Systematic ML Calibration (Random Forest)

This strategy (BallesJr/polymarket-edge-model) uses a Random Forest model trained on historical Polymarket price data and external features (e.g., news sentiment, volume) to predict short-term price direction. The model is calibrated daily with a walk-forward validation approach. Entry signals are generated when model confidence exceeds 70% threshold. The bot runs in paper-trading mode with automated execution. Position sizing uses fractional Kelly. Backtesting uses 6 months of historical data, claiming a 58% win rate with 1.2 Sharpe ratio. Exit is based on model re-evaluation every 5 minutes or at stop-loss. Risk includes overfitting and regime change. Implementation includes a full pipeline from data collection to trade execution.