Adaptive Technical Indicators Using Least Squares and Price-Weighted Windows
Summary
The article distinguishes genuinely adaptive indicators from filters that only appear to adapt. Averaging recent forecast errors changes the effective linear weights, but the resulting indicator is still a fixed linear combination of past prices. Laplace smoothing and a price-domain weighting scheme are explored as further attempts; they adjust to current prices but do not establish feedback-driven adaptation.
The proposed adaptive method updates indicator coefficients to reduce the difference between the indicator and price while limiting how much the coefficients change. A convergence parameter controls the update speed and smoothness, with a suggested starting point tied to the indicator period. The article also describes constructing stable indicators from price movement and finite differences, with coefficient constraints. These are mathematical design examples, not evidence of trading profitability: no market tests or performance results are provided, and parameter choice affects behavior.
Key ideas
- Averaging recent indicator errors can be rewritten as a fixed linear filter, so error correction alone does not make an indicator adaptive.
- Price-based window weights respond to the current price distribution but are better described as market-adjusted than adaptive.
- The proposed least-squares approach updates coefficients to reduce price error while discouraging large coefficient changes.
- A convergence parameter trades faster coefficient adjustment against smoother, slower changes.
- Finite differences can represent discrete price velocity and acceleration in forecast-style indicator formulas.
Tags
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.