🧠 Loading live leaderboard...
One real machine learning fighter on BTC. A calibrated Random Forest learns from Binance OHLCV bars and is evaluated with Brier-aware metrics aligned with the Calibration paper.
≠ AI Arena (LLM decision-makers · here: calibrated Random Forest)
Calibrated Random Forest trained on real OHLCV market data
Real machine learning fighter (calibrated Random Forest) trained on Binance OHLCV data. Brier-aware évaluation aligned with the Calibration paper.
ML Arena V3 exposes one machine learning model: a calibrated Random Forest trained on Binance OHLCV data. The decision rule uses 4h upside probability and the primary évaluation metric is Brier score, coherent with the Calibration paper.
See the complète AI vs ML benchmark and the free training course.
Affiliate links — we may earn a small commission
ML Arena V3 runs a single calibrated Random Forest on real Bitcoin OHLCV data. It uses walk-forward time-series évaluation, a 48-bar embargo, isotonic calibration, and Brier score monitoring.
The fighter stays in cold start until enough 5-minute bars are available for training.
XGBoost gagne encore 80% des competitions Kaggle quant. Le deep learning est sexy mais perd souvent face a un GradientBoosting bien tune sur features.