The thesis
Mean-reversion runs one of the oldest ideas in technical trading: buy weakness, sell strength. Concretely, it accumulates names whose 14-period RSI drops below 30 (oversold) and trims those that push above 70 (overbought). The universe is 24 large-cap US equities spanning tech, financials, healthcare, staples, and energy — AAPL, MSFT, NVDA, JPM, V, JNJ, UNH, PG, WMT, XOM and peers. It is a defensible, well-understood premise, and on a diversified blue-chip universe it should behave.
Headline numbers
The full-sample backtest looks healthy at first glance: a 14.73% total return, final equity of $11,473 on a $10k base, a 7.98% CAGR over 451 days, and a 70.59% win rate across 38 trades. Fees were negligible ($38 total, no FX cost). Those are the strengths, and the high win rate is genuinely on-brand for mean-reversion, which tends to win often and small.
But two caveats sit right next to the return. The Sharpe ratio is only 0.58, meaning that return came with meaningful volatility, and the max drawdown reached 15.64% — larger than the annual return itself. Turnover of 879% also signals a strategy that churns the book aggressively for that modest edge.
Validation says: not yet
This is where the story turns cautionary. The walk-forward validation fails. Across four sequential folds, three were positive — but the pattern decays: +2.06%, +11.10%, +2.21%, and finally −2.84% in the most recent fold (Dec 2025–May 2026), which also carried the worst drawdown at 14.96%. Out-of-sample Sharpe came in negative (−0.33).
The robustness statistics tell the same tale. The Probabilistic Sharpe Ratio (0.785) is borderline, but the Deflated Sharpe Ratio — which penalizes for the 6 trials run — collapses to 0.304. Once you account for how many variants were tested, the evidence that this edge is real, not luck, is weak.
Live activity: a book on pause
The live account reflects that hesitation. The last executed trade was a buy of 21 WMT shares at $115.75 on May 31. Every scheduled run since — six shown here, from July 23 through July 30 — reports 0 executed, 0 rejected. Cash has been parked at $7,569.25 the entire stretch, with total equity drifting between roughly $9,837 and $9,971, i.e. slightly below the $10k base. In other words, no RSI extremes have triggered, and the strategy is effectively sidelined.
Verdict
Mean-reversion is a coherent idea with an attractive win rate, but the numbers argue for patience over promotion. The headline return leans on early folds, the deflated Sharpe undercuts the edge, and the most recent out-of-sample window lost money. Combined with a live book that hasn't found a setup in two months, this reads as a strategy to keep on a short leash — monitored, capital-capped, and re-validated — rather than scaled.