Fitting Forecast Weights with Recency, Alpha, and Joint Optimization
Summary
The document explores three changes to fitting trading-system weights: exponential weighting that emphasizes recent performance, evaluating alpha rather than Sharpe ratio alone, and jointly fitting instrument and forecast weights. The motivation is that old performance can obscure strategy decay, while Sharpe ratio may reward positive market beta that an investor already holds elsewhere. For recency weighting, the author proposes a 15-year half-life for estimating means and Sharpe ratios, retaining older observations while emphasizing more recent ones.
For alpha-based fitting, the author discusses benchmark choice and favors a diversified long-only futures portfolio as a compromise. The method estimates betas on rolling and then expanding 30-year windows, updates them annually, and applies a 15-year half-life to residual returns before using them in weight fitting. Reported comparisons show different tradeoffs in Sharpe ratio, beta, and residual performance across approaches; the excerpt also describes a shift away from faster momentum toward strategies that performed better recently. The source is incomplete, and the displayed tables and conclusions do not establish broad out-of-sample robustness.
Key ideas
- Recent performance can receive greater weight through exponential weighting while older observations retain influence.
- Sharpe ratio based fitting may favor strategies with positive market beta, motivating an alpha based comparison.
- The choice of benchmark changes the estimated alpha and should reflect the investor’s broader portfolio.
- The described approach estimates beta over long windows and weights residual performance with a shorter half-life.
- The reported portfolio comparisons involve tradeoffs, and the excerpt does not establish general robustness.
Tags
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.