The thesis
Mean-reversion is one of the oldest ideas in systematic trading, and this agent runs the textbook version: buy when a name is oversold (RSI below 30), sell when it is overbought (RSI above 70). It operates over a 24-stock universe of US large caps spanning tech, financials, healthcare, staples, and energy — AAPL, MSFT, NVDA, JPM, JNJ, XOM, and the like. The premise is that liquid blue chips overshoot in the short term and snap back toward fair value. It is currently flagged live.
Backtest performance
On paper, the numbers look inviting. Over 451 days the strategy returned 14.73% (final equity $11,473 on a $10,000 base), a 7.98% CAGR, with a 70.59% win rate across 38 trades. That hit rate is the headline strength — mean-reversion tends to win often and small.
The caveats sit just underneath. The Sharpe ratio is a modest 0.58, and max drawdown reached 15.64% — nearly the full annual return at risk in a single decline. Turnover of 879% signals heavy churn; the $38 in fees is trivial here, but on a real book with slippage that activity compounds.
Validation: the real story
This is where enthusiasm should cool. The walk-forward validation failed. Across four folds, three were positive — but the most recent, out-of-sample window (Dec 2025 to May 2026) returned -2.84% with a -0.33 Sharpe and a 14.96% drawdown. Worse, most of the full-sample profit came from a single lucky stretch: fold 2 (Jan–Jul 2025) delivered +11.1% at a 1.32 Sharpe, while the other three folds contributed little or nothing.
The deflated statistics confirm the concern. The Probabilistic Sharpe Ratio is a respectable 0.785, but the Deflated Sharpe Ratio is just 0.304 after adjusting for 6 trials — well short of confidence that the edge is real rather than a product of selection. When one fold carries the strategy and the newest fold is negative, the honest read is not validated.
Recent activity
The live account tells its own quiet story. The last executed trade was a 21-share buy of WMT at $115.75 on May 31 — nearly two months ago. Every scheduled run since (July 20 through 27) reports 0 executed, 0 rejected, with cash pinned at $7,569.25 and total equity drifting between roughly $9,837 and $9,930. In other words, the RSI triggers simply aren't firing in the current low-volatility tape, and the book is slightly underwater versus its starting stake.
Verdict
Mean-reversion has a genuine strength — a high, consistent win rate — but it fails on the metric that matters most: it did not hold up out of sample, and its edge appears concentrated in one favorable period. Combined with a stretch of no signals, this is a strategy to keep on the bench and monitor, not to scale. The next positive out-of-sample fold, not the win rate, is what would change the picture.