Strategy Thesis
The mean-reversion agent operates on a simple, time-tested premise: markets overreact. When a stock's 14-day RSI falls below 30 it is considered oversold and the agent buys; when RSI climbs above 70 the position is closed as overbought. The universe spans 24 large-cap U.S. names across six sectors — technology, financials, healthcare, consumer staples and discretionary, energy, and industrials — giving the strategy broad exposure while staying within liquid, well-covered names where RSI signals have historically carried more signal than noise.
Backtest Highlights
Over a 451-day backtest the strategy generated a 14.73% total return (7.98% CAGR) on 38 trades, with a 70.6% win rate and a 0.58 Sharpe ratio. Those headline numbers are solid for a rule-based system operating without any fundamental overlay. Maximum drawdown came in at 15.64%, manageable but not trivial — indicating the strategy is not immune to multi-week streaks of losses when mean-reversion bets go offside.
Turnover was high at 879%, which is characteristic of active swing-trading strategies cycling in and out of oversold positions. Total fees were a flat $38 (one dollar per trade), a minor drag at this scale.
Walk-Forward Validation: The Problem
The strategy did not pass Headmars's walk-forward validation gate. Splitting the backtest period into four sequential folds tells a revealing story:
| 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 is an outlier — an 11.1% quarter with a Sharpe of 1.32 that inflates the full-period figures. Strip it out and the remaining three folds average a meager 0.48% return. The most recent out-of-sample fold (Fold 4) returned −2.84% with a −0.33 Sharpe and the steepest drawdown of any period. The Probabilistic Sharpe Ratio (PSR) of 0.785 is moderate, but the Deflated Sharpe Ratio (DSR) of just 0.304 — which corrects for the number of trials and selection bias across six strategy iterations — is the sharpest red flag here. It suggests that a meaningful portion of the apparent edge may be the product of parameter fitting rather than durable alpha.
Recent Live Activity
The paper-trade book has been quiet. The most recent executed trade was a WMT buy at $115.75 (21 shares) on May 31. Since then, six consecutive daily runs between June 25 and July 2 report zero executions — no RSI extremes in the current universe have cleared the entry threshold. The portfolio holds $7,569 in cash against a total value ranging from roughly $9,848 to $10,024 over that same window, implying open positions are drifting slightly with market moves but generating no fresh signals.
Risk Assessment
Strengths: High win rate, clean rule-based execution, and a broad blue-chip universe that limits liquidity risk. The strategy is easy to audit and explain.
Risks: Regime sensitivity is the primary concern. RSI mean-reversion strategies tend to underperform in trending markets, and Fold 4 coincides with a period of elevated macro volatility. The extended signal drought also raises the question of whether the strategy's RSI thresholds are calibrated too conservatively for current market conditions. Until the live book generates more trades across diverse regimes, the validation failure should be treated as a real warning rather than statistical noise.