The thesis
bollinger-reversion is about as classic as systematic trading gets: buy when price closes below the lower Bollinger band, sell when it pushes above the upper band. It trades a 24-name large-cap universe spanning tech (AAPL, MSFT, NVDA), payments (V, MA), staples (PG, KO, WMT, COST), healthcare (JNJ, UNH, PFE) and industrials (CAT, HON). The bet is simple — stretched names snap back to their moving average. It's a reasonable prior in range-bound markets and a dangerous one in trending ones.
Backtest performance
Over 451 days the strategy returned 17.55% (9.46% CAGR), ending at $11,755 on a $10,000 base, with a 63.89% win rate across 76 trades. That hit rate is its headline strength. The caveats are just as visible: a Sharpe of only 0.66, a 20.57% max drawdown, and eye-watering 1,720% turnover. High turnover means the strategy lives and dies by execution costs — here fees were a modest $76, but in any less forgiving cost regime that churn would bite.
Validation: the red flag
This is where the story turns. The four-fold walk-forward test is marked as failed, and it fails for an instructive reason. All four folds are positive, which looks encouraging — until you read them in order:
- Fold 1 (Aug 2024–Jan 2025): +7.67%, Sharpe 1.78
- Fold 2: +1.42%, Sharpe 0.25
- Fold 3: +0.77%, Sharpe 0.20
- Fold 4 (ends May 2026): +0.40%, Sharpe 0.14
That is a textbook decay curve. Almost the entire backtest edge came from the first fold; by the most recent out-of-sample window the strategy is barely distinguishable from flat. The out-of-sample return of 0.4% and OOS Sharpe of 0.14 confirm it. The deflated Sharpe ratio of 0.342 — which penalises the 6 trials run — sits well below any sensible deploy threshold, even though the probabilistic Sharpe (0.814) reads benignly on its own. The gate is right to block this one.
Recent live activity
In paper trading the strategy is largely idle. The last six scheduled runs executed almost nothing — mostly "0 executed, 1 rejected" — with the only real action on 2026-10-02 (a JNJ buy and a CAT trim). The live account total has drifted around $9,500–$9,700, below its $10,000 start, consistent with a mean-reverter finding few genuinely stretched names in the current tape.
Verdict
Strengths: a coherent, well-understood thesis and a strong win rate. Risks: a mediocre Sharpe, a 20%+ drawdown, punishing turnover, and — decisively — an edge that visibly erodes out of sample. It remains live for observation, but on this evidence it is not a candidate for real capital.