ETF Rotation with Trend Scoring and Mean-Variance Allocation
Summary
This strategy rotates among a fixed pool of 11 ETFs spanning several asset classes and markets. Every five trading days, it ranks candidates using a 25-day linear-regression slope multiplied by R-squared, then adds a short-term signal based on the five-day and ten-day moving-average ratio. It selects the six highest-scoring ETFs and removes any held ETF whose 20-day gain exceeds the stated take-profit threshold.
The selected assets receive weights from a mean-variance optimization using historical returns, subject to per-ETF minimum and maximum weights and a fully invested portfolio constraint. If optimization fails, the method falls back to an allocation that respects the stated limits. The document reports backtest performance from early 2021 through October 2025, but those results are claims from the source and do not establish future performance. It also flags that results depend heavily on the ETF universe: unsuitable assets may undermine the strategy. The description notes an ambiguity between the stated historical lookback and a different initialization value, so implementation details should be checked.
Key ideas
- The strategy ranks a fixed cross-asset ETF universe using trend strength and a short-term moving-average ratio.
- It rebalances every five trading days and selects the six highest-ranked ETFs.
- A 20-day gain threshold can remove an ETF before the next rebalance.
- Mean-variance optimization assigns weights within specified asset-level bounds, with a fallback allocation if optimization fails.
- Performance depends on the chosen ETF universe, and the lookback-period description contains an implementation ambiguity.
Tags
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.