diff --git a/docs/hyperopt.md b/docs/hyperopt.md index ef3d28188..2755cae2d 100644 --- a/docs/hyperopt.md +++ b/docs/hyperopt.md @@ -156,10 +156,10 @@ Each hyperparameter tuning requires a target. This is usually defined as a loss By default, FreqTrade uses a loss function, which has been with freqtrade since the beginning and optimizes mostly for short trade duration and avoiding losses. -A different version this can be used by using the `--hyperopt-loss ` argument. -This class should be in it's own file within the `user_data/hyperopts/` directory. +A different loss function can be specified by using the `--hyperopt-loss ` argument. +This class should be in its own file within the `user_data/hyperopts/` directory. -Currently, the following loss functions are builtin: `SharpeHyperOptLoss` and `DefaultHyperOptLoss`. +Currently, the following loss functions are builtin: `DefaultHyperOptLoss` (default legacy Freqtrade hyperoptimization loss function), `SharpeHyperOptLoss` (optimizes Sharpe Ratio calculated on the trade returns) and `OnlyProfitHyperOptLoss` (which takes only amount of profit into consideration). ### Creating and using a custom loss function diff --git a/freqtrade/optimize/default_hyperopt_loss.py b/freqtrade/optimize/default_hyperopt_loss.py index 2879c4091..4ab9fbe44 100644 --- a/freqtrade/optimize/default_hyperopt_loss.py +++ b/freqtrade/optimize/default_hyperopt_loss.py @@ -3,27 +3,26 @@ DefaultHyperOptLoss This module defines the default HyperoptLoss class which is being used for Hyperoptimization. """ - from math import exp from pandas import DataFrame from freqtrade.optimize.hyperopt import IHyperOptLoss -# Define some constants: -# set TARGET_TRADES to suit your number concurrent trades so its realistic +# Set TARGET_TRADES to suit your number concurrent trades so its realistic # to the number of days TARGET_TRADES = 600 + # This is assumed to be expected avg profit * expected trade count. # For example, for 0.35% avg per trade (or 0.0035 as ratio) and 1100 trades, -# self.expected_max_profit = 3.85 +# expected max profit = 3.85 # Check that the reported Σ% values do not exceed this! # Note, this is ratio. 3.85 stated above means 385Σ%. EXPECTED_MAX_PROFIT = 3.0 -# max average trade duration in minutes -# if eval ends with higher value, we consider it a failed eval +# Max average trade duration in minutes. +# If eval ends with higher value, we consider it a failed eval. MAX_ACCEPTED_TRADE_DURATION = 300 diff --git a/freqtrade/optimize/hyperopt_loss_onlyprofit.py b/freqtrade/optimize/hyperopt_loss_onlyprofit.py new file mode 100644 index 000000000..a1c50e727 --- /dev/null +++ b/freqtrade/optimize/hyperopt_loss_onlyprofit.py @@ -0,0 +1,38 @@ +""" +OnlyProfitHyperOptLoss + +This module defines the alternative HyperOptLoss class which can be used for +Hyperoptimization. +""" +from pandas import DataFrame + +from freqtrade.optimize.hyperopt import IHyperOptLoss + + +# This is assumed to be expected avg profit * expected trade count. +# For example, for 0.35% avg per trade (or 0.0035 as ratio) and 1100 trades, +# expected max profit = 3.85 +# +# Note, this is ratio. 3.85 stated above means 385Σ%, 3.0 means 300Σ%. +# +# In this implementation it's only used in calculation of the resulting value +# of the objective function as a normalization coefficient and does not +# represent any limit for profits as in the Freqtrade legacy default loss function. +EXPECTED_MAX_PROFIT = 3.0 + + +class OnlyProfitHyperOptLoss(IHyperOptLoss): + """ + Defines the loss function for hyperopt. + + This implementation takes only profit into account. + """ + + @staticmethod + def hyperopt_loss_function(results: DataFrame, trade_count: int, + *args, **kwargs) -> float: + """ + Objective function, returns smaller number for better results. + """ + total_profit = results.profit_percent.sum() + return 1 - total_profit / EXPECTED_MAX_PROFIT diff --git a/freqtrade/optimize/hyperopt_loss_sharpe.py b/freqtrade/optimize/hyperopt_loss_sharpe.py index be1a3d4b4..f74b27744 100644 --- a/freqtrade/optimize/hyperopt_loss_sharpe.py +++ b/freqtrade/optimize/hyperopt_loss_sharpe.py @@ -1,8 +1,9 @@ """ -IHyperOptLoss interface -This module defines the interface for the loss-function for hyperopts -""" +SharpeHyperOptLoss +This module defines the alternative HyperOptLoss class which can be used for +Hyperoptimization. +""" from datetime import datetime from pandas import DataFrame @@ -13,8 +14,9 @@ from freqtrade.optimize.hyperopt import IHyperOptLoss class SharpeHyperOptLoss(IHyperOptLoss): """ - Defines the a loss function for hyperopt. - This implementation uses the sharpe ratio calculation. + Defines the loss function for hyperopt. + + This implementation uses the Sharpe Ratio calculation. """ @staticmethod @@ -22,8 +24,9 @@ class SharpeHyperOptLoss(IHyperOptLoss): min_date: datetime, max_date: datetime, *args, **kwargs) -> float: """ - Objective function, returns smaller number for more optimal results - Using sharpe ratio calculation + Objective function, returns smaller number for more optimal results. + + Uses Sharpe Ratio calculation. """ total_profit = results.profit_percent days_period = (max_date - min_date).days diff --git a/freqtrade/tests/optimize/test_hyperopt.py b/freqtrade/tests/optimize/test_hyperopt.py index a588bab64..fc71200a5 100644 --- a/freqtrade/tests/optimize/test_hyperopt.py +++ b/freqtrade/tests/optimize/test_hyperopt.py @@ -337,6 +337,24 @@ def test_sharpe_loss_prefers_higher_profits(default_conf, hyperopt_results) -> N assert under > correct +def test_onlyprofit_loss_prefers_higher_profits(default_conf, hyperopt_results) -> None: + results_over = hyperopt_results.copy() + results_over['profit_percent'] = hyperopt_results['profit_percent'] * 2 + results_under = hyperopt_results.copy() + results_under['profit_percent'] = hyperopt_results['profit_percent'] / 2 + + default_conf.update({'hyperopt_loss': 'OnlyProfitHyperOptLoss'}) + hl = HyperOptLossResolver(default_conf).hyperoptloss + correct = hl.hyperopt_loss_function(hyperopt_results, len(hyperopt_results), + datetime(2019, 1, 1), datetime(2019, 5, 1)) + over = hl.hyperopt_loss_function(results_over, len(hyperopt_results), + datetime(2019, 1, 1), datetime(2019, 5, 1)) + under = hl.hyperopt_loss_function(results_under, len(hyperopt_results), + datetime(2019, 1, 1), datetime(2019, 5, 1)) + assert over < correct + assert under > correct + + def test_log_results_if_loss_improves(hyperopt, capsys) -> None: hyperopt.current_best_loss = 2 hyperopt.log_results(