Financial Machine Learning Labels: Triple Barriers and Meta-Labeling
Summary
The article compares fixed-horizon labels with methods designed to represent trading decisions more directly. Triple-barrier labeling sets profit-taking and stop-loss thresholds around an event, plus a time limit; the label depends on which threshold is reached first. The suggested profit and loss barriers scale with estimated volatility, while a vertical barrier can use a bar count for activity-based data. It also introduces trend scanning, which compares candidate forward horizons, and meta-labeling, which assesses whether to act on signals from a primary strategy.
The implementation discussion includes volatility estimation, barrier placement, and a Bollinger Band example with classification reports and ROC curves. The article says its meta-labeling results improve signal quality under aggressive filtering, but the supplied text gives no numerical results or enough detail to assess robustness. Labels depend on barrier and event choices, and the methods require careful data construction to avoid leakage. Trend scanning, sample weighting for overlapping events, and probability-based position sizing are described as future work rather than completed parts of this installment.
Key ideas
- Fixed-horizon labels ignore the price path and whether trading risk limits would have been reached first.
- Triple-barrier labels encode profit, loss, and time limits, with thresholds that can scale to volatility.
- Trend scanning selects among forward horizons based on the strength of the observed trend.
- Meta-labeling evaluates whether to accept signals from a primary strategy and can inform trade sizing.
- The reported classification evidence is qualitative in the available text, so robustness cannot be judged from it alone.
Tags
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.