The Thesis
Mean-reversion is one of the oldest ideas in quantitative trading: prices that stretch too far from their average tend to snap back. This strategy operationalizes that with the Relative Strength Index, buying names that look oversold (RSI < 30) and selling those that look overbought (RSI > 70). It runs across a 24-name, large-cap universe spanning tech, financials, healthcare, staples, and energy — AAPL, MSFT, NVDA, JPM, JNJ, WMT, XOM and peers. It is a clean, interpretable rule set, which is exactly what makes it a useful benchmark.
Backtest Performance
On paper, the numbers flatter the strategy. Over 451 days it returned 14.73% (7.98% CAGR), ending at $11,473 on a $10,000 base, with a 70.59% win rate across 38 trades. That headline hit rate is the strategy's most seductive feature. Look closer, though, and the picture cools: a Sharpe of 0.58 is modest, the max drawdown of 15.64% actually exceeds the annualized return, and turnover of 879% means the book churns roughly nine times over — a real drag once slippage and the $38 in fees compound in live conditions.
Validation: Where It Breaks
Here is the uncomfortable part. The strategy failed formal walk-forward validation. Across four folds, three were positive — but the most recent one (Dec 2025 to May 2026) returned -2.84% with a -0.33 Sharpe and a 14.96% drawdown. Aggregate out-of-sample Sharpe was negative. The Deflated Sharpe Ratio of 0.304 — well below any pass threshold, and adjusted for 6 trials — tells us the in-sample edge does not survive honest testing. The PSR of 0.785 is more encouraging, but the DSR is the stricter, more trustworthy read. In short: the pretty backtest is at least partly a fit to history, and the edge appears to be decaying in the newest data.
Recent Activity
Live paper trading reflects that stall. The last executed trade was a buy of 21 shares of WMT at $115.75 on May 31. Every scheduled run since — July 10 through 17 — logged 0 executed, 0 rejected: no names hit the RSI extremes, so the strategy simply sat on $7,569 in cash. Total portfolio value has drifted between roughly $9,957 and $9,986, leaving it marginally underwater versus its starting stake.
Verdict
Mean-reversion remains a valuable, legible baseline with an enviable win rate. But a failing validation grade, a negative most-recent fold, and weeks of idle cash are warnings, not footnotes. Until out-of-sample results stabilize, treat this one as a study in why backtests deserve skepticism — not as a green light.