The thesis
Bollinger-reversion runs one of the oldest ideas in technical trading: price tends to snap back toward its mean after stretching too far. Concretely, it buys when price closes below the lower Bollinger band and sells when it pushes above the upper band, across a 24-name large-cap universe spanning tech, financials, healthcare, staples and energy (AAPL, MSFT, NVDA, JPM, JNJ, XOM and friends). It's currently flagged live.
Backtest headline
Over 451 days the strategy turned a $10,000 base into $11,755 — a +17.55% total return, roughly 9.46% CAGR. The win rate is genuinely good at 63.89% over 76 trades, and fees were negligible ($76 total, no FX cost). Those are the numbers a mean-reversion fan wants to see.
The cracks show elsewhere. Sharpe is a modest 0.66, max drawdown is a steep 20.57%, and turnover is a punishing 1,720% — this book churns its capital roughly seventeen times, so it lives and dies on clean fills and tight costs.
The validation problem
This is where the update turns cautionary. Validation did not pass. The 4-fold walk-forward tells a clear decay story:
| Fold | Window | Return | Sharpe |
|---|---|---|---|
| 1 | 2024-08 → 2025-01 | 7.67% | 1.78 |
| 2 | 2025-01 → 2025-07 | 1.42% | 0.25 |
| 3 | 2025-07 → 2025-12 | 0.77% | 0.20 |
| 4 | 2025-12 → 2026-05 | 0.40% | 0.14 |
All four folds are positive — a point in its favour — but the edge is almost entirely front-loaded into the first window and bleeds toward zero thereafter. The most recent out-of-sample fold returned just 0.40% at a Sharpe of 0.14. The probabilistic Sharpe (PSR 0.814) looks encouraging in isolation, but the deflated Sharpe of 0.342 — which penalises the 6 trials behind this configuration — sits well below any level that would let us call the edge statistically real rather than lucky.
Live activity
Recent live runs reinforce the sluggishness. Across six scheduled sessions from 19–26 October, five executed zero trades (1–2 rejected each), with only the 21 October run firing — a 3-share CAT buy and a 9-share MCD sell. Earlier October trades include NVDA, JPM and JNJ buys against NKE, COST and WMT sells. The live book now stands at roughly $9,910 with $2,667 in cash, modestly below the $10,000 the backtest started from.
Verdict
Bollinger-reversion is a well-behaved, high-win-rate mean-reverter with a plausible thesis and consistently positive folds — but the evidence points to a shrinking, possibly ephemeral edge. The decaying walk-forward curve, a deflated Sharpe that fails the multiple-trials test, a 20%+ drawdown and heavy turnover all argue for keeping it on a short leash. It earns its live slot as a watch-and-learn candidate, not a conviction allocation. The next few out-of-sample months should settle whether there's anything left to reap.