perf(qav3): fine tune entry signal thresholds

Signed-off-by: Jérôme Benoit <jerome.benoit@piment-noir.org>

diff --git a/quickadapter/user_data/freqaimodels/QuickAdapterRegressorV3.py b/quickadapter/user_data/freqaimodels/QuickAdapterRegressorV3.py
index 18f1751635373a01ea5ced101c9d1586641327f7..d1b1a44505baed72d75fba5e34a364a770f9043d 100644 (file)
--- a/quickadapter/user_data/freqaimodels/QuickAdapterRegressorV3.py
+++ b/quickadapter/user_data/freqaimodels/QuickAdapterRegressorV3.py
@@ -382,11 +382,11 @@ class QuickAdapterRegressorV3(BaseRegressionModel):
         label_period_candles: int,
     ) -> tuple[float, float]:
         temperature = float(
-            self.freqai_info.get("prediction_thresholds_temperature", 125.0)
+            self.freqai_info.get("prediction_thresholds_temperature", 175.0)
         )
         extrema = pred_df[EXTREMA_COLUMN].iloc[
             -(
-                max(2, int((fit_live_predictions_candles / 2) / label_period_candles))
+                max(2, int(fit_live_predictions_candles / label_period_candles))
                 * label_period_candles
             ) :
         ]
@@ -972,7 +972,7 @@ def label_objective(
 
     df = df.iloc[
         -(
-            int(fit_live_predictions_candles / label_period_candles)
+            max(2, int(fit_live_predictions_candles / label_period_candles))
             * label_period_candles
         ) :
     ]