The thesis
Mean-reversion runs one of the oldest ideas in technical trading: buy weakness, sell strength. Concretely, it buys names whose 14-day RSI drops below 30 (oversold) and sells when RSI climbs above 70 (overbought). It operates live across a 24-name large-cap universe spanning tech, financials, healthcare, staples and energy — AAPL, MSFT, NVDA, JPM, JNJ, XOM and the like. The logic is clean and interpretable, which is a virtue: you can reason about why any given trade fired.
Recent activity
Activity has been thin. The last four executed trades are all buys — MCD (10 @ $238.58) and HD (8 @ $299.12) in late September, BAC (42 @ $58.21) in mid-September, and a lone WMT buy back in May. Every scheduled run in October executed nothing: six consecutive sessions logged 0 executed, with two orders rejected along the way. Cash has sat frozen at $329.71 throughout, which tells us the book is effectively fully invested and waiting for RSI signals that aren't arriving.
The uncomfortable number is the account total. The paper book stands at $9,637.42 — below the $10,000 starting stake the backtest implies (14.73% on a final equity of $11,473.17). In other words, the strategy that looks like a winner on paper is modestly underwater in live trading right now.
Backtest and validation
On the surface, the backtest is appealing: 14.73% total return over 451 days (7.98% CAGR), a 70.59% win rate across 38 trades, and a tolerable 15.64% max drawdown. Fees were negligible ($38) and FX cost nil.
The validation layer is where the story turns. The walk-forward test failed. Three of four folds were positive, but the most recent fold (Dec 2025–May 2026) returned −2.84% with a −0.33 Sharpe and a 14.96% drawdown — the deterioration is recent, not ancient history. The aggregate out-of-sample return is that same −2.84%, versus +14.73% in-sample. The Deflated Sharpe Ratio sits at just 0.304 (against a PSR of 0.785), reflecting a penalty for the six trials run to find this configuration. A DSR that low means the headline Sharpe of 0.58 is not statistically convincing once selection bias is accounted for.
Verdict
The gap between in-sample glamour and out-of-sample reality is the whole lesson here. A 70% win rate is seductive, but win rate says nothing about the size of losses relative to wins, and the strategy's recent live and out-of-sample performance both skew negative. Turnover of 879% also hints at a strategy more active than its sparse recent trades suggest, depending on regime.
Strengths: transparent logic, broad liquid universe, respectable drawdown control, low costs.
Risks: failed validation, a negative and recent out-of-sample fold, a deflated Sharpe that undercuts the headline, and a live book currently below its starting value. For now this is a strategy to watch and learn from — not one to scale. The validation gate flagged it correctly.