The thesis
Mean-reversion is one of the oldest ideas in systematic trading, and this strategy runs the textbook version: buy the oversold (RSI below 30), sell the overbought (RSI above 70). It operates across a 24-name universe of US large caps — the mega-cap tech cohort, financials like JPM, BAC, V and MA, plus defensives such as JNJ, PG, KO and WMT. The bet is simple: short-term price dislocations in liquid, well-covered stocks tend to snap back.
Backtest performance
On paper, the numbers are inviting. Over 451 days the strategy returned 14.73%, ending at $11,473 from a $10k base, for a 7.98% CAGR. The headline win rate is 70.59% across 38 trades, with fees a negligible $38 and no FX drag. That hit rate is the strongest part of the story — most entries resolved in the strategy's favour.
The caveats sit right next to it. The Sharpe is only 0.58 and the max drawdown is 15.64%, so the return-per-unit-of-risk is thin and the ride is bumpy. Turnover is a heavy 879%, meaning the book churns itself roughly nine times over — a profile that is more fee- and slippage-sensitive in the real world than the clean backtest suggests.
Validation: the strategy did not pass
This is the part that matters most, and it is unflattering. Walk-forward validation across four folds failed. Three of four folds were positive, but the decisive one — fold 4, the most recent window (Dec 2025 to May 2026) — returned -2.84% with a Sharpe of -0.33 and a 14.96% drawdown. Out-of-sample performance is exactly where a 70% win rate needs to prove it isn't curve-fit, and here it buckled.
The risk-adjusted diagnostics agree. With 6 trials in the search, the deflated Sharpe ratio (DSR) is just 0.304 — low enough to question whether the in-sample Sharpe survives multiple-testing adjustment — even though the probabilistic Sharpe (PSR 0.785) looks healthier. When PSR and DSR disagree like this, the DSR is the more honest number because it penalises the search effort that produced the result.
Recent live activity
The live book tells its own story. Equity has drifted from $9,619 on 28 Sep to roughly $9,387 on 5 Oct — below the starting line. The last executed trade was a MCD buy on 24 Sep, preceded by buys in HD, BAC and WMT; every recent trade is an entry. Since then, scheduled runs have executed nothing (with one rejection on 5 Oct), and cash has sat flat at $329.71. In other words, the strategy is fully committed and waiting: RSI has flagged oversold entries but not yet triggered the overbought exits that would realise gains.
Verdict
Mean-reversion has a genuinely high win rate and a clean, intuitive thesis. But the combination of a failed walk-forward test, a negative most-recent fold, a low deflated Sharpe, and a live book currently underwater argues for caution. This is a candidate to keep on a short leash — watch whether the pending positions mean-revert as designed, rather than trusting the backtest headline.