Thesis
Bollinger Reversion is a textbook mean-reversion system: buy when price closes below the lower Bollinger Band, sell when it exceeds the upper band. The logic rests on the empirical tendency of liquid, large-cap equities to revert toward their short-term moving average after sharp dislocations. The 24-stock universe — spanning mega-cap tech (AAPL, MSFT, NVDA, GOOGL), financials (JPM, V, MA), consumer staples (PG, KO, WMT), and healthcare (JNJ, UNH) — is deliberately tilted toward names with deep order books where dislocations are more likely to be noise than regime change.
Backtest Performance
Over 451 trading days the strategy compounded to a 17.55% total return (9.46% CAGR), executing 76 trades at a 63.89% win rate. Those are respectable headline numbers. The Sharpe of 0.66 is modest but positive, and fees were minimal at $1 per trade.
The concern lives in the drawdown profile: a 20.57% maximum drawdown is steep for a strategy that only holds 24 names. Most of that damage was concentrated in early 2025 (Fold 2), where the strategy gave back nearly 20% in a single period while capturing only 1.42% in return. Turnover of 1,720% annualised is also high, suggesting the signal fires often — a dynamic that amplifies both fees and slippage risk in a real-money environment.
Walk-Forward Validation: A Troubling Fade
The four-fold walk-forward tells a consistent and uncomfortable story:
| Fold | Period | Return | Sharpe | Max DD |
|---|---|---|---|---|
| 1 | Aug 2024 – Jan 2025 | 7.67% | 1.78 | 7.06% |
| 2 | Jan 2025 – Jul 2025 | 1.42% | 0.25 | 20.56% |
| 3 | Jul 2025 – Dec 2025 | 0.77% | 0.20 | 6.66% |
| 4 | Dec 2025 – May 2026 | 0.40% | 0.14 | 11.19% |
All four folds were technically positive, but Fold 1's outsized alpha has not repeated. The out-of-sample Sharpe landed at 0.14 and the Deflated Sharpe Ratio (DSR) — which penalises for the six parameter trials run — came in at 0.342, well below the 0.50 threshold required to pass the validation gate. The PSR of 0.814 says the raw Sharpe is probably above zero; the DSR says, once you account for the search process, the edge is statistically thin. The strategy has failed validation.
Recent Activity
Live operation has been quiet. Every scheduled run from June 25 through July 2 recorded zero executed trades, with two to three signals rejected each session. The portfolio has drifted between roughly $9,700 and $9,880 during this window. The last cluster of real activity came in late May and early June — buys in WMT, COST, PG, and GOOGL, followed by a quick exit from PG nine days later at a modest gain. The current cash reserve sits near $1,002 against a total book of ~$9,861.
The quiet period could reflect an orderly market where few names are trading outside their bands — consistent with what mean-reversion strategies experience in low-volatility regimes.
Outlook
Bollinger Reversion demonstrates that a simple, rules-based mean-reversion signal can generate real alpha in large-cap equities — but the evidence of fade is hard to ignore. The strategy's best days may correspond to a volatility environment (late 2024) that has since normalised. Before scaling capital, the team should investigate whether parameter re-optimisation or a volatility filter (e.g., only trading when VIX is elevated) can restore the early-period edge out-of-sample. Until DSR clears the gate, this one stays in observation mode.