The thesis
Mean-reversion runs one of the oldest ideas in technical trading: buy weakness, sell strength. Concretely, it accumulates names when their RSI drops below 30 (oversold) and trims when RSI pushes above 70 (overbought). The universe is 24 large-cap US blue chips across tech, financials, healthcare, staples and energy — AAPL, MSFT, NVDA, JPM, V, JNJ, PG, XOM and peers. It is a low-exotica, liquid-names approach, and it currently runs live as a paper strategy.
What the backtest says
On paper the numbers flatter the strategy. Over 451 days it returned 14.73% (7.98% CAGR), finishing at $11,473 from a $10,000 base, with a 70.59% win rate across 38 trades. Drawdown was contained at 15.64% and total fees were trivial at $38. The headline weakness is a Sharpe of just 0.58 — respectable return, but earned with enough volatility that risk-adjusted it is unremarkable. Turnover of 879% is also high; that it still nets out cheaply says more about the low fee model than about restraint.
Why validation failed
This is where enthusiasm should cool. The walk-forward validation did not pass. Three of four folds were positive — fold 2 was genuinely strong (+11.1%, Sharpe 1.32) — but the newest fold (2025-12-16 to 2026-05-29) lost 2.84% at a Sharpe of -0.33 and a 14.96% drawdown. The out-of-sample read is therefore negative, and the deflated Sharpe ratio of 0.304 (against a PSR of 0.785 over 6 trials) warns that the full-sample Sharpe is inflated by selection. In plain terms: the edge looks real in aggregate but is not stable, and the most recent regime is where it broke.
Live activity
Live behaviour matches a strategy with few signals firing. The last week of scheduled runs executed nothing on most days — a single trade on 2026-10-16, otherwise zeros — which is expected when few names are at RSI extremes. The notable recent action: selling 21 shares of WMT at $118.83, having bought the same lot at $115.75 back in May, a small clean win. September saw oversold buys into MCD, HD and BAC. The live book sits around $9,925 total with $2,820 cash, i.e. modestly below its starting stake — a reminder that the backtest gains have not yet reproduced forward.
Verdict
Strengths: a high hit rate, cheap execution, and a transparent, well-understood thesis on liquid names. Risks: a failed validation driven by a losing recent fold, a mediocre Sharpe, and a live portfolio that is slightly underwater. The honest reading is a strategy with a plausible but fragile edge. It deserves to keep running on paper for more out-of-sample evidence — not capital allocation — until it demonstrates it can navigate the regime that just beat it.