The Thesis
The mean-reversion strategy is built on a simple, time-tested premise: equity prices tend to snap back toward their historical averages after extremes. The agent scans a 24-stock universe of large-cap U.S. equities — spanning tech, financials, healthcare, consumer, and energy — and buys when RSI falls below 30 (oversold) and sells when RSI exceeds 70 (overbought). No leverage, no complex signals. Just a classic reversion bet on blue chips.
Backtest Performance
Over 451 trading days, the strategy returned 14.73% on a starting equity of $10,000, compounding to a final equity of $11,473. Annualised, that's a CAGR of roughly 7.98% — respectable for a passive-flavoured large-cap approach with no overnight risk from smaller names.
The win rate of 70.6% across 38 trades is the headline strength. More than seven in ten trades closed in profit, suggesting the RSI thresholds are doing real work at identifying reversion candidates in this universe. Turnover at 879% sounds alarming but across 38 trades over 15 months it reflects normal cycling rather than hyperactivity. Total fees came to $38 — essentially a dollar a trade.
The Sharpe of 0.58 is modest but positive. Maximum drawdown of 15.64% is the most significant risk flag in the backtest — deep enough to test conviction during a bad patch.
Fold-by-Fold Breakdown
The cross-validation picture is where nuance enters. Of four time folds:
| Fold | Period | Return | Sharpe | Max DD |
|---|---|---|---|---|
| 1 | Aug 2024 – Jan 2025 | +2.06% | 0.57 | 4.24% |
| 2 | Jan 2025 – Jul 2025 | +11.10% | 1.32 | 10.53% |
| 3 | Jul 2025 – Dec 2025 | +2.21% | 0.46 | 8.80% |
| 4 | Dec 2025 – May 2026 | -2.84% | -0.33 | 14.96% |
Fold 2 was the standout — an 11.1% return with a Sharpe above 1.3 likely reflects a market environment where oversold bounces were sharp and reliable. Folds 1 and 3 were quietly profitable but unspectacular. Fold 4, the most recent period, turned negative: -2.84% with a near-15% drawdown. This is the out-of-sample fold, and it drives the validation failure.
Validation Verdict: Not Yet
The strategy did not pass the Headmars validation gate. The Probabilistic Sharpe Ratio (PSR) of 0.785 indicates a reasonable probability that the Sharpe is genuinely positive, but the Deflated Sharpe Ratio (DSR) of 0.304 — which corrects for the number of trials tested (6 in this run) — is the disqualifier. After adjusting for multiple testing, the edge is not statistically robust enough to clear the bar. Out-of-sample Sharpe of -0.33 confirms the concern: what worked in-sample has not translated cleanly to the most recent market regime.
Live Activity: Waiting for a Signal
The last executed trade was a buy of 21 shares of WMT at $115.75 on May 31. Since then — across six consecutive daily runs from June 3 through June 10 — the agent has executed zero trades. Cash sits at $7,569 against a total portfolio value hovering near $10,060–$10,095.
This inactivity is not a bug; it means no stock in the 24-name universe has crossed the RSI-30 threshold in a week of sessions. In a relatively calm or trending-up market, oversold conditions simply don't appear.
Bottom Line
Mean-reversion is a legitimate, historically grounded strategy with a clean signal and a strong win rate. But the validation data is an honest warning: the edge may be regime-dependent. One strong quarter (Fold 2) is carrying most of the return, and the most recent period is underwater. Watch whether Fold 4 conditions persist or reverse before reading too much into the backtest headline.