Filtering and Sizing ADX Crossover Trades with Meta-Labeling
Summary
The article explains why directional indicator crossovers can generate repeated losing entries in range-bound markets, then presents a two-layer filter for ADX trades. The first layer replaces a fixed ADXR regime threshold with a gate optimized on validation data. The second uses a Random Forest classifier and eleven ADX-derived context features to assess gated signals; position size is then scaled by classifier confidence. It also details Wilder's directional movement, true range, smoothing, ADX and ADXR calculations, plus the crossover entry and protective extreme-point rule.
The proposed workflow uses EURUSD hourly data over seven years, triple-barrier labels and walk-forward validation, with both time bars and an alternative tick-based bar set described. The excerpt emphasizes that a single regime threshold cannot represent context such as ADX slope, DI separation, dominance duration and volatility. Its claims depend on the specified data, labeling, optimization and validation design; the supplied material does not provide the final numerical results needed to assess the strength or robustness of the reported comparison.
Key ideas
- DI crossovers can whipsaw when directional movement lacks persistence.
- Wilder's ADXR threshold is a regime filter, while the proposed first layer optimizes a gate using validation data.
- A Random Forest evaluates contextual ADX features for signals that pass the gate.
- Classifier confidence is used to scale position size.
- Walk-forward results depend on the data, labeling and validation choices and require careful interpretation.
Tags
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.