Logistic Regression for Confidence-Based Bollinger Band Position Sizing

Summary

This article modifies a Bollinger Band trading system by using logistic regression to estimate trade confidence and vary position size. The underlying rules sell when price moves above the upper band and buy when it falls below the lower band, with corresponding exits. The system is tested on GBPUSD using the 15-minute timeframe over a stated historical period. The original strategy uses a fixed lot size; the revised version takes larger positions when the model indicates greater confidence and smaller positions when confidence is lower.

The article reports a backtest comparison: the original system made 493 trades, with 62% profitable, a loss of $813, and a Sharpe ratio of -0.33. The revised version made 495 trades, with 63% profitable, profit of $2,427, and a Sharpe ratio of 0.74. These are reported results for one pair, timeframe, and period, not evidence of general performance. The method also increases exposure on selected trades, so sizing risk and the need for further validation remain central considerations.

Key ideas

• The strategy buys below the lower Bollinger Band and sells above the upper band, using band breaches as exits as well.
• A logistic regression model estimates confidence to choose between larger and smaller position sizes.
• The reported backtest shows improved profit and Sharpe ratio versus the fixed-size baseline in its tested GBPUSD sample.
• The reported results cover one market, timeframe, and historical period and do not establish broader robustness.
• Confidence-based sizing changes exposure and should be evaluated alongside drawdown and loss behavior.

Tags

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