The thesis
Mean-reversion runs one of the oldest ideas in systematic trading: prices overshoot, then snap back. Concretely, it buys names whose RSI falls below 30 (oversold) and sells when RSI pushes above 70 (overbought). The universe is a conservative basket of 24 US large-caps — mega-cap tech (AAPL, MSFT, GOOGL, NVDA), financials (JPM, BAC, V, MA), healthcare, staples and industrials. There is nothing exotic here, which is both the strategy's appeal and, as we'll see, part of its problem.
Recent activity
The strategy is live and trading on a scheduled nightly cadence. Most sessions in early November did nothing at all — the 2nd through 5th each logged zero executions — which is expected behaviour: when nothing is oversold, mean-reversion sits on its hands. Activity clustered at the end of October and on the 6th: a 48-share BAC buy at $51.15 (Oct 29), then CAT (4 @ $622.54) and GOOGL (8 @ $280.28) on Oct 30, and most recently 13 shares of NVDA at $188.24 on Nov 6.
The honest read on live results is cautious. Total equity has hovered in the $9,600–$9,900 range all week — below the $10,000 it started from. The strategy is patient, but patience hasn't yet translated into gains since going live.
Backtest and validation
On paper the backtest looks good: 14.73% total return over 451 days (a 7.98% CAGR), a 70.59% win rate across 38 trades, and a manageable fee drag of just $38. Drawdown topped out at 15.64% and the Sharpe ratio was a middling 0.58.
The cracks appear under walk-forward validation, which the strategy failed. Three of four folds were positive — fold 2 was excellent (+11.1%, Sharpe 1.32) — but the most recent fold (Dec 2025–May 2026) lost 2.84% with a negative Sharpe of −0.33 and a 14.96% drawdown. That matters: the out-of-sample window, the one that best proxies "what happens next," is exactly the one that went red.
The statistical measures reinforce the warning. The Deflated Sharpe Ratio sits at just 0.304 against 6 trials — meaning once you account for how many variations were tested, the odds that the true edge is real are not convincing. The Probabilistic Sharpe of 0.785 is more flattering but doesn't rescue the verdict.
Balance sheet: strengths and risks
Strengths: a high and consistent win rate, modest fees, disciplined inactivity when no signal fires, and a liquid, defensible universe. The core edge — buying weakness in quality names — is economically sensible.
Risks: the strategy failed validation, the latest out-of-sample fold lost money, and the deflated Sharpe suggests the headline numbers may be partly luck. Turnover of 879% is high for a strategy that trades infrequently, implying concentrated, chunky positions. And live performance so far is underwater.
Verdict
Mean-reversion is a textbook strategy with a textbook problem: it backtests well and generalises poorly. The 70% win rate is real, but a losing out-of-sample fold and a weak DSR mean we should treat this as a candidate under observation, not a proven edge. Watch whether the recent BAC/GOOGL/NVDA entries revert as the thesis predicts — that's the test that counts.