Using Catch22 Features to Classify Market Volatility Regimes

Summary

The article explains catch22, a compact set of time-series features selected from a much larger library for classification performance and low redundancy. Its features describe distribution shape, autocorrelation, nonlinear dependence, symbolic patterns, scaling, periodicity, and predictability. The author outlines a native MQL5 implementation that computes the full feature vector and checks it against the Python reference implementation.

The proposed trading experiment classifies upcoming volatility regimes using a leak-free dataset, compares catch22 features with classic indicators and their combination, and uses the resulting model as a filter for a basic strategy. The article describes validation and evaluation procedures, but the provided excerpt omits most detailed results, so it does not support a strong conclusion about predictive or trading value. Regime classification is presented as a more suitable use than direct price-direction forecasting; results may also depend on the chosen market, labeling, features, and test design.

Key ideas

• Catch22 offers a compact set of time-series descriptors selected to reduce redundancy while preserving classification usefulness.
• The feature families capture distribution, dependence, spectral, symbolic, scaling, and predictability properties.
• The MQL5 implementation is checked feature by feature against a Python reference.
• The experiment evaluates whether catch22 adds value beyond conventional indicators for volatility-regime classification.
• A regime filter can be tested within a strategy, but the excerpt provides limited evidence about its results.

Tags

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.