The thesis
The premise is about as classic as systematic trading gets: buy when price slips below the lower Bollinger band, sell when it pokes above the upper band. In other words, bet on mean reversion — that stretched prices snap back toward their moving average. The strategy runs this logic across a 24-name universe of large-cap U.S. equities spanning tech (AAPL, MSFT, NVDA), financials (JPM, V, MA), staples (PG, KO, WMT, COST) and healthcare (JNJ, UNH, PFE). It is clean, interpretable, and exactly the kind of well-understood signal you want as a baseline.
Backtest performance
On paper, the numbers are inviting. Over 451 trading days the strategy returned 17.55%, lifting a notional book to $11,755 for a CAGR of 9.46%. Win rate was a healthy 63.89% across 76 trades, and fees were a trivial $76 with no FX cost. The catch is quality-adjusted return: a Sharpe of 0.66 is mediocre, and a maximum drawdown of 20.57% means you'd have stomached a one-fifth equity dip to collect those gains. Turnover of 1,720% also signals a strategy that trades constantly to extract a modest edge.
Where validation breaks down
This is the part that should temper any enthusiasm. The walk-forward validation did not pass. All four folds were positive, which sounds reassuring — until you read them in sequence:
- 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
That is a textbook decay curve. The headline backtest return was effectively earned in the earliest window, with the edge thinning toward nothing more recently. The most recent out-of-sample fold returned just 0.4% at a Sharpe of 0.14. The deflated Sharpe ratio (DSR) of 0.342 — which penalizes the 6 trials run — sits well below the confidence you'd want, even though the probabilistic Sharpe (PSR) of 0.814 looks superficially fine.
Live reality
The live account echoes the warning. Recent scheduled runs have been unproductive: on most days through late September 2026 the strategy logged 0 executed, 1 rejected, with cash frozen at $2,398.39 and total equity drifting between roughly $9,600 and $9,734 — under the $10,000 it would have started with. The last genuine fills were in September (buying MCD and NKE, selling WMT and GOOGL) and earlier round-trips in DIS and PG. Right now the signal is firing but orders are being rejected, so the book is largely idling.
Verdict
bollinger-reversion is a strong teaching example and a plausible diversifier, but it is not validated for conviction capital. The strengths are a high win rate and a transparent, robust signal; the risks are a decaying out-of-sample edge, a failed validation gate, and a live account currently underwater. Treat it as a candidate for refinement — not promotion.