Thesis and Approach
The mean-reversion strategy is one of the oldest ideas in systematic trading: prices tend to snap back toward their long-run average after extreme moves. The Headmars implementation operationalises this with a simple RSI filter — buy when RSI falls below 30 (oversold), sell when it climbs above 70 (overbought) — applied across a 24-ticker universe of large-cap US equities spanning tech, financials, healthcare, consumer staples, energy, and industrials.
The universe is deliberately defensive: household names with deep liquidity and well-documented mean-reverting tendencies during risk-off episodes. The logic is sound in theory. The data, however, tells a more nuanced story.
Backtest Performance
Over 451 days, the strategy returned 14.73% on paper equity (CAGR ~7.98%), with a Sharpe of 0.58 and a maximum drawdown of 15.64%. The win rate of 70.6% across 38 trades is genuinely encouraging — the strategy is right more often than not when it does act. Turnover is modest at 879% annualised (roughly 8–9 full rotations), and total fees were a flat $38, keeping friction low.
These are respectable numbers for a rule-based RSI strategy. But backtest results are only the starting line.
Walk-Forward Validation: Where It Gets Complicated
The four-fold walk-forward validation did not pass. Three of four folds were profitable, which sounds encouraging, but the critical number is the out-of-sample (OOS) return: −2.84%, with an OOS Sharpe of −0.33. The most recent fold (December 2025 – May 2026) drove that result, posting a −2.84% return and a peak drawdown of nearly 15%.
Two additional statistics warrant attention:
- PSR (Probabilistic Sharpe Ratio): 0.785 — there is roughly a 22% chance the true Sharpe is zero or negative. For a live deployment decision, most practitioners want PSR above 0.95.
- DSR (Deflated Sharpe Ratio): 0.304 — after adjusting for the number of trials (6) and the optimism bias inherent in backtesting, the probability that the observed Sharpe reflects genuine edge falls to about 30%. This is the headline overfitting signal.
In plain terms: the backtest's edge does not hold up cleanly when tested on data the strategy has never seen, and the statistical haircut for multiple testing is steep.
Recent Activity
Live paper-trading behaviour mirrors the validation concern. The most recent executed trade was a buy of 21 shares of WMT at $115.75 on 31 May 2026. Since then — six consecutive scheduled runs from 5–12 June — the strategy has executed zero trades, leaving the portfolio sitting at roughly 75% cash ($7,569 of $10,107 total equity).
This isn't necessarily a flaw: RSI-based strategies are inherently patient and only fire when the market hands them a clear signal. But extended dormancy in a trending or range-bound tape can mean significant opportunity cost, and it underscores how infrequent the entry conditions are in this universe under current market conditions.
Strengths and Risks
Strengths: High in-sample win rate, low fees, interpretable rules, a diversified universe, and three out of four folds in the black.
Risks: Negative OOS performance in the most recent period, a DSR well below the 0.5 threshold that would provide statistical confidence, and a strategy that can go weeks without a signal — making it hard to distinguish "patiently waiting" from "regime-broken."
Outlook
Mean-reversion is a legitimate edge in the right conditions, but this incarnation needs further evidence before earning real capital. Extending the backtest window, tightening the trial count, or combining the RSI signal with a volatility or volume filter may improve the DSR. Until validation passes, this strategy is best kept in observation mode — which is exactly where Headmars has it.