The thesis
Bollinger-reversion is a textbook mean-reversion play: buy when price closes below the lower Bollinger band, sell when it pushes above the upper band. It runs live across a 24-name large-cap universe spanning tech (AAPL, MSFT, NVDA), financials (JPM, V, MA), staples (PG, KO, COST), and healthcare (JNJ, UNH). The premise is that liquid blue chips overshoot their volatility envelope and revert — a well-worn but crowded edge.
Backtest scorecard
On paper, the numbers look respectable. Over 451 days the strategy returned 17.55% (final equity $11,755 from a $10k base), a 9.46% CAGR, with a 63.89% win rate across 76 trades. That hit rate is the headline strength — mean-reversion systems typically win often and small.
The caveats sit right beside it. The Sharpe of 0.66 is mediocre for a strategy carrying a 20.57% max drawdown — you endured a fifth of your capital underwater to earn that return. Turnover of 1,720% is punishing: the book churns roughly seventeen times over, and while fees here totalled only $76, that level of trading is fragile to slippage the backtest may not fully price.
Validation: the part that matters
This is where the story turns. Walk-forward validation fails. All four folds are technically positive, but the trend is a steady bleed of edge:
- Fold 1 (Aug 2024–Jan 2025): +7.67%, Sharpe 1.78
- Fold 2: +1.42%, Sharpe 0.25, with a 20.56% drawdown
- Fold 3: +0.77%, Sharpe 0.20
- Fold 4 (Dec 2025–May 2026): +0.40%, Sharpe 0.14
The most recent out-of-sample window returned essentially nothing. The aggregate OOS return of 0.40% and OOS Sharpe of 0.14 tell you the full-sample 17.55% is largely front-loaded into one strong early period. Supporting stats agree: the Probabilistic Sharpe Ratio (0.814) is borderline, but the Deflated Sharpe Ratio of 0.342 — which penalises for the 6 trials run — sits well below the confidence you'd want before trusting the edge is real rather than mined.
Recent activity
Live behaviour reinforces the concern. The last executed trades were in early June (buys in DIS, GOOGL, WMT, COST; a PG round-trip). Every scheduled run since late July has logged 0 executed, 1–2 rejected, with cash frozen at $1,001.98 and total equity drifting from ~$9,962 down through the $9,600s. The strategy simply isn't finding entries that clear its filters — consistent with a mean-reversion signal that has gone quiet in the current regime.
Verdict
Bollinger-reversion has a genuine strength — a high, stable win rate — but it is not validation-clean, and I would not promote it on the strength of the headline return. The decaying fold sequence, weak deflated Sharpe, and stalled live execution all point the same way: the edge was strongest in the past and is thin today. It earns continued paper-trading observation, not fresh capital, until out-of-sample performance recovers.