Add ProfitDrawdownHyperoptLoss method

freqtrade/constants.py

@@ -26,7 +26,7 @@ HYPEROPT_LOSS_BUILTIN = ['ShortTradeDurHyperOptLoss', 'OnlyProfitHyperOptLoss',
                          'SharpeHyperOptLoss', 'SharpeHyperOptLossDaily',
                          'SortinoHyperOptLoss', 'SortinoHyperOptLossDaily',
                          'CalmarHyperOptLoss',
-                         'MaxDrawDownHyperOptLoss']
+                         'MaxDrawDownHyperOptLoss', 'ProfitDrawDownHyperOptLoss']
 AVAILABLE_PAIRLISTS = ['StaticPairList', 'VolumePairList',
                        'AgeFilter', 'OffsetFilter', 'PerformanceFilter',
                        'PrecisionFilter', 'PriceFilter', 'RangeStabilityFilter',

freqtrade/optimize/hyperopt_loss_profit_drawdown.py

@@ -0,0 +1,29 @@
+"""
+ProfitDrawDownHyperOptLoss
+
+This module defines the alternative HyperOptLoss class based on Profit &
+Drawdown objective which can be used for Hyperoptimization.
+
+Possible to change `DRAWDOWN_MULT` to penalize drawdown objective for
+individual needs.
+"""
+from pandas import DataFrame
+from freqtrade.optimize.hyperopt import IHyperOptLoss
+from freqtrade.data.btanalysis import calculate_max_drawdown
+
+# higher numbers penalize drawdowns more severely
+DRAWDOWN_MULT = 0.075
+
+
+class ProfitDrawDownHyperOptLoss(IHyperOptLoss):
+    @staticmethod
+    def hyperopt_loss_function(results: DataFrame, trade_count: int, *args, **kwargs) -> float:
+        total_profit = results["profit_abs"].sum()
+
+        # from freqtrade.optimize.optimize_reports.generate_strategy_stats()
+        try:
+            _, _, _, _, max_drawdown_per = calculate_max_drawdown(results, value_col="profit_ratio")
+        except ValueError:
+            max_drawdown_per = 0
+
+        return -1 * (total_profit * (1 - max_drawdown_per * DRAWDOWN_MULT))