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 universe of large-cap US equities — AAPL, MSFT, GOOGL and NVDA alongside defensives like JNJ, PG, KO and WMT. The logic bets that stretched prices snap back toward their moving average, and on a diversified blue-chip basket that is a reasonable prior.
The backtest looks good
The headline numbers flatter the strategy. Over 451 days it returned 17.55% (9.46% CAGR), finishing at $11,755 on a $10k base, with a 63.89% win rate across 76 trades. That hit rate is the strongest part of the story — mean-reversion systems tend to win often and small, and this one fits the mould.
The caveats sit right next to it. Sharpe is a modest 0.66, max drawdown is a steep 20.57%, and turnover is a punishing 1,720%. High turnover means the edge has to survive real-world frictions; here fees were only $76, but live spreads and slippage could erode a thin margin.
Validation says wait
This is where the case weakens. Four-fold walk-forward analysis fails the gate. Every fold is positive, which sounds encouraging, but the returns decay monotonically: 7.67% → 1.42% → 0.77% → 0.40%, with fold Sharpe collapsing from 1.78 to 0.14. Out-of-sample return is essentially flat at 0.4% (OOS Sharpe 0.14). The deflated Sharpe ratio of 0.342 — after accounting for 6 trials — sits well below the level that would give confidence the edge is real rather than the best draw from a small search. A PSR of 0.814 is decent in isolation, but the DSR is the honest number, and it is telling us the early-fold strength was likely regime luck, not durable signal.
Live activity has stalled
The paper account reflects that softness. The last executed trades were in early June — buys in DIS, GOOGL, WMT and COST, a PG round-trip. Since then, every scheduled run from July 23 through July 30 reports 0 executed, 1–2 rejected, with cash pinned at $1,001.98 and total equity drifting between roughly $9,350 and $9,870 — under the $10k mark. Either no names are hitting the band triggers or orders are bouncing on sizing/cash constraints worth investigating.
Verdict
Bollinger-reversion has a coherent thesis and an attractive win rate, but the evidence for a persistent edge is thin: decaying folds, a weak deflated Sharpe, and a failed validation gate. The recent run of rejected orders suggests it is also idle in the current regime. I'd keep it in paper, tighten the position-sizing logic behind those rejections, and treat the 17.55% backtest as a hypothesis — not a green light.