | **Required.** Amount of crypto-currency your bot will use for each trade. Set it to `"unlimited"` to allow the bot to use all available balance. [More information below](#configuring-amount-per-trade). <br> **Datatype:** Positive float or `"unlimited"`.
| `tradable_balance_ratio` | Ratio of the total account balance the bot is allowed to trade. [More information below](#configuring-amount-per-trade). <br>*Defaults to `0.99` 99%).*<br> **Datatype:** Positive float between `0.1` and `1.0`.
| `available_capital` | Available starting capital for the bot. Useful when running multiple bots on the same exchange account.[More information below](#configuring-amount-per-trade). <br> **Datatype:** Positive float.
| `available_capital` | Available starting capital for the bot. Useful when running multiple bots on the same exchange account.[More information below](#configuring-amount-per-trade). <br> **Datatype:** Positive float.
| `amend_last_stake_amount` | Use reduced last stake amount if necessary. [More information below](#configuring-amount-per-trade). <br>*Defaults to `false`.* <br> **Datatype:** Boolean
| `last_stake_amount_min_ratio` | Defines minimum stake amount that has to be left and executed. Applies only to the last stake amount when it's amended to a reduced value (i.e. if `amend_last_stake_amount` is set to `true`). [More information below](#configuring-amount-per-trade). <br>*Defaults to `0.5`.* <br> **Datatype:** Float (as ratio)
| `amount_reserve_percent` | Reserve some amount in min pair stake amount. The bot will reserve `amount_reserve_percent` + stoploss value when calculating min pair stake amount in order to avoid possible trade refusals. <br>*Defaults to `0.05` (5%).* <br> **Datatype:** Positive Float as ratio.
@@ -155,29 +155,29 @@ Mandatory parameters are marked as **Required**, which means that they are requi
| `trailing_stop_positive_offset` | Offset on when to apply `trailing_stop_positive`. Percentage value which should be positive. More details in the [stoploss documentation](stoploss.md#trailing-stop-loss-only-once-the-trade-has-reached-a-certain-offset). [Strategy Override](#parameters-in-the-strategy). <br>*Defaults to `0.0` (no offset).* <br> **Datatype:** Float
| `trailing_only_offset_is_reached` | Only apply trailing stoploss when the offset is reached. [stoploss documentation](stoploss.md). [Strategy Override](#parameters-in-the-strategy). <br>*Defaults to `false`.* <br> **Datatype:** Boolean
| `fee` | Fee used during backtesting / dry-runs. Should normally not be configured, which has freqtrade fall back to the exchange default fee. Set as ratio (e.g. 0.001 = 0.1%). Fee is applied twice for each trade, once when buying, once when selling. <br> **Datatype:** Float (as ratio)
| `futures_funding_rate` | User-specified funding rate to be used when historical funding rates are not available from the exchange. This does not overwrite real historical rates. It is recommended that this be set to 0 unless you are testing a specific coin and you understand how the funding rate will affect freqtrade's profit calculations. [More information here](leverage.md#unavailable-funding-rates) <br>*Defaults to None.*<br> **Datatype:** Float
| `futures_funding_rate` | User-specified funding rate to be used when historical funding rates are not available from the exchange. This does not overwrite real historical rates. It is recommended that this be set to 0 unless you are testing a specific coin and you understand how the funding rate will affect freqtrade's profit calculations. [More information here](leverage.md#unavailable-funding-rates) <br>*Defaults to `None`.*<br> **Datatype:** Float
| `trading_mode` | Specifies if you want to trade regularly, trade with leverage, or trade contracts whose prices are derived from matching cryptocurrency prices. [leverage documentation](leverage.md). <br>*Defaults to `"spot"`.* <br> **Datatype:** String
| `margin_mode` | When trading with leverage, this determines if the collateral owned by the trader will be shared or isolated to each trading pair [leverage documentation](leverage.md). <br> **Datatype:** String
| `liquidation_buffer` | A ratio specifying how large of a safety net to place between the liquidation price and the stoploss to prevent a position from reaching the liquidation price [leverage documentation](leverage.md). <br>*Defaults to `0.05`.* <br> **Datatype:** Float
| | **Unfilled timeout**
| `unfilledtimeout.entry` | **Required.** How long (in minutes or seconds) the bot will wait for an unfilled entry order to complete, after which the order will be cancelled and repeated at current (new) price, as long as there is a signal. [Strategy Override](#parameters-in-the-strategy).<br> **Datatype:** Integer
| `unfilledtimeout.exit` | **Required.** How long (in minutes or seconds) the bot will wait for an unfilled exit order to complete, after which the order will be cancelled and repeated at current (new) price, as long as there is a signal. [Strategy Override](#parameters-in-the-strategy).<br> **Datatype:** Integer
| `unfilledtimeout.unit` | Unit to use in unfilledtimeout setting. Note: If you set unfilledtimeout.unit to "seconds", "internals.process_throttle_secs" must be inferior or equal to timeout [Strategy Override](#parameters-in-the-strategy). <br> *Defaults to `minutes`.* <br> **Datatype:** String
| `unfilledtimeout.unit` | Unit to use in unfilledtimeout setting. Note: If you set unfilledtimeout.unit to "seconds", "internals.process_throttle_secs" must be inferior or equal to timeout [Strategy Override](#parameters-in-the-strategy). <br> *Defaults to `"minutes"`.* <br> **Datatype:** String
| `unfilledtimeout.exit_timeout_count` | How many times can exit orders time out. Once this number of timeouts is reached, an emergency exit is triggered. 0 to disable and allow unlimited order cancels. [Strategy Override](#parameters-in-the-strategy).<br>*Defaults to `0`.* <br> **Datatype:** Integer
| | **Pricing**
| `entry_pricing.price_side` | Select the side of the spread the bot should look at to get the entry rate. [More information below](#buy-price-side).<br> *Defaults to `same`.* <br> **Datatype:** String (either `ask`, `bid`, `same` or `other`).
| `entry_pricing.price_side` | Select the side of the spread the bot should look at to get the entry rate. [More information below](#entry-price).<br> *Defaults to `"same"`.* <br> **Datatype:** String (either `ask`, `bid`, `same` or `other`).
| `entry_pricing.price_last_balance` | **Required.** Interpolate the bidding price. More information [below](#entry-price-without-orderbook-enabled).
| `entry_pricing.use_order_book` | Enable entering using the rates in [Order Book Entry](#entry-price-with-orderbook-enabled). <br> *Defaults to `True`.*<br> **Datatype:** Boolean
| `entry_pricing.use_order_book` | Enable entering using the rates in [Order Book Entry](#entry-price-with-orderbook-enabled). <br> *Defaults to `true`.*<br> **Datatype:** Boolean
| `entry_pricing.order_book_top` | Bot will use the top N rate in Order Book "price_side" to enter a trade. I.e. a value of 2 will allow the bot to pick the 2nd entry in [Order Book Entry](#entry-price-with-orderbook-enabled). <br>*Defaults to `1`.* <br> **Datatype:** Positive Integer
| `entry_pricing. check_depth_of_market.enabled` | Do not enter if the difference of buy orders and sell orders is met in Order Book. [Check market depth](#check-depth-of-market). <br>*Defaults to `false`.* <br> **Datatype:** Boolean
| `entry_pricing. check_depth_of_market.bids_to_ask_delta` | The difference ratio of buy orders and sell orders found in Order Book. A value below 1 means sell order size is greater, while value greater than 1 means buy order size is higher. [Check market depth](#check-depth-of-market) <br> *Defaults to `0`.* <br> **Datatype:** Float (as ratio)
| `exit_pricing.price_side` | Select the side of the spread the bot should look at to get the exit rate. [More information below](#exit-price-side).<br> *Defaults to `same`.* <br> **Datatype:** String (either `ask`, `bid`, `same` or `other`).
| `exit_pricing.price_side` | Select the side of the spread the bot should look at to get the exit rate. [More information below](#exit-price-side).<br> *Defaults to `"same"`.* <br> **Datatype:** String (either `ask`, `bid`, `same` or `other`).
| `exit_pricing.price_last_balance` | Interpolate the exiting price. More information [below](#exit-price-without-orderbook-enabled).
| `exit_pricing.use_order_book` | Enable exiting of open trades using [Order Book Exit](#exit-price-with-orderbook-enabled). <br> *Defaults to `True`.*<br> **Datatype:** Boolean
| `exit_pricing.use_order_book` | Enable exiting of open trades using [Order Book Exit](#exit-price-with-orderbook-enabled). <br> *Defaults to `true`.*<br> **Datatype:** Boolean
| `exit_pricing.order_book_top` | Bot will use the top N rate in Order Book "price_side" to exit. I.e. a value of 2 will allow the bot to pick the 2nd ask rate in [Order Book Exit](#exit-price-with-orderbook-enabled)<br>*Defaults to `1`.* <br> **Datatype:** Positive Integer
| `custom_price_max_distance_ratio` | Configure maximum distance ratio between current and custom entry or exit price. <br>*Defaults to `0.02` 2%).*<br> **Datatype:** Positive float
| | **TODO**
| `use_exit_signal` | Use exit signals produced by the strategy in addition to the `minimal_roi`. [Strategy Override](#parameters-in-the-strategy). <br>*Defaults to `true`.* <br> **Datatype:** Boolean
| `use_exit_signal` | Use exit signals produced by the strategy in addition to the `minimal_roi`. <br>Setting this to false disables the usage of `"exit_long"` and `"exit_short"` columns. Has no influence on other exit methods (Stoploss, ROI, callbacks). [Strategy Override](#parameters-in-the-strategy). <br>*Defaults to `true`.* <br> **Datatype:** Boolean
| `exit_profit_only` | Wait until the bot reaches `exit_profit_offset` before taking an exit decision. [Strategy Override](#parameters-in-the-strategy). <br>*Defaults to `false`.* <br> **Datatype:** Boolean
| `exit_profit_offset` | Exit-signal is only active above this value. Only active in combination with `exit_profit_only=True`. [Strategy Override](#parameters-in-the-strategy). <br>*Defaults to `0.0`.* <br> **Datatype:** Float (as ratio)
| `ignore_roi_if_entry_signal` | Do not exit if the entry signal is still active. This setting takes preference over `minimal_roi` and `use_exit_signal`. [Strategy Override](#parameters-in-the-strategy). <br>*Defaults to `false`.* <br> **Datatype:** Boolean
@@ -188,7 +188,6 @@ Mandatory parameters are marked as **Required**, which means that they are requi
| `max_entry_position_adjustment` | Maximum additional order(s) for each open trade on top of the first entry Order. Set it to `-1` for unlimited additional orders. [More information here](strategy-callbacks.md#adjust-trade-position). <br> [Strategy Override](#parameters-in-the-strategy). <br>*Defaults to `-1`.*<br> **Datatype:** Positive Integer or -1
| | **Exchange**
| `exchange.name` | **Required.** Name of the exchange class to use. [List below](#user-content-what-values-for-exchangename). <br> **Datatype:** String
| `exchange.sandbox` | Use the 'sandbox' version of the exchange, where the exchange provides a sandbox for risk-free integration. See [here](sandbox-testing.md) in more details.<br> **Datatype:** Boolean
| `exchange.key` | API key to use for the exchange. Only required when you are in production mode.<br>**Keep it in secret, do not disclose publicly.** <br> **Datatype:** String
| `exchange.secret` | API secret to use for the exchange. Only required when you are in production mode.<br>**Keep it in secret, do not disclose publicly.** <br> **Datatype:** String
| `exchange.password` | API password to use for the exchange. Only required when you are in production mode and for exchanges that use password for API requests.<br>**Keep it in secret, do not disclose publicly.** <br> **Datatype:** String
@@ -199,10 +198,10 @@ Mandatory parameters are marked as **Required**, which means that they are requi
| `exchange.ccxt_sync_config` | Additional CCXT parameters passed to the regular (sync) ccxt instance. Parameters may differ from exchange to exchange and are documented in the [ccxt documentation](https://ccxt.readthedocs.io/en/latest/manual.html#instantiation) <br> **Datatype:** Dict
| `exchange.ccxt_async_config` | Additional CCXT parameters passed to the async ccxt instance. Parameters may differ from exchange to exchange and are documented in the [ccxt documentation](https://ccxt.readthedocs.io/en/latest/manual.html#instantiation) <br> **Datatype:** Dict
| `exchange.markets_refresh_interval` | The interval in minutes in which markets are reloaded. <br>*Defaults to `60` minutes.* <br> **Datatype:** Positive Integer
| `exchange.skip_pair_validation` | Skip pairlist validation on startup.<br>*Defaults to `false`<br> **Datatype:** Boolean
| `exchange.skip_open_order_update` | Skips open order updates on startup should the exchange cause problems. Only relevant in live conditions.<br>*Defaults to `false`<br> **Datatype:** Boolean
| `exchange.skip_pair_validation` | Skip pairlist validation on startup.<br>*Defaults to `false`*<br> **Datatype:** Boolean
| `exchange.skip_open_order_update` | Skips open order updates on startup should the exchange cause problems. Only relevant in live conditions.<br>*Defaults to `false`*<br> **Datatype:** Boolean
| `exchange.unknown_fee_rate` | Fallback value to use when calculating trading fees. This can be useful for exchanges which have fees in non-tradable currencies. The value provided here will be multiplied with the "fee cost".<br>*Defaults to `None`<br> **Datatype:** float
| `exchange.log_responses` | Log relevant exchange responses. For debug mode only - use with care.<br>*Defaults to `false`<br> **Datatype:** Boolean
| `exchange.log_responses` | Log relevant exchange responses. For debug mode only - use with care.<br>*Defaults to `false`*<br> **Datatype:** Boolean
| `experimental.block_bad_exchanges` | Block exchanges known to not work with freqtrade. Leave on default unless you want to test if that exchange works now. <br>*Defaults to `true`.* <br> **Datatype:** Boolean
| | **Plugins**
| `edge.*` | Please refer to [edge configuration document](edge.md) for detailed explanation of all possible configuration options.
@@ -213,7 +212,7 @@ Mandatory parameters are marked as **Required**, which means that they are requi
| `telegram.token` | Your Telegram bot token. Only required if `telegram.enabled` is `true`. <br>**Keep it in secret, do not disclose publicly.** <br> **Datatype:** String
| `telegram.chat_id` | Your personal Telegram account id. Only required if `telegram.enabled` is `true`. <br>**Keep it in secret, do not disclose publicly.** <br> **Datatype:** String
| `telegram.balance_dust_level` | Dust-level (in stake currency) - currencies with a balance below this will not be shown by `/balance`. <br> **Datatype:** float
| `telegram.reload` | Allow "reload" buttons on telegram messages. <br>*Defaults to `True`.<br> **Datatype:** boolean
| `telegram.reload` | Allow "reload" buttons on telegram messages. <br>*Defaults to `true`.<br> **Datatype:** boolean
| `telegram.notification_settings.*` | Detailed notification settings. Refer to the [telegram documentation](telegram-usage.md) for details.<br> **Datatype:** dictionary
| `telegram.allow_custom_messages` | Enable the sending of Telegram messages from strategies via the dataprovider.send_msg() function. <br> **Datatype:** Boolean
| | **Webhook**
@@ -251,8 +250,8 @@ Mandatory parameters are marked as **Required**, which means that they are requi
| `db_url` | Declares database URL to use. NOTE: This defaults to `sqlite:///tradesv3.dryrun.sqlite` if `dry_run` is `true`, and to `sqlite:///tradesv3.sqlite` for production instances. <br> **Datatype:** String, SQLAlchemy connect string
| `logfile` | Specifies logfile name. Uses a rolling strategy for log file rotation for 10 files with the 1MB limit per file. <br> **Datatype:** String
| `add_config_files` | Additional config files. These files will be loaded and merged with the current config file. The files are resolved relative to the initial file.<br> *Defaults to `[]`*. <br> **Datatype:** List of strings
| `dataformat_ohlcv` | Data format to use to store historical candle (OHLCV) data. <br> *Defaults to `json`*. <br> **Datatype:** String
| `dataformat_trades` | Data format to use to store historical trades data. <br> *Defaults to `jsongz`*. <br> **Datatype:** String
| `dataformat_ohlcv` | Data format to use to store historical candle (OHLCV) data. <br> *Defaults to `feather`*. <br> **Datatype:** String
| `dataformat_trades` | Data format to use to store historical trades data. <br> *Defaults to `feather`*. <br> **Datatype:** String
| `reduce_df_footprint` | Recast all numeric columns to float32/int32, with the objective of reducing ram/disk usage (and decreasing train/inference timing in FreqAI). (Currently only affects FreqAI use-cases) <br> **Datatype:** Boolean. <br> Default: `False`.
### Parameters in the strategy
@@ -614,6 +613,7 @@ Once you will be happy with your bot performance running in the Dry-run mode, yo
* Orders are simulated, and will not be posted to the exchange.
* Market orders fill based on orderbook volume the moment the order is placed.
* Limit orders fill once the price reaches the defined level - or time out based on `unfilledtimeout` settings.
* Limit orders will be converted to market orders if they cross the price by more than 1%.
* In combination with `stoploss_on_exchange`, the stop_loss price is assumed to be filled.
* Open orders (not trades, which are stored in the database) are kept open after bot restarts, with the assumption that they were not filled while being offline.
@@ -682,16 +682,14 @@ To use a proxy for exchange connections - you will have to define the proxies as
{
"exchange": {
"ccxt_config": {
"aiohttp_proxy": "http://addr:port",
"proxies": {
"http": "http://addr:port",
"https": "http://addr:port"
},
"httpsProxy": "http://addr:port",
}
}
}
```
For more information on available proxy types, please consult the [ccxt proxy documentation](https://docs.ccxt.com/#/README?id=proxy).
## Next step
Now you have configured your config.json, the next step is to [start your bot](bot-usage.md).
@@ -6,7 +6,7 @@ To download data (candles / OHLCV) needed for backtesting and hyperoptimization
If no additional parameter is specified, freqtrade will download data for `"1m"` and `"5m"` timeframes for the last 30 days.
Exchange and pairs will come from `config.json` (if specified using `-c/--config`).
Otherwise`--exchange` becomes mandatory.
Without provided configuration,`--exchange` becomes mandatory.
You can use a relative timerange (`--days 20`) or an absolute starting point (`--timerange 20200101-`). For incremental downloads, the relative approach should be used.
--prepend Allow data prepending. (Data-appending is disabled)
Common arguments:
-v, --verbose Verbose mode (-vv for more, -vvv to get all messages).
--logfile FILE Log to the file specified. Special values are:
--logfile FILE, --log-file FILE
Log to the file specified. Special values are:
'syslog', 'journald'. See the documentation for more
details.
-V, --version show program's version number and exit
@@ -83,40 +83,47 @@ Common arguments:
```
!!! Tip "Downloading all data for one quote currency"
Often, you'll want to download data for all pairs of a specific quote-currency. In such cases, you can use the following shorthand:
`freqtrade download-data --exchange binance --pairs .*/USDT <...>`. The provided "pairs" string will be expanded to contain all active pairs on the exchange.
To also download data for inactive (delisted) pairs, add `--include-inactive-pairs` to the command.
!!! Note "Startup period"
`download-data` is a strategy-independent command. The idea is to download a big chunk of data once, and then iteratively increase the amount of data stored.
For that reason, `download-data` does not care about the "startup-period" defined in a strategy. It's up to the user to download additional days if the backtest should start at a specific point in time (while respecting startup period).
### Pairs file
### Start download
In alternative to the whitelist from `config.json`, a `pairs.json` file can be used.
If you are using Binance for example:
- create a directory `user_data/data/binance` and copy or create the `pairs.json` file in that directory.
- update the `pairs.json` file to contain the currency pairs you are interested in.
A very simple command (assuming an available `config.json` file) can look as follows.
```bash
mkdir -p user_data/data/binance
touch user_data/data/binance/pairs.json
freqtrade download-data --exchange binance
```
The format of the `pairs.json` file is a simple json list.
Mixing different stake-currencies is allowed for this file, since it's only used for downloading.
This will download historical candle (OHLCV) data for all the currency pairs defined in the configuration.
!!! Tip "Downloading all data for one quote currency"
Often, you'll want to download data for all pairs of a specific quote-currency. In such cases, you can use the following shorthand:
`freqtrade download-data --exchange binance --pairs .*/USDT <...>`. The provided "pairs" string will be expanded to contain all active pairs on the exchange.
To also downloaddata for inactive (delisted) pairs, add `--include-inactive-pairs` to the command.
or as regex (in this case, to download all active USDT pairs)
* To use a different directory than the exchange specific default, use `--datadir user_data/data/some_directory`.
* To change the exchange used to download the historical data from, please use a different configuration file (you'll probably need to adjust rate limits etc.)
* To use `pairs.json` from some other directory, use `--pairs-file some_other_dir/pairs.json`.
* To download historical candle (OHLCV) data for only 10 days, use `--days 10` (defaults to 30 days).
* To download historical candle (OHLCV) data from a fixed starting point, use `--timerange 20200101-` - which will download all data from January 1st, 2020.
* Use `--timeframes` to specify what timeframe download the historical candle (OHLCV) data for. Default is `--timeframes 1m 5m` which will download 1-minute and 5-minute data.
* To use exchange, timeframe and list of pairs as defined in your configuration file, use the `-c/--config` option. With this, the script uses the whitelist defined in the config as the list of currency pairs to download data for and does not require the pairs.json file. You can combine `-c/--config` with most other options.
??? Note "Permission denied errors"
If your configuration directory `user_data` was made by docker, you may get the following error:
@@ -131,39 +138,7 @@ Mixing different stake-currencies is allowed for this file, since it's only used
sudo chown -R $UID:$GID user_data
```
### Start download
Then run:
```bash
freqtrade download-data --exchange binance
```
This will download historical candle (OHLCV) data for all the currency pairs you defined in `pairs.json`.
- To use a different directory than the exchange specific default, use `--datadir user_data/data/some_directory`.
- To change the exchange used to download the historical data from, please use a different configuration file (you'll probably need to adjust rate limits etc.)
- To use `pairs.json` from some other directory, use `--pairs-file some_other_dir/pairs.json`.
- To download historical candle (OHLCV) data for only 10 days, use `--days 10` (defaults to 30 days).
- To download historical candle (OHLCV) data from a fixed starting point, use `--timerange 20200101-` - which will download all data from January 1st, 2020.
- Use `--timeframes` to specify what timeframe download the historical candle (OHLCV) data for. Default is `--timeframes 1m 5m` which will download 1-minute and 5-minute data.
- To use exchange, timeframe and list of pairs as defined in your configuration file, use the `-c/--config` option. With this, the script uses the whitelist defined in the config as the list of currency pairs to download data for and does not require the pairs.json file. You can combine `-c/--config` with most other options.
#### Download additional data before the current timerange
### Download additional data before the current timerange
Assuming you downloaded all data from 2022 (`--timerange 20220101-`) - but you'd now like to also backtest with earlier data.
You can do so by using the `--prepend` flag, combined with `--timerange` - specifying an end-date.
Freqtrade currently supports the following data-formats:
* `feather` - a dataformat based on Apache Arrow
* `json` - plain "text" json files
* `jsongz` - a gzip-zipped version of json files
* `hdf5` - a high performance datastore
* `feather` - a dataformat based on Apache Arrow (OHLCV only)
* `parquet` - columnar datastore (OHLCV only)
By default, OHLCV data is stored as `json` data, while trades data is stored as `jsongz` data.
By default, both OHLCV data and trades data are stored in the `feather` format.
This can be changed via the `--data-format-ohlcv` and `--data-format-trades` command line arguments respectively.
To persist this change, you should also add the following snippet to your configuration, so you don't have to insert the above arguments each time:
@@ -228,17 +203,46 @@ time freqtrade list-data --show-timerange --data-format-ohlcv <dataformat>
| Format | Size | timing |
|------------|-------------|-------------|
| `feather` | 72Mb | 3.5s |
| `json` | 149Mb | 25.6s |
| `jsongz` | 39Mb | 27s |
| `hdf5` | 145Mb | 3.9s |
| `feather` | 72Mb | 3.5s |
| `parquet` | 83Mb | 3.8s |
Size has been taken from the BTC/USDT 1m spot combination for the timerange specified above.
To have a best performance/size mix, we recommend the use of either feather or parquet.
To have a best performance/size mix, we recommend using the default feather format, or parquet.
#### Sub-command convert data
### Pairs file
In alternative to the whitelist from `config.json`, a `pairs.json` file can be used.
If you are using Binance for example:
* create a directory `user_data/data/binance` and copy or create the `pairs.json` file in that directory.
* update the `pairs.json` file to contain the currency pairs you are interested in.
```bash
mkdir -p user_data/data/binance
touch user_data/data/binance/pairs.json
```
The format of the `pairs.json` file is a simple json list.
Mixing different stake-currencies is allowed for this file, since it's only used for downloading.
``` json
[
"ETH/BTC",
"ETH/USDT",
"BTC/USDT",
"XRP/ETH"
]
```
!!! Note
The `pairs.json` file is only used when no configuration is loaded (implicitly by naming, or via `--config` flag).
You can force the usage of this file via `--pairs-file pairs.json` - however we recommend to use the pairlist from within the configuration, either via `exchange.pair_whitelist` or `pairs` setting in the configuration.
-v, --verbose Verbose mode (-vv for more, -vvv to get all messages).
--logfile FILE Log to the file specified. Special values are:
--logfile FILE, --log-file FILE
Log to the file specified. Special values are:
'syslog', 'journald'. See the documentation for more
details.
-V, --version show program's version number and exit
@@ -287,10 +292,9 @@ Common arguments:
Path to directory with historical backtesting data.
--userdir PATH, --user-data-dir PATH
Path to userdata directory.
```
##### Example converting data
### Example converting data
The following command will convert all candle (OHLCV) data available in `~/.freqtrade/data/binance` from json to jsongz, saving diskspace in the process.
It'll also remove original json data files (`--erase` parameter).
@@ -299,7 +303,7 @@ It'll also remove original json data files (`--erase` parameter).
By default, `download-data` sub-command downloads Candles (OHLCV) data. Some exchanges also provide historic trade-data via their API.
This data can be useful if you need many different timeframes, since it is only downloaded once, and then resampled locally to the desired timeframes.
Since this data is large by default, the files use gzip by default. They are stored in your data-directory with the naming convention of `<pair>-trades.json.gz` (`ETH_BTC-trades.json.gz`). Incremental mode is also supported, as for historic OHLCV data, so downloading the data once per week with `--days 8` will create an incremental data-repository.
Since this data is large by default, the files use the feather fileformat by default. They are stored in your data-directory with the naming convention of `<pair>-trades.feather` (`ETH_BTC-trades.feather`). Incremental mode is also supported, as for historic OHLCV data, so downloading the data once per week with `--days 8` will create an incremental data-repository.
To use this mode, simply add `--dl-trades` to your call. This will swap the download method to download trades, and resamples the data locally.
To debug freqtrade, we recommend VSCode with the following launch configuration (located in `.vscode/launch.json`).
To debug freqtrade, we recommend VSCode (with the Python extension) with the following launch configuration (located in `.vscode/launch.json`).
Details will obviously vary between setups - but this should work to get you started.
``` json
@@ -102,6 +102,19 @@ This method can also be used to debug a strategy, by setting the breakpoints wit
A similar setup can also be taken for Pycharm - using `freqtrade` as module name, and setting the command line arguments as "parameters".
??? Tip "Correct venv usage"
When using a virtual environment (which you should), make sure that your Editor is using the correct virtual environment to avoid problems or "unknown import" errors.
#### Vscode
You can select the correct environment in VSCode with the command "Python: Select Interpreter" - which will show you environments the extension detected.
If your environment has not been detected, you can also pick a path manually.
#### Pycharm
In pycharm, you can select the appropriate Environment in the "Run/Debug Configurations" window.
This assumes that you have the repository checked out, and the editor is started at the repository root level (so setup.py is at the top level of your repository).
@@ -327,18 +340,18 @@ To check how the new exchange behaves, you can use the following snippet:
``` python
import ccxt
from datetime import datetime
from datetime import datetime, timezone
from freqtrade.data.converter import ohlcv_to_dataframe
ct = ccxt.binance()
ct = ccxt.binance() # Use the exchange you're testing
timeframe = "1d"
pair = "XLM/BTC" # Make sure to use a pair that exists on that exchange!
pair = "BTC/USDT" # Make sure to use a pair that exists on that exchange!
@@ -14,6 +14,9 @@ Start by downloading and installing Docker / Docker Desktop for your platform:
Freqtrade documentation assumes the use of Docker desktop (or the docker compose plugin).
While the docker-compose standalone installation still works, it will require changing all `docker compose` commands from `docker compose` to `docker-compose` to work (e.g. `docker compose up -d` will become `docker-compose up -d`).
??? Warning "Docker on windows"
If you just installed docker on a windows system, make sure to reboot your system, otherwise you might encounter unexplainable Problems related to network connectivity to docker containers.
## Freqtrade with docker
Freqtrade provides an official Docker image on [Dockerhub](https://hub.docker.com/r/freqtradeorg/freqtrade/), as well as a [docker compose file](https://github.com/freqtrade/freqtrade/blob/stable/docker-compose.yml) ready for usage.
@@ -78,7 +81,7 @@ If you've selected to enable FreqUI in the `new-config` step, you will have freq
You can now access the UI by typing localhost:8080 in your browser.
??? Note "UI Access on a remote servers"
??? Note "UI Access on a remote server"
If you're running on a VPS, you should consider using either a ssh tunnel, or setup a VPN (openVPN, wireguard) to connect to your bot.
This will ensure that freqUI is not directly exposed to the internet, which is not recommended for security reasons (freqUI does not support https out of the box).
Setup of these tools is not part of this tutorial, however many good tutorials can be found on the internet.
@@ -128,7 +131,7 @@ All freqtrade arguments will be available by running `docker compose run --rm fr
!!! Note "`docker compose run --rm`"
Including `--rm` will remove the container after completion, and is highly recommended for all modes except trading mode (running with `freqtrade trade` command).
??? Note "Using docker without docker"
??? Note "Using docker without docker compose"
"`docker compose run --rm`" will require a compose file to be provided.
Some freqtrade commands that don't require authentication such as `list-pairs` can be run with "`docker run --rm`" instead.
For example `docker run --rm freqtradeorg/freqtrade:stable list-pairs --exchange binance --quote BTC --print-json`.
@@ -172,7 +175,7 @@ You can then run `docker compose build --pull` to build the docker image, and ru
### Plotting with docker
Commands `freqtrade plot-profit` and `freqtrade plot-dataframe` ([Documentation](plotting.md)) are available by changing the image to `*_plot` in your docker-compose.yml file.
Commands `freqtrade plot-profit` and `freqtrade plot-dataframe` ([Documentation](plotting.md)) are available by changing the image to `*_plot` in your `docker-compose.yml` file.
If you're on windows and just installed Docker (desktop), make sure to reboot your System. Docker can have problems with network connectivity without a restart.
You should obviously also make sure to have your [settings](#accessing-the-ui) accordingly.
!!! Warning
Due to the above, we do not recommend the usage of docker on windows for production setups, but only for experimentation, datadownload and backtesting.
The `Edge Positioning` module uses probability to calculate your win rate and risk reward ratio. It will use these statistics to control your strategy trade entry points, position size and, stoploss.
!!! Danger "Deprecated functionality"
`Edge positioning` (or short Edge) is currently in maintenance mode only (we keep existing functionality alive) and should be considered as deprecated.
It will currently not receive new features until either someone stepped forward to take up ownership of that module - or we'll decide to remove edge from freqtrade.
!!! Warning
When using `Edge positioning` with a dynamic whitelist (VolumePairList), make sure to also use `AgeFilter` and set it to at least `calculate_since_number_of_days` to avoid problems with missing data.
@@ -55,7 +55,7 @@ This configuration enables kraken, as well as rate-limiting to avoid bans from t
## Binance
!!! Warning "Server location and geo-ip restrictions"
Please be aware that binance restrict api access regarding the server country. The currents and nonexhaustive countries blocked are United States, Malaysia (Singapour), Ontario (Canada). Please go to [binance terms > b. Eligibility](https://www.binance.com/en/terms) to find up to date list.
Please be aware that Binance restricts API access regarding the server country. The current and non-exhaustive countries blocked are Canada, Malaysia, Netherlands and United States. Please go to [binance terms > b. Eligibility](https://www.binance.com/en/terms) to find up to date list.
@@ -136,13 +136,41 @@ Freqtrade will not attempt to change these settings.
The Kraken API does only provide 720 historic candles, which is sufficient for Freqtrade dry-run and live trade modes, but is a problem for backtesting.
To download data for the Kraken exchange, using `--dl-trades` is mandatory, otherwise the bot will download the same 720 candles over and over, and you'll not have enough backtest data.
Due to the heavy rate-limiting applied by Kraken, the following configuration section should be used to download data:
To speed up downloading, you can download the [trades zip files](https://support.kraken.com/hc/en-us/articles/360047543791-Downloadable-historical-market-data-time-and-sales-) kraken provides.
These are usually updated once per quarter. Freqtrade expects these files to be placed in `user_data/data/kraken/trades_csv`.
``` json
"ccxt_async_config": {
"enableRateLimit": true,
"rateLimit": 3100
},
A structure as follows can make sense if using incremental files, with the "full" history in one directory, and incremental files in different directories.
The assumption for this mode is that the data is downloaded and unzipped keeping filenames as they are.
Duplicate content will be ignored (based on timestamp) - though the assumption is that there is no gap in the data.
This means, if your "full" history ends in Q4 2022 - then both incremental updates Q1 2023 and Q2 2023 are available.
Not having this will lead to incomplete data, and therefore invalid results while using the data.
@@ -259,10 +287,17 @@ The configuration parameter `exchange.unknown_fee_rate` can be used to specify t
Futures trading on bybit is currently supported for USDT markets, and will use isolated futures mode.
Users with unified accounts (there's no way back) can create a Sub-account which will start as "non-unified", and can therefore use isolated futures.
On startup, freqtrade will set the position mode to "One-way Mode" for the whole (sub)account. This avoids making this call over and over again (slowing down bot operations), but means that changes to this setting may result in exceptions and errors.
On startup, freqtrade will set the position mode to "One-way Mode" for the whole (sub)account. This avoids making this call over and over again (slowing down bot operations), but means that changes to this setting may result in exceptions and errors
As bybit doesn't provide funding rate history, the dry-run calculation is used for live trades as well.
API Keys for live futures trading (Subaccount on non-unified) must have the following permissions:
* Read-write
* Contract - Orders
* Contract - Positions
We do strongly recommend to limit all API keys to the IP you're going to use it from.
!!! Tip "Stoploss on Exchange"
Bybit (futures only) supports `stoploss_on_exchange` and uses `stop-loss-limit` orders. It provides great advantages, so we recommend to benefit from it by enabling stoploss on exchange.
On futures, Bybit supports both `stop-limit` as well as `stop-market` orders. You can use either `"limit"` or `"market"` in the `order_types.stoploss` configuration setting to decide which type to use.
@@ -20,7 +20,7 @@ Futures trading is supported for selected exchanges. Please refer to the [docume
* When you work with your strategy & hyperopt file you should use a proper code editor like VSCode or PyCharm. A good code editor will provide syntax highlighting as well as line numbers, making it easy to find syntax errors (most likely pointed out by Freqtrade during startup).
## Freqtrade common issues
## Freqtrade common questions
### Can freqtrade open multiple positions on the same pair in parallel?
@@ -36,7 +36,7 @@ Running the bot with `freqtrade trade --config config.json` shows the output `fr
This could be caused by the following reasons:
* The virtual environment is not active.
* Run `source .env/bin/activate` to activate the virtual environment.
* Run `source .venv/bin/activate` to activate the virtual environment.
* The installation did not complete successfully.
* Please check the [Installation documentation](installation.md).
@@ -78,6 +78,14 @@ Where possible (e.g. on binance), the use of the exchange's dedicated fee curren
On binance, it's sufficient to have BNB in your account, and have "Pay fees in BNB" enabled in your profile. Your BNB balance will slowly decline (as it's used to pay fees) - but you'll no longer encounter dust (Freqtrade will include the fees in the profit calculations).
Other exchanges don't offer such possibilities, where it's simply something you'll have to accept or move to a different exchange.
### I deposited more funds to the exchange, but my bot doesn't recognize this
Freqtrade will update the exchange balance when necessary (Before placing an order).
RPC calls (Telegram's `/balance`, API calls to `/balance`) can trigger an update at max. once per hour.
If `adjust_trade_position` is enabled (and the bot has open trades eligible for position adjustments) - then the wallets will be refreshed once per hour.
To force an immediate update, you can use `/reload_config` - which will restart the bot.
### I want to use incomplete candles
Freqtrade will not provide incomplete candles to strategies. Using incomplete candles will lead to repainting and consequently to strategies with "ghost" buys, which are impossible to both backtest, and verify after they happened.
@@ -142,6 +150,13 @@ To fix this, redefine order types in the strategy to use "limit" instead of "mar
The same fix should be applied in the configuration file, if order types are defined in your custom config rather than in the strategy.
### I'm trying to start the bot live, but get an API permission error
Errors like `Invalid API-key, IP, or permissions for action` mean exactly what they actually say.
Your API key is either invalid (copy/paste error? check for leading/trailing spaces in the config), expired, or the IP you're running the bot from is not enabled in the Exchange's API console.
Usually, the permission "Spot Trading" (or the equivalent in the exchange you use) will be necessary.
Futures will usually have to be enabled specifically.
### How do I search the bot logs for something?
By default, the bot writes its log into stderr stream. This is implemented this way so that you can easily separate the bot's diagnostics messages from Backtesting, Edge and Hyperopt results, output from other various Freqtrade utility sub-commands, as well as from the output of your custom `print()`'s you may have inserted into your strategy. So if you need to search the log messages with the grep utility, you need to redirect stderr to stdout and disregard stdout.
Required function to set the targets for the model.
@@ -139,6 +139,7 @@ The FreqAI strategy requires including the following lines of code in the standa
/dataframe["close"]
-1
)
returndataframe
```
Notice how the `feature_engineering_*()` is where [features](freqai-feature-engineering.md#feature-engineering) are added. Meanwhile `set_freqai_targets()` adds the labels/targets. A full example strategy is available in `templates/FreqaiExampleStrategy.py`.
@@ -159,7 +160,7 @@ Below are the values you can expect to include/use inside a typical strategy dat
|------------|-------------|
| `df['&*']` | Any dataframe column prepended with `&` in `set_freqai_targets()` is treated as a training target (label) inside FreqAI (typically following the naming convention `&-s*`). For example, to predict the close price 40 candles into the future, you would set `df['&-s_close'] = df['close'].shift(-self.freqai_info["feature_parameters"]["label_period_candles"])` with `"label_period_candles": 40` in the config. FreqAI makes the predictions and gives them back under the same key (`df['&-s_close']`) to be used in `populate_entry/exit_trend()`. <br>**Datatype:** Depends on the output of the model.
| `df['&*_std/mean']` | Standard deviation and mean values of the defined labels during training (or live tracking with `fit_live_predictions_candles`). Commonly used to understand the rarity of a prediction (use the z-score as shown in `templates/FreqaiExampleStrategy.py` and explained [here](#creating-a-dynamic-target-threshold) to evaluate how often a particular prediction was observed during training or historically with `fit_live_predictions_candles`). <br>**Datatype:** Float.
| `df['do_predict']` | Indication of an outlier data point. The return value is integer between -2 and 2, which lets you know if the prediction is trustworthy or not. `do_predict==1` means that the prediction is trustworthy. If the Dissimilarity Index (DI, see details [here](freqai-feature-engineering.md#identifying-outliers-with-the-dissimilarity-index-di)) of the input data point is above the threshold defined in the config, FreqAI will subtract 1 from `do_predict`, resulting in `do_predict==0`. If `use_SVM_to_remove_outliers()` is active, the Support Vector Machine (SVM, see details [here](freqai-feature-engineering.md#identifying-outliers-using-a-support-vector-machine-svm)) may also detect outliers in training and prediction data. In this case, the SVM will also subtract 1 from `do_predict`. If the input data point was considered an outlier by the SVM but not by the DI, or vice versa, the result will be `do_predict==0`. If both the DI and the SVM considers the input data point to be an outlier, the result will be `do_predict==-1`. As with the SVM, if `use_DBSCAN_to_remove_outliers` is active, DBSCAN (see details [here](freqai-feature-engineering.md#identifying-outliers-with-dbscan)) may also detect outliers and subtract 1 from `do_predict`. Hence, if both the SVM and DBSCAN are active and identify a datapoint that was above the DI threshold as an outlier, the result will be `do_predict==-2`. A particular case is when `do_predict == 2`, which means that the model has expired due to exceeding `expired_hours`. <br>**Datatype:** Integer between -2 and 2.
| `df['do_predict']` | Indication of an outlier data point. The return value is integer between -2 and 2, which lets you know if the prediction is trustworthy or not. `do_predict==1` means that the prediction is trustworthy. If the Dissimilarity Index (DI, see details [here](freqai-feature-engineering.md#identifying-outliers-with-the-dissimilarity-index-di)) of the input data point is above the threshold defined in the config, FreqAI will subtract 1 from `do_predict`, resulting in `do_predict==0`. If `use_SVM_to_remove_outliers` is active, the Support Vector Machine (SVM, see details [here](freqai-feature-engineering.md#identifying-outliers-using-a-support-vector-machine-svm)) may also detect outliers in training and prediction data. In this case, the SVM will also subtract 1 from `do_predict`. If the input data point was considered an outlier by the SVM but not by the DI, or vice versa, the result will be `do_predict==0`. If both the DI and the SVM considers the input data point to be an outlier, the result will be `do_predict==-1`. As with the SVM, if `use_DBSCAN_to_remove_outliers` is active, DBSCAN (see details [here](freqai-feature-engineering.md#identifying-outliers-with-dbscan)) may also detect outliers and subtract 1 from `do_predict`. Hence, if both the SVM and DBSCAN are active and identify a datapoint that was above the DI threshold as an outlier, the result will be `do_predict==-2`. A particular case is when `do_predict == 2`, which means that the model has expired due to exceeding `expired_hours`. <br>**Datatype:** Integer between -2 and 2.
| `df['DI_values']` | Dissimilarity Index (DI) values are proxies for the level of confidence FreqAI has in the prediction. A lower DI means the prediction is close to the training data, i.e., higher prediction confidence. See details about the DI [here](freqai-feature-engineering.md#identifying-outliers-with-the-dissimilarity-index-di). <br>**Datatype:** Float.
| `df['%*']` | Any dataframe column prepended with `%` in `feature_engineering_*()` is treated as a training feature. For example, you can include the RSI in the training feature set (similar to in `templates/FreqaiExampleStrategy.py`) by setting `df['%-rsi']`. See more details on how this is done [here](freqai-feature-engineering.md). <br>**Note:** Since the number of features prepended with `%` can multiply very quickly (10s of thousands of features are easily engineered using the multiplictative functionality of, e.g., `include_shifted_candles` and `include_timeframes` as described in the [parameter table](freqai-parameter-table.md)), these features are removed from the dataframe that is returned from FreqAI to the strategy. To keep a particular type of feature for plotting purposes, you would prepend it with `%%`. <br>**Datatype:** Depends on the output of the model.
@@ -236,3 +237,181 @@ If you want to predict multiple targets you must specify all labels in the same
The PyTorch module requires large packages such as `torch`, which should be explicitly requested during `./setup.sh -i` by answering "y" to the question "Do you also want dependencies for freqai-rl or PyTorch (~700mb additional space required) [y/N]?".
Users who prefer docker should ensure they use the docker image appended with `_freqaitorch`.
We do provide an explicit docker-compose file for this in `docker/docker-compose-freqai.yml` - which can be used via `docker compose -f docker/docker-compose-freqai.yml run ...` - or can be copied to replace the original docker file.
This docker-compose file also contains a (disabled) section to enable GPU resources within docker containers. This obviously assumes the system has GPU resources available.
### Structure
#### Model
You can construct your own Neural Network architecture in PyTorch by simply defining your `nn.Module` class inside your custom [`IFreqaiModel` file](#using-different-prediction-models) and then using that class in your `def train()` function. Here is an example of logistic regression model implementation using PyTorch (should be used with nn.BCELoss criterion) for classification tasks.
The `PyTorchModelTrainer` performs the idiomatic PyTorch train loop:
Define our model, loss function, and optimizer, and then move them to the appropriate device (GPU or CPU). Inside the loop, we iterate through the batches in the dataloader, move the data to the device, compute the prediction and loss, backpropagate, and update the model parameters using the optimizer.
In addition, the trainer is responsible for the following:
- saving and loading the model
- converting the data from `pandas.DataFrame` to `torch.Tensor`.
#### Integration with Freqai module
Like all freqai models, PyTorch models inherit `IFreqaiModel`. `IFreqaiModel` declares three abstract methods: `train`, `fit`, and `predict`. we implement these methods in three levels of hierarchy.
From top to bottom:
1. `BasePyTorchModel` - Implements the `train` method. all `BasePyTorch*` inherit it. responsible for general data preparation (e.g., data normalization) and calling the `fit` method. Sets `device` attribute used by children classes. Sets `model_type` attribute used by the parent class.
2. `BasePyTorch*` - Implements the `predict` method. Here, the `*` represents a group of algorithms, such as classifiers or regressors. responsible for data preprocessing, predicting, and postprocessing if needed.
3. `PyTorch*Classifier` / `PyTorch*Regressor` - implements the `fit` method. responsible for the main train flaw, where we initialize the trainer and model objects.
#### Full example
Building a PyTorch regressor using MLP (multilayer perceptron) model, MSELoss criterion, and AdamW optimizer.
Here we create a `PyTorchMLPRegressor` class that implements the `fit` method. The `fit` method specifies the training building blocks: model, optimizer, criterion, and trainer. We inherit both `BasePyTorchRegressor` and `BasePyTorchModel`, where the former implements the `predict` method that is suitable for our regression task, and the latter implements the train method.
??? Note "Setting Class Names for Classifiers"
When using classifiers, the user must declare the class names (or targets) by overriding the `IFreqaiModel.class_names` attribute. This is achieved by setting `self.freqai.class_names` in the FreqAI strategy inside the `set_freqai_targets` method.
For example, if you are using a binary classifier to predict price movements as up or down, you can set the class names as follows:
To see a full example, you can refer to the [classifier test strategy class](https://github.com/freqtrade/freqtrade/blob/develop/tests/strategy/strats/freqai_test_classifier.py).
#### Improving performance with `torch.compile()`
Torch provides a `torch.compile()` method that can be used to improve performance for specific GPU hardware. More details can be found [here](https://pytorch.org/tutorials/intermediate/torch_compile_tutorial.html). In brief, you simply wrap your `model` in `torch.compile()`:
```python
model = PyTorchMLPModel(
input_dim=n_features,
output_dim=1,
**self.model_kwargs
)
model.to(self.device)
model = torch.compile(model)
```
Then proceed to use the model as normal. Keep in mind that doing this will remove eager execution, which means errors and tracebacks will not be informative.
@@ -6,8 +6,8 @@ Low level feature engineering is performed in the user strategy within a set of
| Function | Description |
|---------------|-------------|
| `feature_engineering__expand_all()` | This optional function will automatically expand the defined features on the config defined `indicator_periods_candles`, `include_timeframes`, `include_shifted_candles`, and `include_corr_pairs`.
| `feature_engineering__expand_basic()` | This optional function will automatically expand the defined features on the config defined `include_timeframes`, `include_shifted_candles`, and `include_corr_pairs`. Note: this function does *not* expand across `include_periods_candles`.
| `feature_engineering_expand_all()` | This optional function will automatically expand the defined features on the config defined `indicator_periods_candles`, `include_timeframes`, `include_shifted_candles`, and `include_corr_pairs`.
| `feature_engineering_expand_basic()` | This optional function will automatically expand the defined features on the config defined `include_timeframes`, `include_shifted_candles`, and `include_corr_pairs`. Note: this function does *not* expand across `include_periods_candles`.
| `feature_engineering_standard()` | This optional function will be called once with the dataframe of the base timeframe. This is the final function to be called, which means that the dataframe entering this function will contain all the features and columns from the base asset created by the other `feature_engineering_expand` functions. This function is a good place to do custom exotic feature extractions (e.g. tsfresh). This function is also a good place for any feature that should not be auto-expanded upon (e.g., day of the week).
| `set_freqai_targets()` | Required function to set the targets for the model. All targets must be prepended with `&` to be recognized by the FreqAI internals.
@@ -16,7 +16,7 @@ Meanwhile, high level feature engineering is handled within `"feature_parameters
It is advisable to start from the template `feature_engineering_*` functions in the source provided example strategy (found in `templates/FreqaiExampleStrategy.py`) to ensure that the feature definitions are following the correct conventions. Here is an example of how to set the indicators and labels in the strategy:
Required function to set the targets for the model.
@@ -178,18 +178,20 @@ You can ask for each of the defined features to be included also for informative
`include_shifted_candles` indicates the number of previous candles to include in the feature set. For example, `include_shifted_candles: 2` tells FreqAI to include the past 2 candles for each of the features in the feature set.
In total, the number of features the user of the presented example strat has created is: length of `include_timeframes`* no. features in `feature_engineering_expand_*()` * length of `include_corr_pairlist` * no. `include_shifted_candles` * length of `indicator_periods_candles`
In total, the number of features the user of the presented example strategy has created is: length of `include_timeframes`* no. features in `feature_engineering_expand_*()` * length of `include_corr_pairlist` * no. `include_shifted_candles` * length of `indicator_periods_candles`
$= 3 * 3 * 3 * 2 * 2 = 108$.
!!! note "Learn more about creative feature engineering"
Check out our [medium article](https://emergentmethods.medium.com/freqai-from-price-to-prediction-6fadac18b665) geared toward helping users learn how to creatively engineer features.
### Gain finer control over `feature_engineering_*` functions with `metadata`
### Gain finer control over `feature_engineering_*` functions with `metadata`
All `feature_engineering_*` and `set_freqai_targets()` functions are passed a `metadata` dictionary which contains information about the `pair`, `tf` (timeframe), and `period` that FreqAI is automating for feature building. As such, a user can use `metadata` inside `feature_engineering_*` functions as criteria for blocking/reserving features for certain timeframes, periods, pairs etc.
All `feature_engineering_*` and `set_freqai_targets()` functions are passed a `metadata` dictionary which contains information about the `pair`, `tf` (timeframe), and `period` that FreqAI is automating for feature building. As such, a user can use `metadata` inside `feature_engineering_*` functions as criteria for blocking/reserving features for certain timeframes, periods, pairs etc.
This will block `ta.ROC()` from being added to any timeframes other than `"1h"`.
@@ -210,41 +212,7 @@ Another example, where the user wants to use live metrics from the trade databas
You need to set the standard dictionary in the config so that FreqAI can return proper dataframe shapes. These values will likely be overridden by the prediction model, but in the case where the model has yet to set them, or needs a default initial value, the pre-set values are what will be returned.
## Feature normalization
FreqAI is strict when it comes to data normalization. The train features, $X^{train}$, are always normalized to [-1, 1] using a shifted min-max normalization:
All other data (test data and unseen prediction data in dry/live/backtest) is always automatically normalized to the training feature space according to industry standards. FreqAI stores all the metadata required to ensure that test and prediction features will be properly normalized and that predictions are properly denormalized. For this reason, it is not recommended to eschew industry standards and modify FreqAI internals - however - advanced users can do so by inheriting `train()` in their custom `IFreqaiModel` and using their own normalization functions.
## Data dimensionality reduction with Principal Component Analysis
You can reduce the dimensionality of your features by activating the `principal_component_analysis` in the config:
```json
"freqai": {
"feature_parameters" : {
"principal_component_analysis": true
}
}
```
This will perform PCA on the features and reduce their dimensionality so that the explained variance of the data set is >= 0.999. Reducing data dimensionality makes training the model faster and hence allows for more up-to-date models.
## Inlier metric
The `inlier_metric` is a metric aimed at quantifying how similar the features of a data point are to the most recent historical data points.
You define the lookback window by setting `inlier_metric_window` and FreqAI computes the distance between the present time point and each of the previous `inlier_metric_window` lookback points. A Weibull function is fit to each of the lookback distributions and its cumulative distribution function (CDF) is used to produce a quantile for each lookback point. The `inlier_metric` is then computed for each time point as the average of the corresponding lookback quantiles. The figure below explains the concept for an `inlier_metric_window` of 5.
FreqAI adds the `inlier_metric` to the training features and hence gives the model access to a novel type of temporal information.
This function does **not** remove outliers from the data set.
## Weighting features for temporal importance
### Weighting features for temporal importance
FreqAI allows you to set a `weight_factor` to weight recent data more strongly than past data via an exponential function:
@@ -254,13 +222,103 @@ where $W_i$ is the weight of data point $i$ in a total set of $n$ data points. B
## Building the data pipeline
By default, FreqAI builds a dynamic pipeline based on user congfiguration settings. The default settings are robust and designed to work with a variety of methods. These two steps are a `MinMaxScaler(-1,1)` and a `VarianceThreshold` which removes any column that has 0 variance. Users can activate other steps with more configuration parameters. For example if users add `use_SVM_to_remove_outliers: true` to the `freqai` config, then FreqAI will automatically add the [`SVMOutlierExtractor`](#identifying-outliers-using-a-support-vector-machine-svm) to the pipeline. Likewise, users can add `principal_component_analysis: true` to the `freqai` config to activate PCA. The [DissimilarityIndex](#identifying-outliers-with-the-dissimilarity-index-di) is activated with `DI_threshold: 1`. Finally, noise can also be added to the data with `noise_standard_deviation: 0.1`. Finally, users can add [DBSCAN](#identifying-outliers-with-dbscan) outlier removal with `use_DBSCAN_to_remove_outliers: true`.
!!! note "More information available"
Please review the [parameter table](freqai-parameter-table.md) for more information on these parameters.
### Customizing the pipeline
Users are encouraged to customize the data pipeline to their needs by building their own data pipeline. This can be done by simply setting `dk.feature_pipeline` to their desired `Pipeline` object inside their `IFreqaiModel` `train()` function, or if they prefer not to touch the `train()` function, they can override `define_data_pipeline`/`define_label_pipeline` functions in their `IFreqaiModel`:
!!! note "More information available"
FreqAI uses the the [`DataSieve`](https://github.com/emergentmethods/datasieve) pipeline, which follows the SKlearn pipeline API, but adds, among other features, coherence between the X, y, and sample_weight vector point removals, feature removal, feature name following.
```python
from datasieve.transforms import SKLearnWrapper, DissimilarityIndex
from datasieve.pipeline import Pipeline
from sklearn.preprocessing import QuantileTransformer, StandardScaler
from freqai.base_models import BaseRegressionModel
User defines their custom label pipeline here (if they wish)
"""
label_pipeline = Pipeline([
('qt', SKLearnWrapper(StandardScaler())),
])
return label_pipeline
```
Here, you are defining the exact pipeline that will be used for your feature set during training and prediction. You can use *most* SKLearn transformation steps by wrapping them in the `SKLearnWrapper` class as shown above. In addition, you can use any of the transformations available in the [`DataSieve` library](https://github.com/emergentmethods/datasieve).
You can easily add your own transformation by creating a class that inherits from the datasieve `BaseTransform` and implementing your `fit()`, `transform()` and `inverse_transform()` methods:
```python
from datasieve.transforms.base_transform import BaseTransform
# do/dont do something with X, y, sample_weight, or/and feature_list
return X, y, sample_weight, feature_list
```
!!! note "Hint"
You can define this custom class in the same file as your `IFreqaiModel`.
### Migrating a custom `IFreqaiModel` to the new Pipeline
If you have created your own custom `IFreqaiModel` with a custom `train()`/`predict()` function, *and* you still rely on `data_cleaning_train/predict()`, then you will need to migrate to the new pipeline. If your model does *not* rely on `data_cleaning_train/predict()`, then you do not need to worry about this migration.
More details about the migration can be found [here](strategy_migration.md#freqai---new-data-pipeline).
## Outlier detection
Equity and crypto markets suffer from a high level of non-patterned noise in the form of outlier data points. FreqAI implements a variety of methods to identify such outliers and hence mitigate risk.
### Identifying outliers with the Dissimilarity Index (DI)
The Dissimilarity Index (DI) aims to quantify the uncertainty associated with each prediction made by the model.
The Dissimilarity Index (DI) aims to quantify the uncertainty associated with each prediction made by the model.
You can tell FreqAI to remove outlier data points from the training/test data sets using the DI by including the following statement in the config:
@@ -272,7 +330,7 @@ You can tell FreqAI to remove outlier data points from the training/test data se
}
```
The DI allows predictions which are outliers (not existent in the model feature space) to be thrown out due to low levels of certainty. To do so, FreqAI measures the distance between each training data point (feature vector), $X_{a}$, and all other training data points:
Which will add `DissimilarityIndex` step to your `feature_pipeline` and set the threshold to 1. The DI allows predictions which are outliers (not existent in the model feature space) to be thrown out due to low levels of certainty. To do so, FreqAI measures the distance between each training data point (feature vector), $X_{a}$, and all other training data points:
@@ -306,9 +364,9 @@ You can tell FreqAI to remove outlier data points from the training/test data se
}
```
The SVM will be trained on the training data and any data point that the SVM deems to be beyond the feature space will be removed.
Which will add `SVMOutlierExtractor` step to your `feature_pipeline`. The SVM will be trained on the training data and any data point that the SVM deems to be beyond the feature space will be removed.
FreqAI uses `sklearn.linear_model.SGDOneClassSVM` (details are available on scikit-learn's webpage [here](https://scikit-learn.org/stable/modules/generated/sklearn.linear_model.SGDOneClassSVM.html) (external website)) and you can elect to provide additional parameters for the SVM, such as `shuffle`, and `nu`.
You can elect to provide additional parameters for the SVM, such as `shuffle`, and `nu` via the `feature_parameters.svm_params` dictionary in the config.
The parameter `shuffle` is by default set to `False` to ensure consistent results. If it is set to `True`, running the SVM multiple times on the same data set might result in different outcomes due to `max_iter` being to low for the algorithm to reach the demanded `tol`. Increasing `max_iter` solves this issue but causes the procedure to take longer time.
@@ -326,7 +384,7 @@ You can configure FreqAI to use DBSCAN to cluster and remove outliers from the t
}
```
DBSCAN is an unsupervised machine learning algorithm that clusters data without needing to know how many clusters there should be.
Which will add the `DataSieveDBSCAN` step to your `feature_pipeline`. This is an unsupervised machine learning algorithm that clusters data without needing to know how many clusters there should be.
Given a number of data points $N$, and a distance $\varepsilon$, DBSCAN clusters the data set by setting all data points that have $N-1$ other data points within a distance of $\varepsilon$ as *core points*. A data point that is within a distance of $\varepsilon$ from a *core point* but that does not have $N-1$ other data points within a distance of $\varepsilon$ from itself is considered an *edge point*. A cluster is then the collection of *core points* and *edge points*. Data points that have no other data points at a distance $<\varepsilon$ are considered outliers. The figure below shows a cluster with $N = 3$.
@@ -18,9 +18,10 @@ Mandatory parameters are marked as **Required** and have to be set in one of the
| `purge_old_models` | Number of models to keep on disk (not relevant to backtesting). Default is 2, which means that dry/live runs will keep the latest 2 models on disk. Setting to 0 keeps all models. This parameter also accepts a boolean to maintain backwards compatibility. <br>**Datatype:** Integer. <br> Default: `2`.
| `save_backtest_models` | Save models to disk when running backtesting. Backtesting operates most efficiently by saving the prediction data and reusing them directly for subsequent runs (when you wish to tune entry/exit parameters). Saving backtesting models to disk also allows to use the same model files for starting a dry/live instance with the same model `identifier`. <br>**Datatype:** Boolean. <br> Default: `False` (no models are saved).
| `fit_live_predictions_candles` | Number of historical candles to use for computing target (label) statistics from prediction data, instead of from the training dataset (more information can be found [here](freqai-configuration.md#creating-a-dynamic-target-threshold)). <br>**Datatype:** Positive integer.
| `continual_learning` | Use the final state of the most recently trained model as starting point for the new model, allowing for incremental learning (more information can be found [here](freqai-running.md#continual-learning)). <br>**Datatype:** Boolean. <br> Default: `False`.
| `continual_learning` | Use the final state of the most recently trained model as starting point for the new model, allowing for incremental learning (more information can be found [here](freqai-running.md#continual-learning)). Beware that this is currently a naive approach to incremental learning, and it has a high probability of overfitting/getting stuck in local minima while the market moves away from your model. We have the connections here primarily for experimental purposes and so that it is ready for more mature approaches to continual learning in chaotic systems like the crypto market.<br>**Datatype:** Boolean. <br> Default: `False`.
| `write_metrics_to_disk` | Collect train timings, inference timings and cpu usage in json file. <br>**Datatype:** Boolean. <br> Default: `False`
| `data_kitchen_thread_count` | <br> Designate the number of threads you want to use for data processing (outlier methods, normalization, etc.). This has no impact on the number of threads used for training. If user does not set it (default), FreqAI will use max number of threads - 2 (leaving 1 physical core available for Freqtrade bot and FreqUI) <br>**Datatype:** Positive integer.
| `activate_tensorboard` | <br> Indicate whether or not to activate tensorboard for the tensorboard enabled modules (currently Reinforcment Learning, XGBoost, Catboost, and PyTorch). Tensorboard needs Torch installed, which means you will need the torch/RL docker image or you need to answer "yes" to the install question about whether or not you wish to install Torch. <br>**Datatype:** Boolean. <br> Default: `True`.
### Feature parameters
@@ -41,7 +42,6 @@ Mandatory parameters are marked as **Required** and have to be set in one of the
| `svm_params` | All parameters available in Sklearn's `SGDOneClassSVM()`. See details about some select parameters [here](freqai-feature-engineering.md#identifying-outliers-using-a-support-vector-machine-svm). <br>**Datatype:** Dictionary.
| `use_DBSCAN_to_remove_outliers` | Cluster data using the DBSCAN algorithm to identify and remove outliers from training and prediction data. See details about how it works [here](freqai-feature-engineering.md#identifying-outliers-with-dbscan). <br>**Datatype:** Boolean.
| `inlier_metric_window` | If set, FreqAI adds an `inlier_metric` to the training feature set and set the lookback to be the `inlier_metric_window`, i.e., the number of previous time points to compare the current candle to. Details of how the `inlier_metric` is computed can be found [here](freqai-feature-engineering.md#inlier-metric). <br>**Datatype:** Integer. <br> Default: `0`.
| `noise_standard_deviation` | If set, FreqAI adds noise to the training features with the aim of preventing overfitting. FreqAI generates random deviates from a gaussian distribution with a standard deviation of `noise_standard_deviation` and adds them to all data points. `noise_standard_deviation` should be kept relative to the normalized space, i.e., between -1 and 1. In other words, since data in FreqAI is always normalized to be between -1 and 1, `noise_standard_deviation: 0.05` would result in 32% of the data being randomly increased/decreased by more than 2.5% (i.e., the percent of data falling within the first standard deviation). <br>**Datatype:** Integer. <br> Default: `0`.
| `outlier_protection_percentage` | Enable to prevent outlier detection methods from discarding too much data. If more than `outlier_protection_percentage` % of points are detected as outliers by the SVM or DBSCAN, FreqAI will log a warning message and ignore outlier detection, i.e., the original dataset will be kept intact. If the outlier protection is triggered, no predictions will be made based on the training dataset. <br>**Datatype:** Float. <br> Default: `30`.
| `reverse_train_test_order` | Split the feature dataset (see below) and use the latest data split for training and test on historical split of the data. This allows the model to be trained up to the most recent data point, while avoiding overfitting. However, you should be careful to understand the unorthodox nature of this parameter before employing it. <br>**Datatype:** Boolean. <br> Default: `False` (no reversal).
@@ -85,6 +85,28 @@ Mandatory parameters are marked as **Required** and have to be set in one of the
| `net_arch` | Network architecture which is well described in [`stable_baselines3` doc](https://stable-baselines3.readthedocs.io/en/master/guide/custom_policy.html#examples). In summary: `[<sharedlayers>, dict(vf=[<non-sharedvaluenetworklayers>], pi=[<non-sharedpolicynetworklayers>])]`. By default this is set to `[128, 128]`, which defines 2 shared hidden layers with 128 units each.
| `randomize_starting_position` | Randomize the starting point of each episode to avoid overfitting. <br> **Datatype:** bool. <br> Default: `False`.
| `drop_ohlc_from_features` | Do not include the normalized ohlc data in the feature set passed to the agent during training (ohlc will still be used for driving the environment in all cases) <br> **Datatype:** Boolean. <br> **Default:** `False`
| `progress_bar` | Display a progress bar with the current progress, elapsed time and estimated remaining time. <br> **Datatype:** Boolean. <br> Default: `False`.
### PyTorch parameters
#### general
| Parameter | Description |
|------------|-------------|
| | **Model training parameters within the `freqai.model_training_parameters` sub dictionary**
| `learning_rate` | Learning rate to be passed to the optimizer. <br> **Datatype:** float. <br> Default: `3e-4`.
| `model_kwargs` | Parameters to be passed to the model class. <br> **Datatype:** dict. <br> Default: `{}`.
| `trainer_kwargs` | Parameters to be passed to the trainer class. <br> **Datatype:** dict. <br> Default: `{}`.
#### trainer_kwargs
| Parameter | Description |
|--------------|-------------|
| | **Model training parameters within the `freqai.model_training_parameters.model_kwargs` sub dictionary**
| `n_epochs` | The `n_epochs` parameter is a crucial setting in the PyTorch training loop that determines the number of times the entire training dataset will be used to update the model's parameters. An epoch represents one full pass through the entire training dataset. Overrides `n_steps`. Either `n_epochs` or `n_steps` must be set. <br><br> **Datatype:** int. optional. <br> Default: `10`.
| `n_steps` | An alternative way of setting `n_epochs` - the number of training iterations to run. Iteration here refer to the number of times we call `optimizer.step()`. Ignored if `n_epochs` is set. A simplified version of the function: <br><br> n_epochs = n_steps / (n_obs / batch_size) <br><br> The motivation here is that `n_steps` is easier to optimize and keep stable across different n_obs - the number of data points. <br> <br> **Datatype:** int. optional. <br> Default: `None`.
| `batch_size` | The size of the batches to use during training. <br><br> **Datatype:** int. <br> Default: `64`.
### Additional parameters
@@ -92,5 +114,5 @@ Mandatory parameters are marked as **Required** and have to be set in one of the
|------------|-------------|
| | **Extraneous parameters**
| `freqai.keras` | If the selected model makes use of Keras (typical for TensorFlow-based prediction models), this flag needs to be activated so that the model save/loading follows Keras standards. <br> **Datatype:** Boolean. <br> Default: `False`.
| `freqai.conv_width` | The width of a convolutional neural network input tensor. This replaces the need for shifting candles (`include_shifted_candles`) by feeding in historical data points as the second dimension of the tensor. Technically, this parameter can also be used for regressors, but it only adds computational overhead and does not change the model training/prediction. <br> **Datatype:** Integer. <br> Default: `2`.
| `freqai.conv_width` | The width of a neural network input tensor. This replaces the need for shifting candles (`include_shifted_candles`) by feeding in historical data points as the second dimension of the tensor. Technically, this parameter can also be used for regressors, but it only adds computational overhead and does not change the model training/prediction. <br> **Datatype:** Integer. <br> Default: `2`.
| `freqai.reduce_df_footprint` | Recast all numeric columns to float32/int32, with the objective of reducing ram/disk usage and decreasing train/inference timing. This parameter is set in the main level of the Freqtrade configuration file (not inside FreqAI). <br> **Datatype:** Boolean. <br> Default: `False`.
@@ -20,7 +20,7 @@ With the current framework, we aim to expose the training environment via the co
We envision the majority of users focusing their effort on creative design of the `calculate_reward()` function [details here](#creating-a-custom-reward-function), while leaving the rest of the environment untouched. Other users may not touch the environment at all, and they will only play with the configuration settings and the powerful feature engineering that already exists in FreqAI. Meanwhile, we enable advanced users to create their own model classes entirely.
The framework is built on stable_baselines3 (torch) and OpenAI gym for the base environment class. But generally speaking, the model class is well isolated. Thus, the addition of competing libraries can be easily integrated into the existing framework. For the environment, it is inheriting from `gym.env` which means that it is necessary to write an entirely new environment in order to switch to a different library.
The framework is built on stable_baselines3 (torch) and OpenAI gym for the base environment class. But generally speaking, the model class is well isolated. Thus, the addition of competing libraries can be easily integrated into the existing framework. For the environment, it is inheriting from `gym.Env` which means that it is necessary to write an entirely new environment in order to switch to a different library.
where `ReinforcementLearner` will use the templated `ReinforcementLearner` from `freqai/prediction_models/ReinforcementLearner` (or a custom user defined one located in `user_data/freqaimodels`). The strategy, on the other hand, follows the same base [feature engineering](freqai-feature-engineering.md) with `feature_engineering_*` as a typical Regressor. The difference lies in the creation of the targets, Reinforcement Learning doesn't require them. However, FreqAI requires a default (neutral) value to be set in the action column:
Required function to set the targets for the model.
@@ -53,17 +53,19 @@ where `ReinforcementLearner` will use the templated `ReinforcementLearner` from
# For RL, there are no direct targets to set. This is filler (neutral)
# until the agent sends an action.
dataframe["&-action"]=0
returndataframe
```
Most of the function remains the same as for typical Regressors, however, the function below shows how the strategy must pass the raw price data to the agent so that it has access to raw OHLCV in the training environment:
# The following features are necessary for RL models
dataframe[f"%-raw_close"]=dataframe["close"]
dataframe[f"%-raw_open"]=dataframe["open"]
dataframe[f"%-raw_high"]=dataframe["high"]
dataframe[f"%-raw_low"]=dataframe["low"]
returndataframe
```
Finally, there is no explicit "label" to make - instead it is necessary to assign the `&-action` column which will contain the agent's actions when accessed in `populate_entry/exit_trends()`. In the present example, the neutral action to 0. This value should align with the environment used. FreqAI provides two environments, both use 0 as the neutral action.
@@ -133,130 +135,139 @@ Parameter details can be found [here](freqai-parameter-table.md), but in general
## Creating a custom reward function
As you begin to modify the strategy and the prediction model, you will quickly realize some important differences between the Reinforcement Learner and the Regressors/Classifiers. Firstly, the strategy does not set a target value (no labels!). Instead, you set the `calculate_reward()` function inside the `MyRLEnv` class (see below). A default `calculate_reward()` is provided inside `prediction_models/ReinforcementLearner.py` to demonstrate the necessary building blocks for creating rewards, but users are encouraged to create their own custom reinforcement learning model class (see below) and save it to `user_data/freqaimodels`. It is inside the `calculate_reward()` where creative theories about the market can be expressed. For example, you can reward your agent when it makes a winning trade, and penalize the agent when it makes a losing trade. Or perhaps, you wish to reward the agent for entering trades, and penalize the agent for sitting in trades too long. Below we show examples of how these rewards are all calculated:
!!! danger "Not for production"
Warning!
The reward function provided with the Freqtrade source code is a showcase of functionality designed to show/test as many possible environment control features as possible. It is also designed to run quickly on small computers. This is a benchmark, it is *not* for live production. Please beware that you will need to create your own custom_reward() function or use a template built by other users outside of the Freqtrade source code.
As you begin to modify the strategy and the prediction model, you will quickly realize some important differences between the Reinforcement Learner and the Regressors/Classifiers. Firstly, the strategy does not set a target value (no labels!). Instead, you set the `calculate_reward()` function inside the `MyRLEnv` class (see below). A default `calculate_reward()` is provided inside `prediction_models/ReinforcementLearner.py` to demonstrate the necessary building blocks for creating rewards, but this is *not* designed for production. Users *must* create their own custom reinforcement learning model class or use a pre-built one from outside the Freqtrade source code and save it to `user_data/freqaimodels`. It is inside the `calculate_reward()` where creative theories about the market can be expressed. For example, you can reward your agent when it makes a winning trade, and penalize the agent when it makes a losing trade. Or perhaps, you wish to reward the agent for entering trades, and penalize the agent for sitting in trades too long. Below we show examples of how these rewards are all calculated:
!!! note "Hint"
The best reward functions are ones that are continuously differentiable, and well scaled. In other words, adding a single large negative penalty to a rare event is not a good idea, and the neural net will not be able to learn that function. Instead, it is better to add a small negative penalty to a common event. This will help the agent learn faster. Not only this, but you can help improve the continuity of your rewards/penalties by having them scale with severity according to some linear/exponential functions. In other words, you'd slowly scale the penalty as the duration of the trade increases. This is better than a single large penalty occuring at a single point in time.
Reinforcement Learning models benefit from tracking training metrics. FreqAI has integrated Tensorboard to allow users to track training and evaluation performance across all coins and across all retrainings. Tensorboard is activated via the following command:
```bash
cd freqtrade
tensorboard --logdir user_data/models/unique-id
```
where `unique-id` is the `identifier` set in the `freqai` configuration file. This command must be run in a separate shell to view the output in their browser at 127.0.0.1:6006 (6006 is the default port used by Tensorboard).
where `unique-id` is the `identifier` set in the `freqai` configuration file. This command must be run in a separate shell to view the output in the browser at 127.0.0.1:6006 (6006 is the default port used by Tensorboard).
### Custom logging
## Custom logging
FreqAI also provides a built in episodic summary logger called `self.tensorboard_log` for adding custom information to the Tensorboard log. By default, this function is already called once per step inside the environment to record the agent actions. All values accumulated for all steps in a single episode are reported at the conclusion of each episode, followed by a full reset of all metrics to 0 in preparation for the subsequent episode.
`self.tensorboard_log` can also be used anywhere inside the environment, for example, it can be added to the `calculate_reward` function to collect more detailed information about how often various parts of the reward were called:
```py
classMyRLEnv(Base5ActionRLEnv):
"""
User made custom environment. This class inherits from BaseEnvironment and gym.env.
Users can override any functions from those parent classes. Here is an example
of a user customized `calculate_reward()` function.
"""
defcalculate_reward(self,action:int)->float:
ifnotself._is_valid(action):
self.tensorboard_log("invalid")
return-2
```python
classMyRLEnv(Base5ActionRLEnv):
"""
User made custom environment. This class inherits from BaseEnvironment and gym.Env.
Users can override any functions from those parent classes. Here is an example
of a user customized `calculate_reward()` function.
"""
defcalculate_reward(self,action:int)->float:
ifnotself._is_valid(action):
self.tensorboard_log("invalid")
return-2
```
!!! Note
The `self.tensorboard_log()` function is designed for tracking incremented objects only i.e. events, actions inside the training environment. If the event of interest is a float, the float can be passed as the second argument e.g. `self.tensorboard_log("float_metric1", 0.23)`. In this case the metric values are not incremented.
### Choosing a base environment
## Choosing a base environment
FreqAI provides three base environments, `Base3ActionRLEnvironment`, `Base4ActionEnvironment` and `Base5ActionEnvironment`. As the names imply, the environments are customized for agents that can select from 3, 4 or 5 actions. The `Base3ActionEnvironment` is the simplest, the agent can select from hold, long, or short. This environment can also be used for long-only bots (it automatically follows the `can_short` flag from the strategy), where long is the enter condition and short is the exit condition. Meanwhile, in the `Base4ActionEnvironment`, the agent can enter long, enter short, hold neutral, or exit position. Finally, in the `Base5ActionEnvironment`, the agent has the same actions as Base4, but instead of a single exit action, it separates exit long and exit short. The main changes stemming from the environment selection include:
@@ -131,6 +131,9 @@ You can choose to adopt a continual learning scheme by setting `"continual_learn
???+ danger "Continual learning enforces a constant parameter space"
Since `continual_learning` means that the model parameter space *cannot* change between trainings, `principal_component_analysis` is automatically disabled when `continual_learning` is enabled. Hint: PCA changes the parameter space and the number of features, learn more about PCA [here](freqai-feature-engineering.md#data-dimensionality-reduction-with-principal-component-analysis).
???+ danger "Experimental functionality"
Beware that this is currently a naive approach to incremental learning, and it has a high probability of overfitting/getting stuck in local minima while the market moves away from your model. We have the mechanics available in FreqAI primarily for experimental purposes and so that it is ready for more mature approaches to continual learning in chaotic systems like the crypto market.
## Hyperopt
You can hyperopt using the same command as for [typical Freqtrade hyperopt](hyperopt.md):
@@ -158,7 +161,14 @@ This specific hyperopt would help you understand the appropriate `DI_values` for
## Using Tensorboard
CatBoost models benefit from tracking training metrics via Tensorboard. You can take advantage of the FreqAI integration to track training and evaluation performance across all coins and across all retrainings. Tensorboard is activated via the following command:
!!! note "Availability"
FreqAI includes tensorboard for a variety of models, including XGBoost, all PyTorch models, Reinforcement Learning, and Catboost. If you would like to see Tensorboard integrated into another model type, please open an issue on the [Freqtrade GitHub](https://github.com/freqtrade/freqtrade/issues)
!!! danger "Requirements"
Tensorboard logging requires the FreqAI torch installation/docker image.
The easiest way to use tensorboard is to ensure `freqai.activate_tensorboard` is set to `True` (default setting) in your configuration file, run FreqAI, then open a separate shell and run:
where `unique-id` is the `identifier` set in the `freqai` configuration file. This command must be run in a separate shell if you wish to view the output in your browser at 127.0.0.1:6060 (6060 is the default port used by Tensorboard).
!!! note "Deactivate for improved performance"
Tensorboard logging can slow down training and should be deactivated for production use.
You will see the boot-up process of automatic data downloading, followed by simultaneous training and trading.
You will see the boot-up process of automatic data downloading, followed by simultaneous training and trading.
!!! danger "Not for production"
The example strategy provided with the Freqtrade source code is designed for showcasing/testing a wide variety of FreqAI features. It is also designed to run on small computers so that it can be used as a benchmark between developers and users. It is *not* designed to be run in production.
An example strategy, prediction model, and config to use as a starting points can be found in
`freqtrade/templates/FreqaiExampleStrategy.py`, `freqtrade/freqai/prediction_models/LightGBMRegressor.py`, and
Catboost will not be installed on arm devices (raspberry, Mac M1, ARM based VPS, ...), since it does not provide wheels for this platform.
!!! Note "python 3.11"
Some dependencies (Catboost, Torch) currently don't support python 3.11. Freqtrade therefore only supports python 3.10 for these models/dependencies.
Tests involving these dependencies are skipped on 3.11.
Catboost will not be installed on low-powered arm devices (raspberry), since it does not provide wheels for this platform.
### Usage with docker
If you are using docker, a dedicated tag with FreqAI dependencies is available as `:freqai`. As such - you can replace the image line in your docker compose file with `image: freqtradeorg/freqtrade:develop_freqai`. This image contains the regular FreqAI dependencies. Similar to native installs, Catboost will not be available on ARM based devices.
If you are using docker, a dedicated tag with FreqAI dependencies is available as `:freqai`. As such - you can replace the image line in your docker compose file with `image: freqtradeorg/freqtrade:develop_freqai`. This image contains the regular FreqAI dependencies. Similar to native installs, Catboost will not be available on ARM based devices. If you would like to use PyTorch or Reinforcement learning, you should use the torch or RL tags, `image: freqtradeorg/freqtrade:develop_freqaitorch`, `image: freqtradeorg/freqtrade:develop_freqairl`.
!!! note "docker-compose-freqai.yml"
We do provide an explicit docker-compose file for this in `docker/docker-compose-freqai.yml` - which can be used via `docker compose -f docker/docker-compose-freqai.yml run ...` - or can be copied to replace the original docker file. This docker-compose file also contains a (disabled) section to enable GPU resources within docker containers. This obviously assumes the system has GPU resources available.
### FreqAI position in open-source machine learning landscape
@@ -105,6 +107,13 @@ This is for performance reasons - FreqAI relies on making quick predictions/retr
it needs to download all the training data at the beginning of a dry/live instance. FreqAI stores and appends
new candles automatically for future retrains. This means that if new pairs arrive later in the dry run due to a volume pairlist, it will not have the data ready. However, FreqAI does work with the `ShufflePairlist` or a `VolumePairlist` which keeps the total pairlist constant (but reorders the pairs according to volume).
## Additional learning materials
Here we compile some external materials that provide deeper looks into various components of FreqAI:
- [Real-time head-to-head: Adaptive modeling of financial market data using XGBoost and CatBoost](https://emergentmethods.medium.com/real-time-head-to-head-adaptive-modeling-of-financial-market-data-using-xgboost-and-catboost-995a115a7495)
- [FreqAI - from price to prediction](https://emergentmethods.medium.com/freqai-from-price-to-prediction-6fadac18b665)
## Credits
FreqAI is developed by a group of individuals who all contribute specific skillsets to the project.
@@ -433,9 +433,14 @@ While this strategy is most likely too simple to provide consistent profit, it s
`range` property may also be used with `DecimalParameter` and `CategoricalParameter`. `RealParameter` does not provide this property due to infinite search space.
??? Hint "Performance tip"
During normal hyperopting, indicators are calculated once and supplied to each epoch, linearly increasing RAM usage as a factor of increasing cores. As this also has performance implications, hyperopt provides `--analyze-per-epoch` which will move the execution of `populate_indicators()` to the epoch process, calculating a single value per parameter per epoch instead of using the `.range` functionality. In this case, `.range` functionality will only return the actually used value. This will reduce RAM usage, but increase CPU usage. However, your hyperopting run will be less likely to fail due to Out Of Memory (OOM) issues.
During normal hyperopting, indicators are calculated once and supplied to each epoch, linearly increasing RAM usage as a factor of increasing cores. As this also has performance implications, there are two alternatives to reduce RAM usage
In either case, you should try to use space ranges as small as possible this will improve CPU/RAM usage in both scenarios.
* Move `ema_short` and `ema_long` calculations from `populate_indicators()` to `populate_entry_trend()`. Since `populate_entry_trend()` gonna be calculated every epochs, you don't need to use `.range` functionality.
* hyperopt provides `--analyze-per-epoch` which will move the execution of `populate_indicators()` to the epoch process, calculating a single value per parameter per epoch instead of using the `.range` functionality. In this case, `.range` functionality will only return the actually used value.
These alternatives will reduce RAM usage, but increase CPU usage. However, your hyperopting run will be less likely to fail due to Out Of Memory (OOM) issues.
Whether you are using `.range` functionality or the alternatives above, you should try to use space ranges as small as possible since this will improve CPU/RAM usage.
@@ -25,6 +25,7 @@ You may also use something like `.*DOWN/BTC` or `.*UP/BTC` to exclude leveraged
* [`ProducerPairList`](#producerpairlist)
* [`RemotePairList`](#remotepairlist)
* [`AgeFilter`](#agefilter)
* [`FullTradesFilter`](#fulltradesfilter)
* [`OffsetFilter`](#offsetfilter)
* [`PerformanceFilter`](#performancefilter)
* [`PrecisionFilter`](#precisionfilter)
@@ -184,6 +185,8 @@ The RemotePairList is defined in the pairlists section of the configuration sett
"pairlists": [
{
"method": "RemotePairList",
"mode": "whitelist",
"processing_mode": "filter",
"pairlist_url": "https://example.com/pairlist",
"number_assets": 10,
"refresh_period": 1800,
@@ -194,6 +197,14 @@ The RemotePairList is defined in the pairlists section of the configuration sett
]
```
The optional `mode` option specifies if the pairlist should be used as a `blacklist` or as a `whitelist`. The default value is "whitelist".
The optional `processing_mode` option in the RemotePairList configuration determines how the retrieved pairlist is processed. It can have two values: "filter" or "append".
In "filter" mode, the retrieved pairlist is used as a filter. Only the pairs present in both the original pairlist and the retrieved pairlist are included in the final pairlist. Other pairs are filtered out.
In "append" mode, the retrieved pairlist is added to the original pairlist. All pairs from both lists are included in the final pairlist without any filtering.
The `pairlist_url` option specifies the URL of the remote server where the pairlist is located, or the path to a local file (if file:/// is prepended). This allows the user to use either a remote server or a local file as the source for the pairlist.
The user is responsible for providing a server or local file that returns a JSON object with the following structure:
@@ -201,7 +212,7 @@ The user is responsible for providing a server or local file that returns a JSON
```json
{
"pairs": ["XRP/USDT", "ETH/USDT", "LTC/USDT"],
"refresh_period": 1800,
"refresh_period": 1800
}
```
@@ -226,6 +237,17 @@ be caught out buying before the pair has finished dropping in price.
This filter allows freqtrade to ignore pairs until they have been listed for at least `min_days_listed` days and listed before `max_days_listed`.
#### FullTradesFilter
Shrink whitelist to consist only in-trade pairs when the trade slots are full (when `max_open_trades` isn't being set to `-1` in the config).
When the trade slots are full, there is no need to calculate indicators of the rest of the pairs (except informative pairs) since no new trade can be opened. By shrinking the whitelist to just the in-trade pairs, you can improve calculation speeds and reduce CPU usage. When a trade slot is free (either a trade is closed or `max_open_trades` value in config is increased), then the whitelist will return to normal state.
When multiple pairlist filters are being used, it's recommended to put this filter at second position directly below the primary pairlist, so when the trade slots are full, the bot doesn't have to download data for the rest of the filters.
!!! Warning "Backtesting"
`FullTradesFilter` does not support backtesting mode.
#### OffsetFilter
Offsets an incoming pairlist by a given `offset` value.
@@ -366,7 +388,7 @@ If the trading range over the last 10 days is <1% or >99%, remove the pair from
"lookback_days": 10,
"min_rate_of_change": 0.01,
"max_rate_of_change": 0.99,
"refresh_period": 1440
"refresh_period": 86400
}
]
```
@@ -421,7 +443,7 @@ The below example blacklists `BNB/BTC`, uses `VolumePairList` with `20` assets,
@@ -24,18 +24,12 @@ The easiest way to install and run Freqtrade is to clone the bot Github reposito
The `stable` branch contains the code of the last release (done usually once per month on an approximately one week old snapshot of the `develop` branch to prevent packaging bugs, so potentially it's more stable).
!!! Note
Python3.8 or higher and the corresponding `pip` are assumed to be available. The install-script will warn you and stop if that's not the case. `git` is also needed to clone the Freqtrade repository.
Python3.9 or higher and the corresponding `pip` are assumed to be available. The install-script will warn you and stop if that's not the case. `git` is also needed to clone the Freqtrade repository.
Also, python headers (`python<yourversion>-dev` / `python<yourversion>-devel`) must be available for the installation to complete successfully.
!!! Warning "Up-to-date clock"
The clock on the system running the bot must be accurate, synchronized to a NTP server frequently enough to avoid problems with communication to the exchanges.
!!! Error "Running setup.py install for gym did not run successfully."
If you get an error related with gym we suggest you to downgrade setuptools it to version 65.5.0 you can do it with the following command:
```bash
pip install setuptools==65.5.0
```
------
## Requirements
@@ -48,11 +42,11 @@ These requirements apply to both [Script Installation](#script-installation) and
Use this csv-filename to store lookahead-analysis-
results
```
!!! Note ""
The above Output was reduced to options `lookahead-analysis` adds on top of regular backtesting commands.
### Summary
Checks a given strategy for look ahead bias via lookahead-analysis
Look ahead bias means that the backtest uses data from future candles thereby not making it viable beyond backtesting
and producing false hopes for the one backtesting.
### Introduction
Many strategies - without the programmer knowing - have fallen prey to look ahead bias.
Any backtest will populate the full dataframe including all time stamps at the beginning.
If the programmer is not careful or oblivious how things work internally
(which sometimes can be really hard to find out) then it will just look into the future making the strategy amazing
but not realistic.
This command is made to try to verify the validity in the form of the aforementioned look ahead bias.
### How does the command work?
It will start with a backtest of all pairs to generate a baseline for indicators and entries/exits.
After the backtest ran, it will look if the `minimum-trade-amount` is met
and if not cancel the lookahead-analysis for this strategy.
After setting the baseline it will then do additional runs for every entry and exit separately.
When a verification-backtest is done, it will compare the indicators as the signal (either entry or exit) and report the bias.
After all signals have been verified or falsified a result-table will be generated for the user to see.
### Caveats
-`lookahead-analysis` can only verify / falsify the trades it calculated and verified.
If the strategy has many different signals / signal types, it's up to you to select appropriate parameters to ensure that all signals have triggered at least once. Not triggered signals will not have been verified.
This could lead to a false-negative (the strategy will then be reported as non-biased).
-`lookahead-analysis` has access to everything that backtesting has too.
Please don't provoke any configs like enabling position stacking.
If you decide to do so, then make doubly sure that you won't ever run out of `max_open_trades` amount and neither leftover money in your wallet.
@@ -49,7 +49,7 @@ Enable subscribing to an instance by adding the `external_message_consumer` sect
| `wait_timeout` | Timeout until we ping again if no message is received. <br>*Defaults to `300`.*<br>**Datatype:** Integer - in seconds.
| `ping_timeout` | Ping timeout <br>*Defaults to `10`.*<br>**Datatype:** Integer - in seconds.
| `sleep_time` | Sleep time before retrying to connect.<br>*Defaults to `10`.*<br>**Datatype:** Integer - in seconds.
| `remove_entry_exit_signals` | Remove signal columns from the dataframe (set them to 0) on dataframe receipt.<br>*Defaults to `False`.*<br>**Datatype:** Boolean.
| `remove_entry_exit_signals` | Remove signal columns from the dataframe (set them to 0) on dataframe receipt.<br>*Defaults to `false`.*<br>**Datatype:** Boolean.
| `message_size_limit` | Size limit per message<br>*Defaults to `8`.*<br>**Datatype:** Integer - Megabytes.
Instead of (or as well as) calculating indicators in `populate_indicators()` the follower instance listens on the connection to a producer instance's messages (or multiple producer instances in advanced configurations) and requests the producer's most recently analyzed dataframes for each pair in the active whitelist.
This page explains how to validate your strategy for inaccuracies due to recursive issues with certain indicators.
A recursive formula defines any term of a sequence relative to its preceding term(s). An example of a recursive formula is a<sub>n</sub> = a<sub>n-1</sub> + b.
Why does this matter for Freqtrade? In backtesting, the bot will get full data of the pairs according to the timerange specified. But in a dry/live run, the bot will be limited by the amount of data each exchanges gives.
For example, to calculate a very basic indicator called `steps`, the first row's value is always 0, while the following rows' values are equal to the value of the previous row plus 1. If I were to calculate it using the latest 1000 candles, then the `steps` value of the first row is 0, and the `steps` value at the last closed candle is 999.
What happens if the calculation is using only the latest 500 candles? Then instead of 999, the `steps` value at last closed candle is 499. The difference of the value means your backtest result can differ from your dry/live run result.
The `recursive-analysis` command requires historic data to be available. To learn how to get data for the pairs and exchange you're interested in,
head over to the [Data Downloading](data-download.md) section of the documentation.
This command is built upon preparing different lengths of data and calculates indicators based on them.
This does not backtest the strategy itself, but rather only calculates the indicators. After calculating the indicators of different startup candle values (`startup_candle_count`) are done, the values of last rows across all specified `startup_candle_count` are compared to see how much variance they show compared to the base calculation.
Command settings:
- Use the `-p` option to set your desired pair to analyze. Since we are only looking at indicator values, using more than one pair is redundant. Preferably use a pair with a relatively high price and at least moderate volatility, such as BTC or ETH, to avoid rounding issues that can make the results inaccurate. If no pair is set on the command, the pair used for this analysis is the first pair in the whitelist.
- It is recommended to set a long timerange (at least 5000 candles) so that the initial indicators' calculation that is going to be used as a benchmark has very small or no recursive issues itself. For example, for a 5m timeframe, a timerange of 5000 candles would be equal to 18 days.
-`--cache` is forced to "none" to avoid loading previous indicators calculation automatically.
In addition to the recursive formula check, this command also carries out a simple lookahead bias check on the indicator values only. For a full lookahead check, use [Lookahead-analysis](lookahead-analysis.md).
### Why are odd-numbered default startup candles used?
The default value for startup candles are odd numbers. When the bot fetches candle data from the exchange's API, the last candle is the one being checked by the bot and the rest of the data are the "startup candles".
For example, Binance allows 1000 candles per API call. When the bot receives 1000 candles, the last candle is the "current candle", and the preceding 999 candles are the "startup candles". By setting the startup candle count as 1000 instead of 999, the bot will try to fetch 1001 candles instead. The exchange API will then send candle data in a paginated form, i.e. in case of the Binance API, this will be two groups- one of length 1000 and another of length 1. This results in the bot thinking the strategy needs 1001 candles of data, and so it will download 2000 candles worth of data instead, which means there will be 1 "current candle" and 1999 "startup candles".
Furthermore, exchanges limit the number of consecutive bulk API calls, e.g. Binance allows 5 calls. In this case, only 5000 candles can be downloaded from Binance API without hitting the API rate limit, which means the max `startup_candle_count` you can have is 4999.
Please note that this candle limit may be changed in the future by the exchanges without any prior notice.
### How does the command work?
- Firstly an initial indicator calculation is carried out using the supplied timerange to generate a benchmark for indicator values.
- After setting the benchmark it will then carry out additional runs for each of the different startup candle count values.
- The command will then compare the indicator values at the last candle rows and report the differences in a table.
## Understanding the recursive-analysis output
This is an example of an output results table where at least one indicator has a recursive formula issue:
The column headers indicate the different `startup_candle_count` used in the analysis. The values in the table indicate the variance of the calculated indicators compared to the benchmark value.
`nan%` means the value of that indicator cannot be calculated due to lack of data. In this example, you cannot calculate RSI with length 30 with just 21 candles (1 current candle + 20 startup candles).
Users should assess the table per indicator to decide if the specified `startup_candle_count` results in a sufficiently small variance so that the indicator does not have any effect on entries and/or exits.
As such, aiming for absolute zero variance (shown by `-` value) might not be the best option, because some indicators might require you to use such a long `startup_candle_count` to have zero variance.
## Caveats
-`recursive-analysis` will only calculate and compare the indicator values at the last row. The output table reports the percentage differences between the different startup candle count calculations and the original benchmark calculation. Whether it has any actual impact on your entries and exits is not included.
- The ideal scenario is that indicators will have no variance (or at least very close to 0%) despite the startup candle being varied. In reality, indicators such as EMA are using a recursive formula to calculate indicator values, so the goal is not necessarily to have zero percentage variance, but to have the variance low enough (and therefore `startup_candle_count` high enough) that the recursion inherent in the indicator will not have any real impact on trading decisions.
-`recursive-analysis` will only run calculations on `populate_indicators` and `@informative` decorator(s). If you put any indicator calculation on `populate_entry_trend` or `populate_exit_trend`, it won't be calculated.
@@ -9,9 +9,6 @@ This same command can also be used to update freqUI, should there be a new relea
Once the bot is started in trade / dry-run mode (with `freqtrade trade`) - the UI will be available under the configured port below (usually `http://127.0.0.1:8080`).
!!! info "Alpha release"
FreqUI is still considered an alpha release - if you encounter bugs or inconsistencies please open a [FreqUI issue](https://github.com/freqtrade/frequi/issues/new/choose).
!!! Note "developers"
Developers should not use this method, but instead use the method described in the [freqUI repository](https://github.com/freqtrade/frequi) to get the source-code of freqUI.
| `reload_config` | Reloads the configuration file.
| `trades` | List last trades. Limited to 500 trades per call.
| `trade/<tradeid>` | Get specific trade.
| `delete_trade<trade_id>` | Remove trade from the database. Tries to close open orders. Requires manual handling of this trade on the exchange.
| `trade/<tradeid>` | DELETE - Remove trade from the database. Tries to close open orders. Requires manual handling of this trade on the exchange.
| `trade/<tradeid>/open-order` | DELETE - Cancel open order for this trade.
| `trade/<tradeid>/reload` | GET - Reload a trade from the Exchange. Only works in live, and can potentially help recover a trade that was manually sold on the exchange.
| `show_config` | Shows part of the current configuration with relevant settings to operation.
We also recommend to set `datadir` to something identifying downloaded data as sandbox data, to avoid having sandbox data mixed with data from the real exchange.
This can be done by adding the `"datadir"` key to the configuration.
Now, whenever you use this configuration, your data directory will be set to this directory.
---
## You should now be ready to test your sandbox
Ensure Freqtrade logs show the sandbox URL, and trades made are shown in sandbox. Also make sure to select a pair which shows at least some decent value (which very often is BTC/<somestablecoin>).
## Common problems with sandbox exchanges
Sandbox exchange instances often have very low volume, which can cause some problems which usually are not seen on a real exchange instance.
### Old Candles problem
Since Sandboxes often have low volume, candles can be quite old and show no volume.
To disable the error "Outdated history for pair ...", best increase the parameter `"outdated_offset"` to a number that seems realistic for the sandbox you're using.
### Unfilled orders
Sandboxes often have very low volumes - which means that many trades can go unfilled, or can go unfilled for a very long time.
To mitigate this, you can try to match the first order on the opposite orderbook side using the following configuration:
``` jsonc
"order_types": {
"entry": "limit",
"exit": "limit"
// ...
},
"entry_pricing": {
"price_side": "other",
// ...
},
"exit_pricing":{
"price_side": "other",
// ...
},
```
The configuration is similar to the suggested configuration for market orders - however by using limit-orders you can avoid moving the price too much, and you can set the worst price you might get.
@@ -23,10 +23,22 @@ These modes can be configured with these values:
'stoploss_on_exchange_limit_ratio': 0.99
```
!!! Note
Stoploss on exchange is only supported for Binance (stop-loss-limit), Huobi (stop-limit), Kraken (stop-loss-market, stop-loss-limit), Gate (stop-limit), and Kucoin (stop-limit and stop-market) as of now.
<ins>Do not set too low/tight stoploss value if using stop loss on exchange!</ins>
If set to low/tight then you have greater risk of missing fill on the order and stoploss will not work.
Stoploss on exchange is only supported for the following exchanges, and not all exchanges support both stop-limit and stop-market.
The Order-type will be ignored if only one mode is available.
| Exchange | stop-loss type |
|----------|-------------|
| Binance | limit |
| Binance Futures | market, limit |
| Huobi | limit |
| kraken | market, limit |
| Gate | limit |
| Okx | limit |
| Kucoin | stop-limit, stop-market|
!!! Note "Tight stoploss"
<ins>Do not set too low/tight stoploss value when using stop loss on exchange!</ins>
If set to low/tight you will have greater risk of missing fill on the order and stoploss will not work.
### stoploss_on_exchange and stoploss_on_exchange_limit_ratio
@@ -197,11 +209,6 @@ You can also keep a static stoploss until the offset is reached, and then trail
If `trailing_only_offset_is_reached = True` then the trailing stoploss is only activated once the offset is reached. Until then, the stoploss remains at the configured `stoploss`.
This option can be used with or without `trailing_stop_positive`, but uses `trailing_stop_positive_offset` as offset.
This page explains some advanced concepts available for strategies.
If you're just getting started, please be familiar with the methods described in the [Strategy Customization](strategy-customization.md) documentation and with the [Freqtrade basics](bot-basics.md) first.
If you're just getting started, please familiarize yourself with the [Freqtrade basics](bot-basics.md) and methods described in [Strategy Customization](strategy-customization.md) first.
[Freqtrade basics](bot-basics.md) describes in which sequence each method described below is called, which can be helpful to understand which method to use for your custom needs.
The call sequence of the methods described here is covered under [bot execution logic](bot-basics.md#bot-execution-logic). Those docs are also helpful in deciding which method is most suitable for your customisation needs.
!!! Note
All callback methods described below should only be implemented in a strategy if they are actually used.
Callback methods should *only* be implemented if a strategy uses them.
!!! Tip
You can get a strategy template containing all below methods by running `freqtrade new-strategy --strategy MyAwesomeStrategy --template advanced`
Start off with a strategy template containing all available callback methods by running `freqtrade new-strategy --strategy MyAwesomeStrategy --template advanced`
## Storing information
Storing information can be accomplished by creating a new dictionary within the strategy class.
The name of the variable can be chosen at will, but should be prefixed with `cust_` to avoid naming collisions with predefined strategy variables.
The name of the variable can be chosen at will, but should be prefixed with `custom_` to avoid naming collisions with predefined strategy variables.
```python
classAwesomeStrategy(IStrategy):
@@ -227,8 +227,8 @@ for val in self.buy_ema_short.range:
# Combine all dataframes, and reassign the original dataframe column
dataframe = pd.concat(frames, axis=1)
```
Freqtrade does however also counter this by running `dataframe.copy()` on the dataframe right after the `populate_indicators()` method - so performance implications of this should be low to non-existant.
@@ -164,6 +164,31 @@ E.g. If the `current_rate` is 200 USD, then returning `0.02` will set the stoplo
During backtesting, `current_rate` (and `current_profit`) are provided against the candle's high (or low for short trades) - while the resulting stoploss is evaluated against the candle's low (or high for short trades).
The absolute value of the return value is used (the sign is ignored), so returning `0.05` or `-0.05` have the same result, a stoploss 5% below the current price.
Returning None will be interpreted as "no desire to change", and is the only safe way to return when you'd like to not modify the stoploss.
Stoploss on exchange works similar to `trailing_stop`, and the stoploss on exchange is updated as configured in `stoploss_on_exchange_interval` ([More details about stoploss on exchange](stoploss.md#stop-loss-on-exchange-freqtrade)).
!!! Note "Use of dates"
All time-based calculations should be done based on `current_time` - using `datetime.now()` or `datetime.utcnow()` is discouraged, as this will break backtesting support.
!!! Tip "Trailing stoploss"
It's recommended to disable `trailing_stop` when using custom stoploss values. Both can work in tandem, but you might encounter the trailing stop to move the price higher while your custom function would not want this, causing conflicting behavior.
### Adjust stoploss after position adjustments
Depending on your strategy, you may encounter the need to adjust the stoploss in both directions after a [position adjustment](#adjust-trade-position).
For this, freqtrade will make an additional call with `after_fill=True` after an order fills, which will allow the strategy to move the stoploss in any direction (also widening the gap between stoploss and current price, which is otherwise forbidden).
!!! Note "backwards compatibility"
This call will only be made if the `after_fill` parameter is part of the function definition of your `custom_stoploss` function.
As such, this will not impact (and with that, surprise) existing, running strategies.
### Custom stoploss examples
The next section will show some examples on what's possible with the custom stoploss function.
Of course, many more things are possible, and all examples can be combined at will.
#### Trailing stop via custom stoploss
To simulate a regular trailing stoploss of 4% (trailing 4% behind the maximum reached price) you would use the following very simple method:
@@ -179,7 +204,8 @@ class AwesomeStrategy(IStrategy):
Custom stoploss logic, returning the new distance relative to current_rate (as ratio).
e.g. returning -0.05 would create a stoploss 5% below current_rate.
@@ -187,7 +213,7 @@ class AwesomeStrategy(IStrategy):
For full documentation please go to https://www.freqtrade.io/en/latest/strategy-advanced/
When not implemented by a strategy, returns the initial stoploss value
When not implemented by a strategy, returns the initial stoploss value.
Only called when use_custom_stoploss is set to True.
:param pair: Pair that's currently analyzed
@@ -195,25 +221,13 @@ class AwesomeStrategy(IStrategy):
:param current_time: datetime object, containing the current datetime
:param current_rate: Rate, calculated based on pricing settings in exit_pricing.
:param current_profit: Current profit (as ratio), calculated based on current_rate.
:param after_fill: True if the stoploss is called after the order was filled.
:param **kwargs: Ensure to keep this here so updates to this won't break your strategy.
:return float: New stoploss value, relative to the currentrate
:return float: New stoploss value, relative to the current_rate
"""
return -0.04
```
Stoploss on exchange works similar to `trailing_stop`, and the stoploss on exchange is updated as configured in `stoploss_on_exchange_interval` ([More details about stoploss on exchange](stoploss.md#stop-loss-on-exchange-freqtrade)).
!!! Note "Use of dates"
All time-based calculations should be done based on `current_time` - using `datetime.now()` or `datetime.utcnow()` is discouraged, as this will break backtesting support.
!!! Tip "Trailing stoploss"
It's recommended to disable `trailing_stop` when using custom stoploss values. Both can work in tandem, but you might encounter the trailing stop to move the price higher while your custom function would not want this, causing conflicting behavior.
### Custom stoploss examples
The next section will show some examples on what's possible with the custom stoploss function.
Of course, many more things are possible, and all examples can be combined at will.
#### Time based trailing stop
Use the initial stoploss for the first 60 minutes, after this change to 10% trailing stoploss, and after 2 hours (120 minutes) we use a 5% trailing stoploss.
@@ -229,14 +243,45 @@ class AwesomeStrategy(IStrategy):
#### Time based trailing stop with after-fill adjustments
Use the initial stoploss for the first 60 minutes, after this change to 10% trailing stoploss, and after 2 hours (120 minutes) we use a 5% trailing stoploss.
If an additional order fills, set stoploss to -10% below the new `open_rate` ([Averaged across all entries](#position-adjust-calculations)).
# return maximum stoploss value, keeping current stoploss price unchanged
return 1
return None
```
See [Dataframe access](strategy-advanced.md#dataframe-access) for more information about dataframe use in strategy callbacks.
@@ -364,15 +413,89 @@ See [Dataframe access](strategy-advanced.md#dataframe-access) for more informati
#### Stoploss relative to open price
Stoploss values returned from `custom_stoploss()` always specify a percentage relative to `current_rate`. In order to set a stoploss relative to the *open* price, we need to use `current_profit` to calculate what percentage relative to the `current_rate` will give you the same result as if the percentage was specified from the open price.
Stoploss values returned from `custom_stoploss()` must specify a percentage relative to `current_rate`, but sometimes you may want to specify a stoploss relative to the _entry_ price instead.
`stoploss_from_open()` is a helper function to calculate a stoploss value that can be returned from `custom_stoploss` which will be equivalent to the desired trade profit above the entry point.
The helper function [`stoploss_from_open()`](strategy-customization.md#stoploss_from_open) can be used to convert from an open price relative stop, to a current price relative stop which can be returned from `custom_stoploss()`.
??? Example "Returning a stoploss relative to the open price from the customstoploss function"
Say the open price was $100, and `current_price` is $121 (`current_profit` will be `0.21`).
If we want a stop price at 7% above the open price we can call `stoploss_from_open(0.07, current_profit, False)` which will return `0.1157024793`. 11.57% below $121 is $107, which is the same as 7% above $100.
This function will consider leverage - so at 10x leverage, the actual stoploss would be 0.7% above $100 (0.7% * 10x = 7%).
``` python
from datetime import datetime
from freqtrade.persistence import Trade
from freqtrade.strategy import IStrategy, stoploss_from_open
Full examples can be found in the [Custom stoploss](strategy-advanced.md#custom-stoploss) section of the Documentation.
!!! Note
Providing invalid input to `stoploss_from_open()` may produce "CustomStoploss function did not return valid stoploss" warnings.
This may happen if `current_profit` parameter is below specified `open_relative_stop`. Such situations may arise when closing trade
is blocked by `confirm_trade_exit()` method. Warnings can be solved by never blocking stop loss sells by checking `exit_reason` in
`confirm_trade_exit()`, or by using `return stoploss_from_open(...) or 1` idiom, which will request to not change stop loss when
`current_profit <open_relative_stop`.
#### Stoploss percentage from absolute price
Stoploss values returned from `custom_stoploss()` always specify a percentage relative to `current_rate`. In order to set a stoploss at specified absolute price level, we need to use `stop_rate` to calculate what percentage relative to the `current_rate` will give you the same result as if the percentage was specified from the open price.
The helper function [`stoploss_from_absolute()`](strategy-customization.md#stoploss_from_absolute) can be used to convert from an absolute price, to a current price relative stop which can be returned from `custom_stoploss()`.
The helper function `stoploss_from_absolute()` can be used to convert from an absolute price, to a current price relative stop which can be returned from `custom_stoploss()`.
??? Example "Returning a stoploss using absolute price from the custom stoploss function"
If we want to trail a stop price at 2xATR below current price we can call `stoploss_from_absolute(current_rate+(side* candle['atr'] *2),current_rate,is_short=trade.is_short,leverage=trade.leverage)`.
For futures, we need to adjust the direction (up or down), as well as adjust for leverage, since the [`custom_stoploss`](strategy-callbacks.md#custom-stoploss) callback returns the ["risk for this trade"](stoploss.md#stoploss-and-leverage) - not the relative price movement.
``` python
from datetime import datetime
from freqtrade.persistence import Trade
from freqtrade.strategy import IStrategy, stoploss_from_absolute, timeframe_to_prev_date
@@ -387,6 +510,9 @@ Each of these methods are called right before placing an order on the exchange.
!!! Note
If your custom pricing function return None or an invalid value, price will fall back to `proposed_rate`, which is based on the regular pricing configuration.
!!! Note
Using custom_entry_price, the Trade object will be available as soon as the first entry order associated with the trade is created, for the first entry, `trade` parameter value will be `None`.
### Custom order entry and exit price example
``` python
@@ -397,7 +523,7 @@ class AwesomeStrategy(IStrategy):
@@ -793,6 +920,8 @@ Returning any other price will cancel the existing order, and replace it with a
The trade open-date (`trade.open_date_utc`) will remain at the time of the very first order placed.
Please make sure to be aware of this - and eventually adjust your logic in other callbacks to account for this, and use the date of the first filled order instead.
If the cancellation of the original order fails, then the order will not be replaced - though the order will most likely have been canceled on exchange. Having this happen on initial entries will result in the deletion of the order, while on position adjustment orders, it'll result in the trade size remaining as is.
!!! Warning "Regular timeout"
Entry `unfilledtimeout` mechanism (as well as `check_entry_timeout()`) takes precedence over this.
Entry Orders that are cancelled via the above methods will not have this callback called. Be sure to update timeout values to match your expectations.
@@ -168,10 +168,12 @@ Most indicators have an instable startup period, in which they are either not av
To account for this, the strategy can be assigned the `startup_candle_count` attribute.
This should be set to the maximum number of candles that the strategy requires to calculate stable indicators. In the case where a user includes higher timeframes with informative pairs, the `startup_candle_count` does not necessarily change. The value is the maximum period (in candles) that any of the informatives timeframes need to compute stable indicators.
In this example strategy, this should be set to 100 (`startup_candle_count = 100`), since the longest needed history is 100 candles.
You can use [recursive-analysis](recursive-analysis.md) to check and find the correct `startup_candle_count` to be used.
In this example strategy, this should be set to 400 (`startup_candle_count = 400`), since the minimum needed history for ema100 calculation to make sure the value is correct is 400 candles.
Assuming `startup_candle_count` is set to 100, backtesting knows it needs 100 candles to generate valid buy signals. It will load data from `20190101 - (100 * 5m)` - which is ~2018-12-31 15:30:00.
Assuming `startup_candle_count` is set to 400, backtesting knows it needs 400 candles to generate valid buy signals. It will load data from `20190101 - (400 * 5m)` - which is ~2018-12-30 11:40:00.
If this data is available, indicators will be calculated with this extended timerange. The instable startup period (up to 2019-01-01 00:00:00) will then be removed before starting backtesting.
!!! Note
If data for the startup period is not available, then the timerange will be adjusted to account for this startup period - so Backtesting would start at 2019-01-01 08:30:00.
If data for the startup period is not available, then the timerange will be adjusted to account for this startup period - so Backtesting would start at 2019-01-02 09:20:00.
Edit the method `populate_exit_trend()` into your strategy file to update your exit strategy.
The exit-signal is only used for exits if `use_exit_signal` is set to true in the configuration.
The exit-signal can be suppressed by setting `use_exit_signal` to false in the configuration or strategy.
`use_exit_signal` will not influence [signal collision rules](#colliding-signals) - which will still apply and can prevent entries.
It's important to always return the dataframe without removing/modifying the columns `"open", "high", "low", "close", "volume"`, otherwise these fields would contain something unexpected.
@@ -342,16 +344,12 @@ The above configuration would therefore mean:
The calculation does include fees.
To disable ROI completely, set it to an insanely high number:
To disable ROI completely, set it to an empty dictionary:
```python
minimal_roi = {
"0": 100
}
minimal_roi = {}
```
While technically not completely disabled, this would exit once the trade reaches 10000% Profit.
To use times based on candle duration (timeframe), the following snippet can be handy.
This will allow you to change the timeframe for the strategy, and ROI times will still be set as candles (e.g. after 3 candles ...)
@@ -590,6 +588,67 @@ for more information.
will overwrite previously defined method and not produce any errors due to limitations of Python programming language. In such cases you will find that indicators
created in earlier-defined methods are not available in the dataframe. Carefully review method names and make sure they are unique!
### *merge_informative_pair()*
This method helps you merge an informative pair to a regular dataframe without lookahead bias.
It's there to help you merge the dataframe in a safe and consistent way.
Options:
- Rename the columns for you to create unique columns
- Merge the dataframe without lookahead bias
- Forward-fill (optional)
For a full sample, please refer to the [complete data provider example](#complete-data-provider-sample) below.
All columns of the informative dataframe will be available on the returning dataframe in a renamed fashion:
!!! Example "Column renaming"
Assuming `inf_tf = '1d'` the resulting columns will be:
``` python
'date', 'open', 'high', 'low', 'close', 'rsi' # from the original dataframe
'date_1d', 'open_1d', 'high_1d', 'low_1d', 'close_1d', 'rsi_1d' # from the informative dataframe
```
??? Example "Column renaming - 1h"
Assuming `inf_tf = '1h'` the resulting columns will be:
``` python
'date', 'open', 'high', 'low', 'close', 'rsi' # from the original dataframe
'date_1h', 'open_1h', 'high_1h', 'low_1h', 'close_1h', 'rsi_1h' # from the informative dataframe
```
??? Example "Custom implementation"
A custom implementation for this is possible, and can be done as follows:
``` python
# Shift date by 1 candle
# This is necessary since the data is always the "open date"
# and a 15m candle starting at 12:15 should not know the close of the 1h candle from 12:00 to 13:00
# FFill to have the 1d value available in every row throughout the day.
# Without this, comparisons would only work once per day.
dataframe = dataframe.ffill()
```
!!! Warning "Informative timeframe < timeframe"
Using informative timeframes smaller than the dataframe timeframe is not recommended with this method, as it will not use any of the additional information this would provide.
To use the more detailed information properly, more advanced methods should be applied (which are out of scope for freqtrade documentation, as it'll depend on the respective need).
## Additional data (DataProvider)
The strategy provides access to the `DataProvider`. This allows you to get additional data to use in your strategy.
@@ -814,146 +873,6 @@ class SampleStrategy(IStrategy):
***
## Helper functions
### *merge_informative_pair()*
This method helps you merge an informative pair to a regular dataframe without lookahead bias.
It's there to help you merge the dataframe in a safe and consistent way.
Options:
- Rename the columns for you to create unique columns
- Merge the dataframe without lookahead bias
- Forward-fill (optional)
For a full sample, please refer to the [complete data provider example](#complete-data-provider-sample) below.
All columns of the informative dataframe will be available on the returning dataframe in a renamed fashion:
!!! Example "Column renaming"
Assuming `inf_tf = '1d'` the resulting columns will be:
``` python
'date', 'open', 'high', 'low', 'close', 'rsi' # from the original dataframe
'date_1d', 'open_1d', 'high_1d', 'low_1d', 'close_1d', 'rsi_1d' # from the informative dataframe
```
??? Example "Column renaming - 1h"
Assuming `inf_tf = '1h'` the resulting columns will be:
``` python
'date', 'open', 'high', 'low', 'close', 'rsi' # from the original dataframe
'date_1h', 'open_1h', 'high_1h', 'low_1h', 'close_1h', 'rsi_1h' # from the informative dataframe
```
??? Example "Custom implementation"
A custom implementation for this is possible, and can be done as follows:
``` python
# Shift date by 1 candle
# This is necessary since the data is always the "open date"
# and a 15m candle starting at 12:15 should not know the close of the 1h candle from 12:00 to 13:00
# FFill to have the 1d value available in every row throughout the day.
# Without this, comparisons would only work once per day.
dataframe = dataframe.ffill()
```
!!! Warning "Informative timeframe < timeframe"
Using informative timeframes smaller than the dataframe timeframe is not recommended with this method, as it will not use any of the additional information this would provide.
To use the more detailed information properly, more advanced methods should be applied (which are out of scope for freqtrade documentation, as it'll depend on the respective need).
***
### *stoploss_from_open()*
Stoploss values returned from `custom_stoploss` must specify a percentage relative to `current_rate`, but sometimes you may want to specify a stoploss relative to the entry point instead. `stoploss_from_open()` is a helper function to calculate a stoploss value that can be returned from `custom_stoploss` which will be equivalent to the desired trade profit above the entry point.
??? Example "Returning a stoploss relative to the open price from the custom stoploss function"
Say the open price was $100, and `current_price` is $121 (`current_profit` will be `0.21`).
If we want a stop price at 7% above the open price we can call `stoploss_from_open(0.07, current_profit, False)` which will return `0.1157024793`. 11.57% below $121 is $107, which is the same as 7% above $100.
This function will consider leverage - so at 10x leverage, the actual stoploss would be 0.7% above $100 (0.7% * 10x = 7%).
``` python
from datetime import datetime
from freqtrade.persistence import Trade
from freqtrade.strategy import IStrategy, stoploss_from_open
Full examples can be found in the [Custom stoploss](strategy-advanced.md#custom-stoploss) section of the Documentation.
!!! Note
Providing invalid input to `stoploss_from_open()` may produce "CustomStoploss function did not return valid stoploss" warnings.
This may happen if `current_profit` parameter is below specified `open_relative_stop`. Such situations may arise when closing trade
is blocked by `confirm_trade_exit()` method. Warnings can be solved by never blocking stop loss sells by checking `exit_reason` in
`confirm_trade_exit()`, or by using `return stoploss_from_open(...) or 1` idiom, which will request to not change stop loss when
`current_profit <open_relative_stop`.
### *stoploss_from_absolute()*
In some situations it may be confusing to deal with stops relative to current rate. Instead, you may define a stoploss level using an absolute price.
??? Example "Returning a stoploss using absolute price from the custom stoploss function"
If we want to trail a stop price at 2xATR below current price we can call `stoploss_from_absolute(current_rate-(candle['atr']*2),current_rate,is_short=trade.is_short)`.
``` python
from datetime import datetime
from freqtrade.persistence import Trade
from freqtrade.strategy import IStrategy, stoploss_from_absolute
The strategy provides access to the `wallets` object. This contains the current balances on the exchange.
@@ -1089,6 +1008,10 @@ The following lists some common patterns which should be avoided to prevent frus
- don't use `dataframe['volume'].mean()`. This uses the full DataFrame for backtesting, including data from the future. Use `dataframe['volume'].rolling(<window>).mean()` instead
- don't use `.resample('1h')`. This uses the left border of the interval, so moves data from an hour to the start of the hour. Use `.resample('1h', label='right')` instead.
!!! Tip "Identifying problems"
You may also want to check the 2 helper commands [lookahead-analysis](lookahead-analysis.md) and [recursive-analysis](recursive-analysis.md), which can each help you figure out problems with your strategy in different ways.
Please treat them as what they are - helpers to identify most common problems. A negative result of each does not guarantee that there's none of the above errors included.
### Colliding signals
When conflicting signals collide (e.g. both `'enter_long'` and `'exit_long'` are 1), freqtrade will do nothing and ignore the entry signal. This will avoid trades that enter, and exit immediately. Obviously, this can potentially lead to missed entries.
Required function to set the targets for the model.
@@ -728,3 +729,86 @@ Targets now get their own, dedicated method.
return dataframe
```
### FreqAI - New data Pipeline
If you have created your own custom `IFreqaiModel` with a custom `train()`/`predict()` function, *and* you still rely on `data_cleaning_train/predict()`, then you will need to migrate to the new pipeline. If your model does *not* rely on `data_cleaning_train/predict()`, then you do not need to worry about this migration. That means that this migration guide is relevant for a very small percentage of power-users. If you stumbled upon this guide by mistake, feel free to inquire in depth about your problem in the Freqtrade discord server.
The conversion involves first removing `data_cleaning_train/predict()` and replacing them with a `define_data_pipeline()` and `define_label_pipeline()` function to your `IFreqaiModel` class:
1. Data normalization and cleaning is now homogenized with the new pipeline definition. This is created in the new `define_data_pipeline()` and `define_label_pipeline()` functions. The `data_cleaning_train()` and `data_cleaning_predict()` functions are no longer used. You can override `define_data_pipeline()` to create your own custom pipeline if you wish.
2. Data normalization and cleaning is now homogenized with the new pipeline definition. This is created in the new `define_data_pipeline()` and `define_label_pipeline()` functions. The `data_cleaning_train()` and `data_cleaning_predict()` functions are no longer used. You can override `define_data_pipeline()` to create your own custom pipeline if you wish.
3. Data denormalization is done with the new pipeline. Replace this with the lines below.
@@ -187,11 +187,13 @@ official commands. You can ask at any moment for help with `/help`.
| `/forcelong <pair> [rate]` | Instantly buys the given pair. Rate is optional and only applies to limit orders. (`force_entry_enable` must be set to True)
| `/forceshort <pair> [rate]` | Instantly shorts the given pair. Rate is optional and only applies to limit orders. This will only work on non-spot markets. (`force_entry_enable` must be set to True)
| `/delete <trade_id>` | Delete a specific trade from the Database. Tries to close open orders. Requires manual handling of this trade on the exchange.
| `/reload_trade <trade_id>` | Reload a trade from the Exchange. Only works in live, and can potentially help recover a trade that was manually sold on the exchange.
| `/cancel_open_order <trade_id> | /coo <trade_id>` | Cancel an open order for a trade.
| **Metrics** |
| `/profit [<n>]` | Display a summary of your profit/loss from close trades and some stats about your performance, over the last n days (all trades by default)
| `/performance` | Show performance of each finished trade grouped by pair
| `/balance` | Show account balance per currency
| `/balance` | Show bot managed balance per currency
| `/balance full` | Show account balance per currency
| `/daily <n>` | Shows profit or loss per day, over the last n days (n defaults to 7)
| `/weekly <n>` | Shows profit or loss per week, over the last n weeks (n defaults to 8)
| `/monthly <n>` | Shows profit or loss per month, over the last n months (n defaults to 6)
@@ -202,7 +204,6 @@ official commands. You can ask at any moment for help with `/help`.
| `/blacklist [pair]` | Show the current blacklist, or adds a pair to the blacklist.
| `/edge` | Show validated pairs by Edge if it is enabled.
## Telegram commands in action
Below, example of Telegram message you will receive for each command.
@@ -279,19 +280,26 @@ Return a summary of your profit/loss and performance.
> ∙ `33.095 EUR`
>
> **Total Trade Count:** `138`
> **Bot started:** `2022-07-11 18:40:44`
> **First Trade opened:** `3 days ago`
> **Latest Trade opened:** `2 minutes ago`
> **Avg. Duration:** `2:33:45`
> **Best Performing:** `PAY/BTC: 50.23%`
> **Trading volume:** `0.5 BTC`
> **Profit factor:** `1.04`
> **Win / Loss:** `102 / 36`
> **Winrate:** `73.91%`
> **Expectancy (Ratio):** `4.87 (1.66)`
> **Max Drawdown:** `9.23% (0.01255 BTC)`
The relative profit of `1.2%` is the average profit per trade.
The relative profit of `15.2 Σ%` is be based on the starting capital - so in this case, the starting capital was `0.00485701 * 1.152 = 0.00738 BTC`.
Starting capital is either taken from the `available_capital` setting, or calculated by using current wallet size - profits.
Profit Factor is calculated as gross profits / gross losses - and should serve as an overall metric for the strategy.
Expectancy corresponds to the average return per currency unit at risk, i.e. the winrate and the risk-reward ratio (the average gain of winning trades compared to the average loss of losing trades).
Expectancy Ratio is expected profit or loss of a subsequent trade based on the performance of all past trades.
Max drawdown corresponds to the backtesting metric `Absolute Drawdown (Account)` - calculated as `(Absolute Drawdown) / (DrawdownHigh + startingBalance)`.
Bot started date will refer to the date the bot was first started. For older bots, this will default to the first trade's open date.
@@ -141,7 +141,8 @@ Most properties here can be None as they are dependant on the exchange response.
`amount` | float | Amount in base currency
`filled` | float | Filled amount (in base currency)
`remaining` | float | Remaining amount
`cost` | float | Cost of the order - usually average * filled
`cost` | float | Cost of the order - usually average * filled (*Exchange dependant on futures, may contain the cost with or without leverage and may be in contracts.*)
`stake_amount` | float | Stake amount used for this order. *Added in 2023.7.*
`order_date` | datetime | Order creation date **use `order_date_utc` instead**
`order_date_utc` | datetime | Order creation date (in UTC)
`order_fill_date` | datetime | Order fill date **use `order_fill_utc` instead**
@@ -80,12 +80,18 @@ When using the Form-Encoded or JSON-Encoded configuration you can configure any
The result would be a POST request with e.g. `Status: running` body and `Content-Type: text/plain` header.
Optional parameters are available to enable automatic retries for webhook messages. The `webhook.retries` parameter can be set for the maximum number of retries the webhook request should attempt if it is unsuccessful (i.e. HTTP response status is not 200). By default this is set to `0` which is disabled. An additional `webhook.retry_delay` parameter can be set to specify the time in seconds between retry attempts. By default this is set to `0.1` (i.e. 100ms). Note that increasing the number of retries or retry delay may slow down the trader if there are connectivity issues with the webhook. Example configuration for retries:
## Additional configurations
The `webhook.retries` parameter can be set for the maximum number of retries the webhook request should attempt if it is unsuccessful (i.e. HTTP response status is not 200). By default this is set to `0` which is disabled. An additional `webhook.retry_delay` parameter can be set to specify the time in seconds between retry attempts. By default this is set to `0.1` (i.e. 100ms). Note that increasing the number of retries or retry delay may slow down the trader if there are connectivity issues with the webhook.
You can also specify `webhook.timeout` - which defines how long the bot will wait until it assumes the other host as unresponsive (defaults to 10s).
Example configuration for retries:
```json
"webhook":{
"enabled":true,
"url":"https://<YOURHOOKURL>",
"timeout":10,
"retries":3,
"retry_delay":0.2,
"status":{
@@ -109,6 +115,8 @@ Custom messages can be sent to Webhook endpoints via the `self.dp.send_msg()` fu
Different payloads can be configured for different events. Not all fields are necessary, but you should configure at least one of the dicts, otherwise the webhook will never be called.
## Webhook Message types
### Entry
The fields in `webhook.entry` are filled when the bot executes a long/short. Parameters are filled using string.format.
Install ta-lib according to the [ta-lib documentation](https://github.com/mrjbq7/ta-lib#windows).
As compiling from source on windows has heavy dependencies (requires a partial visual studio installation), there is also a repository of unofficial pre-compiled windows Wheels [here](https://www.lfd.uci.edu/~gohlke/pythonlibs/#ta-lib), which need to be downloaded and installed using `pip install TA_Lib-0.4.25-cp38-cp38-win_amd64.whl` (make sure to use the version matching your python version).
As compiling from source on windows has heavy dependencies (requires a partial visual studio installation), Freqtrade provides these dependencies (in the binary wheel format) for the latest 3 Python versions (3.9, 3.10 and 3.11) and for 64bit Windows.
These Wheels are also used by CI running on windows, and are therefore tested together with freqtrade.
Freqtrade provides these dependencies for the latest 3 Python versions (3.8, 3.9, 3.10 and 3.11) and for 64bit Windows.
Other versions must be downloaded from the above link.
``` powershell
cd \path\freqtrade
python -m venv .env
.env\Scripts\activate.ps1
python -m venv .venv
.venv\Scripts\activate.ps1
# optionally install ta-lib from wheel
# Eventually adjust the below filename to match the downloaded wheel
pip install --find-links build_helpers\ TA-Lib -U
@@ -45,8 +45,6 @@ freqtrade
The above installation script assumes you're using powershell on a 64bit windows.
Commands for the legacy CMD windows console may differ.
> Thanks [Owdr](https://github.com/Owdr) for the commands. Source: [Issue #222](https://github.com/freqtrade/freqtrade/issues/222)
Some files were not shown because too many files have changed in this diff
Show More
Reference in New Issue
Block a user
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
