ML Models

Train direction classifiers with walk-forward validation and inspect predictive signals.

Markets (select one or more — multi-market models learn shared patterns)

Timeframes (select up to 4 — samples are pooled across markets & timeframes)

Optimize hyperparameters & indicator parameters (auto)Learn from last training (warm start)Auto-train until accuracy target

Trains a logistic-regression direction classifier on 17 engineered features (multi-horizon momentum, trend distance, ADX, stochastic, Williams %R, volatility, volume) with class-weighted, L2-regularized walk-forward validation. Reports out-of-sample accuracy, AUC, F1, a confusion matrix, and a long/flat equity simulation. Requires 160+ bars of history. Research signals only — not investment advice.

Trained Models

No models yet — train your first classifier above.