The Thesis
Mean-reversion runs a classic, well-understood premise: buy the oversold, sell the overbought. Mechanically, it enters when the 14-day RSI drops below 30 and exits when RSI climbs above 70. The universe is 24 US large-caps spanning tech (AAPL, MSFT, NVDA), financials (JPM, BAC, V, MA), healthcare (JNJ, UNH, PFE), and staples (PG, KO, WMT, COST). It is a defensible, liquid, low-exotica setup — the kind of strategy whose edge, if any, comes from discipline rather than novelty.
Backtest Performance
The headline numbers flatter the strategy. Over 451 days it returned 14.73% (7.98% CAGR), finishing at $11,473 on a $10k base, with a 70.59% win rate across 38 trades. That hit rate is the standout figure, and it fits the mean-reversion profile: many small, high-probability wins.
But context tempers the enthusiasm. The Sharpe ratio of 0.58 is modest — the returns came with real volatility, including a 15.64% max drawdown. Turnover of 879% is heavy, meaning the equity curve leans on constant repositioning; the $38 in fees is trivial only because this is a small paper book. A high win rate paired with a middling Sharpe hints that losing trades, though rarer, were sizeable.
The Validation Problem
This is where the case weakens. The strategy fails walk-forward validation. Across four folds, three were positive — but the most recent fold (Dec 2025–May 2026) returned -2.84% with a Sharpe of -0.33 and a near-15% drawdown. That negative fold is the out-of-sample result, which is the honest test of whether the edge persists forward rather than curve-fitting the past.
The deflated statistics say as much. The Probabilistic Sharpe Ratio of 0.785 is respectable, but the Deflated Sharpe Ratio — which penalises the 6 trials run — drops to 0.304, below the threshold one would want before trusting the result. In plain terms: after accounting for how many variants were tested, the observed performance is plausibly luck.
Live Reality Check
The live paper account reflects this caution. Recent scheduled runs (16–23 Sep) show a portfolio hovering between $9,772 and $9,841 — still below the $10k start. Activity is sparse: most days execute zero trades, with single buys in HD (8 shares @ $299.12) and BAC (42 @ $58.21) the only recent moves. With RSI thresholds this strict, the strategy sits in cash for long stretches ($2,720 idle), which is by design but drags on returns in a rising market.
Verdict
Mean-reversion has an intuitive thesis, a high win rate, and low trading costs — genuine strengths. But the failed validation, negative out-of-sample fold, and sub-water live account are red flags that the backtest overstates the forward edge. The strategy is worth keeping live for continued observation, not scaling. The next fold, and whether live performance climbs back above $10k, will tell us far more than the headline return does.