The Thesis
Bollinger-reversion is about as classical as systematic trading gets: buy when price closes below the lower Bollinger band, sell when it pushes above the upper band. It is a pure mean-reversion bet — the wager that stretched moves snap back toward a moving average. The strategy trades a 24-name large-cap universe spanning tech, financials, staples, healthcare and energy (AAPL, MSFT, NVDA, JPM, V, JNJ, PG, WMT, XOM and peers), so it is diversified across sectors rather than concentrated in one regime.
Headline Performance
On paper, the full-sample backtest looks appealing. Over 451 days it returned 17.55%, ending on $11,755.25 of equity, a 9.46% CAGR, with a 63.89% win rate across 76 trades. The catch is quality-of-return: Sharpe is a modest 0.66, max drawdown reached 20.57%, and turnover ran to a punchy 1,720% — this strategy trades a lot to earn its keep, though total fees were a trivial $76.
Where It Breaks: Walk-Forward Validation
This is where the story turns. The strategy failed validation. All four walk-forward folds were positive, which sounds reassuring, but the returns decay almost monotonically: fold 1 delivered 7.67% at a strong 1.78 Sharpe, then folds 2–4 collapsed to 1.42%, 0.77% and finally 0.40% — with fold-4 Sharpe of just 0.14. The most recent out-of-sample window is effectively flat.
The risk statistics reinforce the concern. The Probabilistic Sharpe Ratio (0.814) is decent, but the Deflated Sharpe Ratio is only 0.342 — and deflation matters here because six trials were tested. A DSR that low suggests the headline Sharpe does not comfortably survive multiple-testing adjustment. In plain terms: the early performance looks like it may have been the easy part of the sample, not a repeatable edge.
Recent Live Activity
Live behaviour matches the sputtering out-of-sample picture. Across the last six scheduled runs (Sept 18–25), most executed zero trades and logged rejections, with only two active days: Sept 18 (buy NKE, sell GOOGL) and Sept 22 (buy MCD, sell WMT). The paper account currently sits around $9,650–$9,780 total with roughly $2,398 in cash — the band signals simply are not firing often, and when they do the account has drifted below its starting line.
The Verdict
Strengths: a well-understood, sector-diversified thesis, a genuinely high win rate, and every validation fold in positive territory. Risks: a failed validation gate, a low deflated Sharpe against six trials, a 20%+ drawdown, and a clear performance fade both out-of-sample and in recent live runs. The honest read is that bollinger-reversion captured a favourable early regime and has since flattened. I would keep it running as a paper experiment and a diversifier, but its recent record does not justify escalating capital until it demonstrates edge in a fresh out-of-sample window.