The thesis
Bollinger-reversion is a classic mean-reversion play: buy when price closes below the lower Bollinger band, sell when it pushes above the upper band. The bet is that extreme moves revert, so the strategy fades strength and accumulates weakness. It runs over a 24-name large-cap universe spanning tech (AAPL, MSFT, NVDA, GOOGL), financials (JPM, V, MA), staples (PG, KO, WMT, COST) and energy (XOM, CVX). It is currently live.
Recent activity
The strategy has been active nightly. In the last week it bought XOM (15 @ $166.26), NVDA (13 @ $184.15), GOOGL (8 @ $283.35) and CAT (3 @ $647.16), and turned two tidy round-trips: AAPL bought at $329.82 on Nov 3 and sold at $352.27 the next day (+6.8%), and MSFT bought at $514.53 on Nov 2, sold at $557.86 on Nov 4 (+8.4%). Those are exactly the fast reversion wins the thesis predicts.
That said, the live book is roughly flat. The Nov 6 run reports total equity of $9,956.62, and daily totals have hovered just under $10,000 all week ($9,830–$10,079). The engine is executing cleanly — most runs show 1–2 fills and zero rejections — but it has not yet pulled ahead live.
Backtest performance
Over 451 days the backtest returns 17.55% (final equity $11,755), a 9.46% CAGR, with a 63.89% win rate across 76 trades and a Sharpe of 0.66. Fees were a trivial $76 with no FX cost. The headline numbers are solid, if unspectacular for a Sharpe below 1.
The warning sign is the max drawdown of 20.57% — large relative to a mid-single-digit CAGR — and turnover of 1,720%, confirming this is a high-churn strategy whose edge must survive frequent trading.
Validation: why it fails the gate
This is where the picture turns cautious. Validation did not pass. All four walk-forward folds are positive, which sounds encouraging, but they decay monotonically:
- 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 (Dec 2025–May 2026): +0.40%, Sharpe 0.14
Nearly all the return came from the earliest window; the most recent out-of-sample fold earned just 0.4% at a Sharpe of 0.14. The Probabilistic Sharpe Ratio of 0.814 looks healthy, but the Deflated Sharpe Ratio of 0.342 — which penalises the 6 trials behind this result — says the edge is not statistically convincing once selection bias is accounted for.
Verdict
Bollinger-reversion has a coherent, time-tested thesis and a clean, low-cost backtest with a genuinely good win rate. But the honest read is that its edge appears to have faded: strong early, marginal lately, and failing the deflated-Sharpe gate. The live account's flat week is consistent with that. I would treat it as a monitor-and-learn candidate rather than a conviction allocation until a recent fold reasserts the edge.