The thesis
Mean-reversion is one of the oldest ideas in systematic trading: prices overshoot, then snap back. This strategy operationalises that with the Relative Strength Index — buy names that look oversold (RSI below 30), sell those that look overbought (RSI above 70). It runs over a 24-name universe of large-cap U.S. equities spanning tech, financials, healthcare, staples, and energy — AAPL, MSFT, NVDA, JPM, JNJ, WMT, XOM and the like. The logic is clean and the universe is liquid, which is exactly where mean-reversion tends to behave.
The backtest looks good — on the surface
Over 451 days the strategy returned 14.73%, ending at $11,473 on a $10k base, for a 7.98% CAGR. The win rate is an eye-catching 70.6% across 38 trades, and the max drawdown of 15.64% is contained. Fees were negligible ($38 total, no FX cost).
But the Sharpe ratio of 0.58 is the tell. High win rate plus modest Sharpe is the classic mean-reversion signature: many small wins, occasionally offset by a few larger losses. Turnover of 879% is also worth flagging — this is an active strategy, and in a live account with wider spreads or slippage, that churn erodes the edge that fees alone don't capture.
Validation: it did not pass
This is the part that matters most, and it's where the story turns cautious. Walk-forward validation ran 4 folds; 3 were positive, but the most recent fold (Dec 2025 – May 2026) returned -2.84% with a Sharpe of -0.33. That last fold is also the out-of-sample window, so the headline out-of-sample numbers are simply negative.
The deflated statistics reinforce the concern. The Probabilistic Sharpe Ratio of 0.785 is respectable, but the Deflated Sharpe Ratio — which penalises for the 6 trials run during search — falls to 0.304. That gap between PSR and DSR is a warning that some of the in-sample performance may be selection luck. The validation gate flagged this failed, and rightly so.
Recent activity: parked in cash
Live behaviour matches the caution. The last executed trade was a buy of 21 WMT at $115.75 on 31 May 2026 — over three months ago. Every scheduled run since (28 Aug through 4 Sep) reports 0 executed, 0 rejected, sitting on $7,569 in cash with total account value drifting between roughly $9,736 and $9,852. In other words, the paper account is modestly below its $10k start while the backtest shows a double-digit gain — a sober reminder that historical curves and live results diverge.
The idleness isn't necessarily a fault: if no name is genuinely oversold or overbought, the correct move is to do nothing. RSI-based entries are naturally sparse in trending or range-bound calm.
Verdict
Mean-reversion has a coherent thesis, a clean rule set, and a flattering backtest — but the failed walk-forward, negative out-of-sample fold, and low deflated Sharpe mean it hasn't earned unconditional trust. Treat the 14.73% figure as a hypothesis, not a promise, and watch whether the next active period breaks the recent losing streak or confirms it.