ML Models & Strategy Layer

ML models (5)

Model	Type	Role
Direction Predictor	Gradient-boosted ensemble (LightGBM + XGBoost + CatBoost)	Binary direction prediction
Regime Detector	Gaussian Hidden Markov Model	Classifies market state: trending, mean-reverting, volatile, quiet
Cascade Predictor	LightGBM + CatBoost	Liquidation-cascade risk blocking
DeepLOB	CNN + Transformer	Limit-order-book microstructure signal (low ensemble weight, high-frequency signal)
RL Meta-Optimizer	Soft Actor-Critic (SAC)	Dynamic position-sizing and parameter tuning against live performance

Training runs through a 16-module pipeline (training/) covering drift monitoring, daily retraining, multi-symbol orchestration, walk-forward validation, and backtesting — model updates are not manual, one-off retrains.

Strategy modules (16)

Rather than running 16 independent bots, the platform coordinates 16 strategy modules through a consensus layer:

• antifragile.py — meta-framework orchestrating barbell allocation across the ensemble
• strategy_mux.py — multi-strategy consensus and weighted voting
• decision_engine.py — risk gate and signal fusion pipeline
• decision_council.py — individual model voting with veto logic
• confluence_engine.py — multi-timeframe confluence scoring
• regime_adaptive.py — regime-aware parameter shifting
• rl_agent.py — the SAC meta-optimizer described above
• auto_evolve.py — live edge-decay tracking and parameter evolution
• model_orchestra.py — ensemble inference coordinator
• microstructure_filter.py — order-book toxicity and execution-quality guard
• tail_risk.py — fat-tail hedge positioning
• opportunity_scorer.py, vol_harvester.py, guerrilla_integration.py, confluence_engine.py, profiles.py — signal ranking, volatility-reversion capture, tactical cascade handling, and per-profile parameterization

Why ensemble, not single-model

A single model failing or drifting degrades gracefully rather than taking the whole system down — the consensus layer down-weights underperforming signals continuously (auto_evolve.py) instead of requiring a manual intervention to notice and fix. This is the same principle documented in Engineering Practice: the system is built to keep functioning without a person in the loop for routine degradation.