Thesis
Mean-reversion runs one of the oldest ideas in systematic trading: buy weakness, sell strength. Concretely, it buys names whose 14-period RSI drops below 30 (oversold) and sells when RSI pushes above 70 (overbought), across a 24-stock universe of US large caps spanning tech (AAPL, MSFT, NVDA), financials (JPM, V, MA), staples (PG, KO, COST), healthcare (JNJ, UNH), and energy (XOM, CVX). It is currently classified as live.
Backtest performance
Over 451 days the strategy returned 14.73% (final equity $11,473 on a $10k base), a 7.98% CAGR, with a Sharpe of 0.58 and a max drawdown of 15.64%. The headline number is the win rate: 70.59% across 38 trades. That is a genuinely attractive hit rate — but it comes with a caveat familiar to mean-reversion systems. Turnover ran to 879%, so the edge is spread thin across many round-trips, and a Sharpe below 0.6 tells you the ride was choppier than the win rate alone suggests.
Validation: the strategy does not pass
This is where the honest picture emerges. Walk-forward validation failed. Three of four out-of-sample folds were positive, but the story is uneven:
- Fold 1 (Aug 2024–Jan 2025): +2.06%
- Fold 2 (Jan–Jul 2025): +11.10%, Sharpe 1.32
- Fold 3 (Jul–Dec 2025): +2.21%
- Fold 4 (Dec 2025–May 2026): −2.84%, Sharpe −0.33
The full 14.73% return leans heavily on Fold 2; strip it out and the remaining folds are close to flat-to-negative. Most concerning, the most recent fold is the losing one, and the aggregate out-of-sample return is −2.84% with a negative OOS Sharpe. The deflated Sharpe ratio of 0.304 (against a probabilistic Sharpe of 0.785, from 6 trials) says that once you account for how many configurations were tried, the confidence in a real edge is weak. That combination — failed folds, deteriorating recency, low DSR — is exactly why the gate is closed.
Recent activity: quietly dormant
Live behaviour matches a strategy waiting for setups that aren't appearing. The last six scheduled runs (Sep 1–8) each report 0 executed, 0 rejected — no RSI extremes triggered. The last actual fill was a buy of 21 WMT at $115.75 on May 31. The paper account sits at roughly $9,798 total, with $7,569 in cash — about 77% uninvested and modestly below its $10k baseline. So live performance is currently negative and largely on the sidelines.
Verdict
The attraction is real: a clear, explainable thesis and a 70% win rate. But the risks are equally clear. The backtest edge is concentrated in one fold, out-of-sample results are negative, and the deflated Sharpe undercuts the headline. It has also gone quiet for over three months. Mean-reversion earns a place in the research pipeline, not a capital allocation — keep it live for observation, but validation is right to withhold its blessing until the recent-fold weakness reverses.