improve training and model data

utils/reward_calculator.py

@@ -23,7 +23,7 @@ class RewardCalculator:
         self.trade_timestamps = []
         self.frequency_threshold = 10  # Trades per minute threshold for penalty
         self.max_frequency_penalty = 0.05
-
+    
     def record_pnl(self, pnl):
         """Record P&L for risk adjustment calculations"""
         self.trade_pnls.append(pnl)
@@ -36,7 +36,7 @@ class RewardCalculator:
         self.trade_timestamps.append(time())
         if len(self.trade_timestamps) > 100:
             self.trade_timestamps.pop(0)
-
+    
     def _calculate_frequency_penalty(self):
         """Calculate penalty for high-frequency trading"""
         if len(self.trade_timestamps) < 2:
@@ -47,9 +47,9 @@ class RewardCalculator:
         trades_per_minute = (len(self.trade_timestamps) / time_span) * 60
         if trades_per_minute > self.frequency_threshold:
             penalty = min(self.max_frequency_penalty, (trades_per_minute - self.frequency_threshold) * 0.001)
-            return penalty
+        return penalty
         return 0.0
-
+    
     def _calculate_risk_adjustment(self, reward):
         """Adjust rewards based on risk (simple Sharpe ratio implementation)"""
         if len(self.trade_pnls) < 5:
@@ -62,7 +62,7 @@ class RewardCalculator:
         sharpe = mean_return / std_return
         adjustment_factor = np.clip(1.0 + 0.5 * sharpe, 0.5, 2.0)
         return reward * adjustment_factor
-
+    
     def _calculate_holding_reward(self, position_held_time, price_change):
         """Calculate reward for holding a position"""
         base_holding_reward = 0.0005 * (position_held_time / 60.0)