Multi-Timeframe Linear Regression with Weighted Signal Validation
Summary
This strategy combines short-, medium-, and long-horizon linear regressions into a directional trading signal. It evaluates each regression using slope, R-squared, and correlation thresholds, then weights the three outputs, with the short horizon receiving the greatest stated weight. A confidence measure also reflects statistical strength, agreement among timeframes, and historical prediction accuracy. Position size is described as confidence-sensitive, with a daily loss limit intended to halt trading.
The document outlines conditions for entries and exits and gives example settings, including regression windows of 20, 50, and 100 periods and a daily loss cap. It also provides source fragments, but no complete empirical results or performance metrics. Its own caveats include parameter sensitivity, lag during sharp reversals, changing market conditions, overfitting, and computational demands. The proposed responses include forward testing, cross-validation, regime classification, and volatility-aware risk controls; these are recommendations, not evidence that the strategy is profitable or robust.
Key ideas
- Three linear regressions across different horizons are combined using weighted signals.
- Regression significance is screened using slope, R-squared, and correlation thresholds.
- Signal confidence incorporates statistical strength, agreement across horizons, and historical validation.
- Position sizing and a daily loss limit are intended to adjust and constrain risk.
- The document supplies design details but no performance results, and it flags lag, overfitting, and parameter sensitivity.
Tags
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.