Machine learning · Advanced · Weekly — model retraining recommended quarterly
ML-Enhanced RSI
How it works, and when it doesn't.
Not all RSI signals are equal. The ML layer learns which market conditions produce reliable bounces vs. which RSI dips lead to further declines.
Why it works
ML-Enhanced RSI uses a gradient boosted classifier to predict whether an RSI < 30 signal will result in a profitable bounce within 5 days. The model is trained on 10 years of features including VIX, volume, sector breadth, and multi-timeframe RSI.
The signal logic
- Same base signal as RSI Oscillator MR (strategy #10)
- Trained XGBoost classifier filters signals by 5-day forward return probability
- Only trade signals where model confidence > 65% for a profitable bounce
When it fails
Out-of-sample regime changes that weren't in the training data — particularly rapid macro regime shifts the model hasn't seen before.
- Best market
- The ML model performs best in the regimes it was trained on
- Worst market
- Novel market conditions unlike anything in training data
Backtest
This is a research tool rather than a directional strategy; it has no live signal of its own.
ML-Enhanced RSI, live.
Re-scored after every close.
The platform runs this and 32 other strategies against the live tape each day — with the evidence behind every number.