Systematic Random Forest Calibration Trading

The strategy involves a systematic trading pipeline that trains a Random Forest model on historical Polymarket data and external features to predict the true probability of event outcomes. When the model's calibrated probability deviates significantly from the market's implied probability, the bot enters a position (long the undervalued outcome, short the overvalued). Entry is based on a confidence threshold derived from model calibration; exit occurs at resolution or when the edge disappears. The pipeline includes backtesting and automated paper trading. Results from historical testing show the model can identify consistent edges in certain market categories.