Combining Indicator Signals with a Naive Bayes Trading Model

Summary

This article presents a probabilistic framework for combining technical indicator signals. It treats a profitable entry as the outcome of interest and estimates each indicator’s historical hit rate, then uses a naive Bayes formulation to estimate the combined probability. The setup simplifies evaluation by using bar data, a fixed holding period, no take-profit or stop-loss, unchanged position size, and only successful versus unsuccessful entries. It omits the wait state and combines buy and sell outcomes into a general entry hypothesis. The article reports that individual standard indicators tend to have hit rates near chance, with stated values typically in the 0.51–0.55 range, and gives theoretical examples of how combining signals could raise estimated accuracy. It also describes practical testing and claims the observed results align with the theoretical calculation, though portions of the test evidence are absent from the excerpt. The key limitation is the conditional independence assumption: correlated indicators can make the theoretical benefit too optimistic. Spread, trade management, signal timing, and other system parameters also require further modeling.

Key ideas

• The model estimates the probability of a successful entry from historical indicator signal outcomes.
• Naive Bayes combines individual indicator probabilities under an assumption of conditional independence.
• The illustrative trading setup fixes holding time and excludes stops, targets, and position-size changes.
• Correlated indicators can reduce the real benefit of combining signals below theoretical estimates.
• The framework can be extended to model indicator states, trade management, and timing tolerance, with added complexity.

Tags

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