The thesis
Mean-reversion runs one of the oldest ideas in technical trading: buy weakness, sell strength. Concretely, it opens positions when the 14-day RSI falls below 30 (oversold) and exits when RSI climbs above 70 (overbought). It works a 24-name universe of US large caps spanning tech (AAPL, MSFT, NVDA), financials (JPM, V, MA), staples (PG, KO, WMT, COST) and energy (XOM, CVX). The logic is intuitive and cheap to run — the question, as always, is whether the edge survives contact with out-of-sample data.
Backtest: attractive on the surface
Over 451 trading days the strategy returned 14.73% (7.98% CAGR), finishing at $11,473 on a $10k base. The headline number that catches the eye is the 70.59% win rate across 38 trades. But a high hit rate is characteristic of mean-reversion — it collects many small winners and occasionally wears a large loser, which is exactly what the 15.64% max drawdown and a modest 0.58 Sharpe suggest. Turnover of 879% is high but not alarming for a signal that cycles positions frequently, and fees were a negligible $38.
Validation: this is where it breaks
The walk-forward validation did not pass, and the reasons matter more than the backtest gloss:
- Three of four folds were positive, but the most recent fold (Dec 2025 – May 2026) lost 2.84% with a negative Sharpe of -0.33 and its deepest drawdown (14.96%).
- The out-of-sample Sharpe is negative, meaning the edge decayed precisely on the freshest data.
- The Deflated Sharpe Ratio of 0.304 — adjusting for 6 trials — is weak, even though the Probabilistic Sharpe (0.785) looks respectable in isolation.
Read together, this is a strategy whose historical performance was front-loaded (fold 2 alone returned 11.1% at a 1.32 Sharpe) and whose recent behaviour has degraded. That is the classic fingerprint of a signal that worked in one regime and is now struggling.
Recent activity: mostly silence
The live sleeve tells its own story. The last executed trade was a WMT buy on 31 May 2026 — over three months ago. Every scheduled run since (21–28 August) reports 0 executed, 0 rejected, holding $7,569 in cash against a total account value hovering around $9,730–$9,790. That is below the $10k starting line, so the live book is modestly underwater while the backtest showed a gain. The flat cash balance and lack of fills indicate few names have hit the RSI<30 trigger recently — unsurprising in a grinding, non-oversold market.
Verdict
Strengths: a transparent, well-understood thesis; a genuinely high win rate; low trading costs. Risks: validation failed, out-of-sample returns turned negative, the deflated Sharpe is thin, and the live account is both underwater and inactive. Mean-reversion earns its keep in choppy, range-bound tape and suffers when trends persist. Until the recent-fold weakness reverses on live data, this one belongs on the watch list rather than in a real allocation.