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. The idea is that extreme deviations from a moving-average envelope tend to snap back. It trades a broad, liquid universe of 24 large-cap US names spanning tech, financials, healthcare, staples, and energy — from AAPL and NVDA to JPM, JNJ, KO, and XOM.
Recent activity
The strategy is live and running on a scheduled daily cadence. Over the week of October 1–8, 2026, it logged mostly quiet sessions: several runs executed zero trades, with one or two fills on active days (and a couple of rejections on Oct 1 and Oct 2). Its paper book sits around $9,600–$9,680 total equity, with cash swinging between roughly $490 and $2,800 as positions open and close.
Recent fills are consistent with the mean-reversion mandate. On the buy side it has added JPM (7 shares @ $328.78), JNJ (9 @ $255.36), MCD (9 @ $250.63), and a large NKE position (70 @ $35.64) — names bought into weakness. On the sell side it trimmed strength: COST (2 @ $934.52), CAT (1 @ $847.03), GOOGL (6 @ $351.91), and WMT (21 @ $109.83). The behaviour matches the stated rules.
Backtest performance
Over a 451-day backtest the strategy returned 17.55%, ending at $11,755 equity, for a 9.46% CAGR. The win rate is a healthy 63.89% across 76 trades, with a Sharpe of 0.66 and a max drawdown of 20.57%. Turnover is high at roughly 1,721%, though total fees of just $76 and zero FX cost kept frictions modest in this paper setting.
The validation problem
Here the picture darkens. The walk-forward validation did not pass, and the reason is visible in the folds. All four out-of-sample windows were positive — but the edge decays monotonically: fold 1 returned 7.67% (Sharpe 1.78), then 1.42%, 0.77%, and finally just 0.40% (Sharpe 0.14) in the most recent window ending May 2026. The aggregate out-of-sample return is a thin 0.4% at a 0.14 Sharpe, far below the full-sample 0.66.
The statistics echo this. The Probabilistic Sharpe Ratio of 0.814 is encouraging, but the Deflated Sharpe Ratio — which penalises for the 6 trials run — falls to 0.342, below a confidence threshold. In plain terms: much of the headline return looks front-loaded into the earliest window, and the recent-period edge is marginal.
Verdict
Strengths are real: a clear, rule-based thesis, a high win rate, and consistently positive folds. But the fading out-of-sample returns and the failed validation gate are a genuine risk flag. This reads less like a broken strategy than a thinning one — worth keeping live and watched, not yet worth trusting with conviction capital.