Using Supervised Learning to Generate Darvas Box Breakout Signals
Summary
The article uses a Darvas box breakout as a case study for supervised machine learning in algorithmic trading. It distinguishes generating trade signals with a model from using a model merely to filter trades produced by an existing strategy. The workflow collects market features around box breakouts, trains a classifier, and uses predicted buy or sell confidence to trigger trades when it exceeds a threshold. Features include technical indicators, recent price behavior, and distances to the box boundaries. The discussion also proposes continuous confidence-based signals and combining predictions from models trained on different timeframes.
The author reports that, in this example, continuous signals improved backtest performance as confidence rose, but provides no detailed metrics in the supplied text. Direct signal generation may increase the available training sample, yet it also makes performance depend on the model’s predictive ability rather than retaining a backbone strategy’s edge. The article treats its techniques as exploratory, notes that model training details are incomplete, and cautions that outcomes depend on the scenario; it does not establish general effectiveness.
Key ideas
- Supervised learning can generate trades directly or filter signals from an existing strategy.
- Darvas box boundaries provide breakout triggers for collecting features and making predictions.
- Confidence thresholds can turn model probabilities into buy or sell decisions.
- The example reports better backtest performance from continuous confidence signals, without detailed results in the supplied text.
- Direct signal generation offers more samples but depends on the model’s predictive power.
Tags
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.