The thesis
The mean-reversion strategy runs one of the oldest ideas in technical trading: buy what looks oversold and sell what looks overbought. Concretely, it enters when RSI drops below 30 and exits when RSI climbs above 70, applied across a universe of 24 US large caps spanning tech (AAPL, MSFT, NVDA), financials (JPM, V, MA), healthcare (JNJ, UNH), staples (PG, KO, WMT) and energy (XOM, CVX). It is currently flagged live.
Backtest: encouraging on the surface
Over 451 trading days the full-sample backtest returned 14.73%, ending on equity of $11,473 from a $10k base — a CAGR of roughly 7.98%. The headline strengths are a 70.59% win rate across 38 trades and modest costs ($38 in fees, no FX drag). Risk-adjusted, though, the picture cools: a Sharpe of 0.58 and a max drawdown of 15.64% mean you took meaningful volatility for that return. Turnover of 879% is also high for a strategy that trades this infrequently, hinting at concentrated, chunky positions rather than smooth compounding.
Validation: the strategy did not pass
This is where enthusiasm should be tempered. Walk-forward validation across four folds failed. Three of the four folds were positive, but the most recent one — covering 2025-12-16 to 2026-05-29 — returned -2.84% with a negative Sharpe of -0.33 and a 14.96% drawdown, its worst on record. The aggregate out-of-sample return was that same -2.84%, versus the rosy 14.73% in-sample.
The deflated statistics reinforce the caution. The Probabilistic Sharpe Ratio sits at 0.785, but the Deflated Sharpe Ratio is only 0.304 after adjusting for 6 trials — below the threshold one would want before trusting the edge as real rather than lucky. In plain terms: the full-sample glow looks partly like curve-fit, and the strategy stumbled most in the freshest, most relevant period.
Recent activity: quiet and slightly underwater
The live paper account tells a sober story. Its last actual fill was a buy of 21 WMT shares at $115.75 on 2026-05-31. Every scheduled run since — six logged between 2026-08-25 and 2026-09-01 — executed zero trades and rejected zero, with cash parked at $7,569.25. Total account value has drifted between roughly $9,724 and $9,795, i.e. still below the $10k starting line. No RSI extremes have triggered entries or exits for months, so the book is mostly idle cash plus a single legacy position.
Verdict
Mean-reversion has a clean, explainable thesis and a genuinely high hit rate — attractive traits. But the evidence stack points to caution: a failed validation, a negative out-of-sample fold in the most recent window, a weak deflated Sharpe, and a live account that is quiet and marginally down. The right read is promising but unproven. Until it demonstrates positive out-of-sample behaviour — and actually deploys capital when signals fire — this one belongs on a short leash, not a larger allocation.