The thesis
Mean-reversion is one of the oldest bets in the book: prices that stretch too far from their average tend to snap back. This strategy operationalizes that with RSI, buying when a name is oversold (RSI < 30) and selling when it's overbought (RSI > 70). It runs on a 24-stock large-cap universe spanning tech (AAPL, MSFT, NVDA), financials (JPM, V, MA), staples (PG, KO, WMT), and energy (XOM, CVX) — a broadly diversified, liquid set that suits a signal built on statistical extremes rather than sector-specific narratives.
Headline performance
On the full backtest, the numbers look inviting. Over 451 days the strategy returned 14.73% (final equity $11,473 on a $10k base), a 7.98% CAGR, with a 70.59% win rate across 38 trades. Drawdown was contained at 15.64%, and total fees were a trivial $38.
The caveat sits right next to the win rate: a Sharpe of 0.58 is modest. A high hit rate paired with a middling Sharpe tells a familiar mean-reversion story — many small wins, punctuated by occasional larger losses when a 'snap back' turns into a trend that keeps going. Turnover of 879% also means the equity is being recycled aggressively, so execution quality matters more than the fee line suggests.
Validation: the part that matters
This is where enthusiasm should cool. Walk-forward validation failed. Across four folds, three were positive, but the most recent out-of-sample window (Dec 2025 – May 2026) returned -2.84% with a negative Sharpe of -0.33 and its worst drawdown yet (14.96%). The earlier folds were strong — fold 2 alone returned 11.1% at a 1.32 Sharpe — but the deterioration into the freshest, most relevant period is exactly the pattern robust validation is designed to catch.
The deflated statistics reinforce the concern. Against 6 trials, the Probabilistic Sharpe Ratio is 0.785 but the Deflated Sharpe Ratio drops to 0.304 — once you account for selection across trials, the edge is no longer convincingly distinguishable from luck.
Recent activity: a strategy in neutral
The live book reflects this hesitation. The last executed trade was a 21-share WMT buy at $115.75 on May 31. Every scheduled run since — six consecutive sessions from Aug 6 through Aug 13 — logged 0 executed, 0 rejected, holding $7,569 in cash against a total account value hovering near $9,930–$9,999. In plain terms: no names are hitting the RSI extremes, so the strategy is simply waiting. That's disciplined behavior, but it also means it's currently below its $10k starting mark and not compounding.
Verdict
Mean-reversion has a coherent thesis, low costs, and a genuinely high win rate. But the failed walk-forward, the negative most-recent fold, and a Deflated Sharpe near 0.3 argue against treating the 14.73% backtest as forward-looking truth. Keep it live in observation mode, watch whether the RSI signal re-engages in the current regime, and demand a passing out-of-sample fold before scaling capital toward it.