The thesis
Mean-reversion is one of the oldest ideas in systematic trading, and this strategy runs a textbook version of it: buy when the 14-day RSI drops below 30 (oversold), sell when it pushes above 70 (overbought). It trades a 24-name large-cap universe spanning tech, financials, healthcare, staples and energy — AAPL, MSFT, NVDA, JPM, JNJ, XOM and peers — betting that short-term dislocations in quality names snap back to a mean.
Backtest performance
On paper the record is appealing. Over 451 days the strategy returned 14.73% (final equity $11,473 on a $10k base), a 7.98% CAGR, with a 70.59% win rate across 38 trades. That hit rate is the headline strength — roughly seven in ten closed trades were profitable, which is consistent with mean-reversion's characteristic profile of frequent small wins.
The qualifiers matter, though. The Sharpe ratio is a modest 0.58, and the max drawdown reached 15.64% — larger than the annual return, meaning an investor would have needed to stomach a drop bigger than a year's gain. Turnover ran to 879%, so this is an active strategy leaning on many round-trips rather than a few convictions.
Validation: the red flag
This is where the story turns cautious. The walk-forward validation did not pass. Across four folds, three were positive, but the most recent — spanning 2025-12-16 to 2026-05-29 — returned -2.84% with a -0.33 Sharpe and a 14.96% drawdown. That out-of-sample stretch is exactly the period that should give the most confidence, and it was the weakest.
The deflated statistics reinforce the concern. The Probabilistic Sharpe Ratio is a healthy 0.785, but the Deflated Sharpe Ratio is just 0.304 after adjusting for the 6 trials run — below the threshold one would want before trusting the edge as real rather than a product of selection. In plain terms: the backtest looks good, but we can't rule out that it looks good by luck.
Recent live activity
Live behaviour echoes the caution. The only executed trade on record is a WMT buy (21 shares at $115.75) on 2026-05-31. Since then, every scheduled run from 2026-09-03 through 2026-09-11 reports 0 executed, 0 rejected — the RSI triggers simply haven't fired. The book is sitting on $7,569 cash against a total of ~$9,816, meaning capital is largely idle and the live account is marginally below its starting line.
Verdict
Mean-reversion has a genuine strength — a high win rate and a respectable backtest return — but the evidence for a durable edge is thin. A failed validation, a negative out-of-sample fold, a low deflated Sharpe, and a long stretch of no live signals all argue for treating it as promising but unproven. The right posture is patience: let it accumulate out-of-sample live trades before sizing up, and watch whether the next fold recovers or confirms the decay.