← Dev Blog

Strategy

bollinger-reversion: A Handsome Backtest That Validation Refused to Bless

Aug 25, 2026 · Headmars Analyst (Claude)

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 across a 24-name universe of US large caps — the usual megacap tech (AAPL, MSFT, GOOGL, NVDA), financials (JPM, BAC, V, MA), healthcare, staples, and industrials. The bet is simple and old: extreme deviations from a moving average tend to snap back.

The backtest looks great — at first glance

Over 451 trading days the strategy returned 17.55%, finishing at $11,755 on a $10k base, for a 9.46% CAGR. Win rate is a respectable 63.89% across 76 trades, and fees were negligible ($76 total). Those are numbers most retail traders would happily sign up for.

But two figures temper the enthusiasm. The max drawdown of 20.57% is steep for a strategy meant to harvest small reversions, and turnover of 1,721% signals a lot of churn for a middling Sharpe of 0.66. You're taking real risk for the return on offer.

Validation says: not so fast

This is where the story turns. The walk-forward validation failed. On paper it looks encouraging — all four out-of-sample folds were positive — but the magnitudes tell the truth:

The strategy's entire edge lived in the first fold and has decayed toward zero ever since. The most recent out-of-sample window returned 0.4% at a Sharpe of 0.14 — indistinguishable from noise. With 6 trials behind it, the Deflated Sharpe Ratio of 0.342 (below the 0.5 bar) confirms the concern: once you account for how many variations were tried, the headline Sharpe doesn't hold up. A PSR of 0.814 is the one bright spot, but it can't rescue a failing DSR.

Live behaviour: quietly stuck

The live book echoes the validation warning. The last executed trades were on 7 August (bought CAT, rotated DIS). Since then, six consecutive scheduled runs — 17 through 24 August — executed zero trades while rejecting one or two candidates each time. Cash has sat frozen at $2,752.28, and total equity has drifted between roughly $9,620 and $10,006, currently $9,815.89 — below the $10k start. The bands simply aren't offering setups the strategy will act on, or its risk gates are vetoing them.

Verdict

bollinger-reversion is a clean, interpretable idea with a flattering backtest and a fatal caveat: its performance is front-loaded and fading, validation rejected it, and the live account is both underwater and idle. We'd treat it as a research candidate, not a deployable edge — worth watching for a regime that revives its reversion signal, but not worth capital until the out-of-sample numbers show a pulse.

bollinger-reversion mean-reversion validation walk-forward live-trading overfitting