The Thesis
bollinger-reversion runs a classic mean-reversion playbook: buy when a name drops below its lower Bollinger band and sell when it pushes above the upper band. The universe is 24 large-cap U.S. names spanning tech (AAPL, MSFT, NVDA), financials (JPM, V, MA), staples (PG, KO, COST) and healthcare (JNJ, UNH). It's a well-understood, liquid playground — exactly where band-reversion is supposed to behave.
Backtest Performance
On the full sample the numbers look inviting. The strategy returned 17.55% over 451 days (a 9.46% CAGR), closing at $11,755 of final equity from 76 trades with a 63.89% win rate and a Sharpe of 0.66. The win rate is the headline strength — nearly two-thirds of trades landed on the right side.
The caveats are just as visible. Max drawdown reached 20.57%, a steep price for a 0.66 Sharpe, and turnover was an eye-watering 1,720%. That churn rack up costs; here fees totalled $76, but in a live book with slippage the drag would bite harder.
Validation: The Warning Signs
This is where the story turns. The walk-forward validation failed. All four folds were positive, which sounds reassuring — until you read them in order:
| 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 |
That is a monotonic decay — the edge is almost entirely front-loaded into the earliest window and fades toward noise. Out-of-sample return sits at just 0.4% with a Sharpe of 0.14. The Deflated Sharpe Ratio of 0.342 (against 6 trials) confirms the concern: adjusted for selection, the confidence that this is real alpha is weak, even though the raw Probabilistic Sharpe (0.814) looks healthier.
Recent Live Activity
The live book echoes the validation story. The last executed trades were back in early June — buys in DIS, GOOGL, PG, WMT and COST — and every scheduled run since mid-July has done nothing, logging "0 executed" with occasional rejections. Portfolio value has drifted down from about $9,925 (July 16) to $9,748 (July 20), with cash pinned at $1,001.98. The bands simply aren't triggering clean entries, and the strategy is riding its existing positions lower.
The Verdict
bollinger-reversion is a tidy idea with a genuine strength — a high win rate — but the evidence points to an edge that has eroded. The fold-by-fold decay, near-zero out-of-sample return, sub-threshold DSR and a stalled, drifting live portfolio all say the same thing: what looked like alpha in 2024 reads more like a fading regime effect today. It stays worth watching as a live probe, but on this data it does not clear the bar for conviction capital.