Learning RSI Trading Zones with a Markov-Based Probabilistic Model

Summary

This article describes a data-driven way to choose RSI entry zones for a forex strategy. It groups RSI readings into ten bands, labels whether price rises or falls over a future horizon, and aggregates those outcomes by band to estimate likely direction. The author trains on part of a 300,000-row, one-minute NZDJPY dataset and evaluates the zones on held-out data. While the most favorable bands in training did not prove reliable in validation, the 11–20 band for buys and 71–80 band for sells showed the strongest reported validation accuracy for their respective sides.

A greedy model based on the estimated probabilities achieved 52% test accuracy, while those selected zones had validation accuracies of 51.4% and 75.8%. The article then describes implementing the rules in an MQL5 Expert Advisor, with exits based either on RSI zone changes or price crossing a moving average. These are classification accuracy results, not net trading returns; the text gives limited detail on costs, robustness across markets, or the interpretation of overlapping future labels. One RSI band had no training observations and was assigned an arbitrary negative value.

Key ideas

• The method groups RSI readings into ten bands and records the direction of price movement over a future horizon.
• A training transition summary is used to estimate which RSI bands tend to precede price appreciation or depreciation.
• The strongest training bands did not remain the best choices in validation, so the author selected zones using held-out data.
• The selected NZDJPY buy and sell zones had different validation accuracies, and the greedy model's test accuracy was 52%.
• The reported metrics measure directional classification accuracy and do not establish profitability after trading costs.

Tags

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