Strategy Overview
The mean-reversion agent operates on a straightforward premise: buy when RSI drops below 30 (oversold), sell when it climbs above 70 (overbought). Its universe spans 24 large-cap US names across technology, financials, healthcare, consumer staples, energy, and industrials — a deliberately diversified basket designed to surface RSI extremes across market regimes without sector concentration.
The thesis is one of the oldest in systematic trading. When a high-quality stock is beaten down relative to its recent range, the odds favour a snap-back. The same logic applies in reverse. Done right, it harvests short-term mean-reversion without predicting direction — just identifying temporary extremes.
Backtest Performance
Over 451 trading days, the backtest produced a 14.73% total return (7.98% CAGR) with a Sharpe ratio of 0.58. The win rate of 70.6% across 38 trades is the headline number — nearly three out of four signals resolved in the expected direction. Maximum drawdown reached 15.64%, which is meaningful but not alarming for an equity-only strategy.
Turnover at 879% annualised is high on paper, but with only 38 trades over 15 months the absolute transaction cost came to $38 — a rounding error at this capital scale. FX costs were zero, reflecting the all-USD universe.
Walk-Forward Validation: Where It Gets Complicated
The strategy did not pass Headmars' automated validation gate, and the fold breakdown explains why.
| Fold | Period | Return | Sharpe | Max DD | Trades |
|---|---|---|---|---|---|
| 1 | Aug 2024 – Jan 2025 | +2.06% | 0.57 | 4.24% | 8 |
| 2 | Jan 2025 – Jul 2025 | +11.10% | 1.32 | 10.53% | 10 |
| 3 | Jul 2025 – Dec 2025 | +2.21% | 0.46 | 8.80% | 6 |
| 4 | Dec 2025 – May 2026 | −2.84% | −0.33 | 14.96% | 10 |
Fold 2 is doing a lot of the heavy lifting — its 11.1% return and 1.32 Sharpe account for most of the aggregate gain. Folds 1 and 3 are modest positives. Fold 4, the out-of-sample period, is a loss. The probabilistic Sharpe ratio (PSR) of 0.785 is respectable, but the deflated Sharpe ratio (DSR) of 0.304 — which penalises for the number of trials — signals that with 6 parameter trials, the results may not clear the bar for statistical robustness.
Recent Activity
The agent has been quiet. Daily runs from June 10–17 executed zero trades, with cash holdings steady at $7,569 and total portfolio value oscillating between roughly $10,040 and $10,108. This is the strategy behaving as designed: when no stock in the universe breaches an RSI extreme, the agent sits on its hands. It is not stuck — it is selective.
The most recent executed trade was a buy of 21 shares of WMT at $115.75 on May 31, which would reflect Walmart touching oversold territory at that time.
Strengths and Risks
Strengths: High win rate, disciplined entry/exit rules, low turnover in practice, and a universe confined to liquid large-caps that reduces execution risk. The agent's current inactivity demonstrates it isn't forcing trades in neutral conditions.
Risks: The failing validation gate should not be ignored. Fold 4's negative OOS performance — combined with a low DSR — suggests the strong Fold 2 performance may reflect a market environment (high volatility, sharp rotations) that the strategy exploits well but cannot count on consistently. RSI-only signals also carry well-known risks in trending markets: oversold can get more oversold. The 15.64% maximum drawdown is a live risk, not just a backtest artefact.
Bottom Line
Mean-reversion is a credible, well-understood strategy with a clean signal. The backtest numbers are encouraging and the current patience in sideways markets is a feature, not a bug. But the out-of-sample deterioration in the most recent fold is the kind of evidence the validation system exists to catch. Before this strategy graduates to real capital, a closer look at regime-conditioning — or at minimum a longer out-of-sample window — would strengthen the case considerably.