The thesis
Mean-reversion runs one of the oldest ideas in technical trading: buy weakness, sell strength. Concretely, it buys names whose 14-day RSI drops below 30 (oversold) and sells when RSI pushes above 70 (overbought). The universe is 24 US large caps spanning tech, financials, healthcare, staples, and energy — AAPL, MSFT, NVDA, JPM, V, JNJ, WMT, XOM and peers. It is a diversified, liquid playground, which is the right place to test a mechanical signal.
Backtest: a genuinely appealing headline
Over 451 days the strategy returned 14.73% (7.98% CAGR), finishing with $11,473 on a $10,000 book. The standout is a 70.59% win rate across 38 trades — mean-reversion tends to be right often, which is characteristic of the style. Max drawdown was contained at 15.64%.
Two cautions sit underneath the headline. The Sharpe ratio is a modest 0.58, meaning the returns came with meaningful volatility relative to their size. And turnover was 879% — the book churned roughly nine times over. Fees were only $38 here, but at any real scale that churn is a tax on the edge.
Validation: this is where it gets honest
The walk-forward test did not pass. Across four sequential folds, three were positive — folds 1 through 3 returned 2.06%, 11.10%, and 2.21%. The problem is the most recent fold (Dec 2025 → May 2026), which lost -2.84% with a negative Sharpe of -0.33 and a 14.96% drawdown. The out-of-sample window is precisely the period we care most about, and it is red.
The deflated metrics tell the same story from a different angle. The Probabilistic Sharpe Ratio is a healthy 0.785, but the Deflated Sharpe Ratio is just 0.304 after accounting for six trials — once you penalize for how many variants were searched, the confidence that this edge is real drops sharply. That gap between PSR and DSR is the classic fingerprint of a strategy that looks better in-sample than it will out-of-sample.
Recent activity: live, but idle
The strategy is live yet quiet. Its last executed trade was a 21-share WMT buy at $115.75 on May 31. Since then, six consecutive scheduled runs (Aug 10–17) reported 0 executed, 0 rejected — no name has hit the RSI thresholds. Cash is parked at $7,569 and total account value has drifted between roughly $9,921 and $9,998, i.e. marginally underwater versus its $10,000 start.
Verdict
Mean-reversion has a real, recognizable edge — high hit rate, disciplined rules, sensible universe. But a failed walk-forward, a negative most-recent fold, and a deflated Sharpe near 0.3 are exactly the warning signs our validation gate exists to catch. High turnover and a long idle stretch add practical friction. This is a strategy to keep on the leaderboard and watch, not one to size up. The next few RSI signals — whenever the market finally hands it one — will tell us whether the recent weakness was noise or decay.