The thesis
bollinger-reversion is one of our built-in, hand-written strategies, and its rule set is about as classic as technical trading gets: buy when price closes below the lower Bollinger band, sell when it pushes above the upper band. It operates on a 24-name universe of US large caps — the familiar mega-cap tech, financials, healthcare and staples roster (AAPL, MSFT, GOOGL, NVDA, JPM, V, JNJ, PG, XOM and friends). It is live, and it was approved at risk score 0 back in May.
The bet is simple: in liquid, well-behaved blue chips, short-term dislocations tend to snap back. No earnings models, no macro calls — just bands and patience.
Recent activity
The live book has been doing exactly what a mean-reverter should. On 2 November it bought MSFT at $514.53; on 4 November it sold at $557.86 — roughly an 8% round trip. AAPL tells the same story: bought at $329.82 on the 3rd, sold at $352.27 the next day for about +7%. Earlier in the window it opened positions in CAT ($647.16) and GOOGL ($283.35). Execution has been clean — across the last five scheduled runs, every order filled and none were rejected. Account value has drifted from about $9,958 to $10,079 on a $10,000 base, so call it a modest ~+0.8% while the strategy churns in and out of positions.
The backtest headline
Over 451 days the full backtest returned 17.55% (final equity $11,755), a 9.46% CAGR with a 63.89% win rate across 76 trades. On paper, attractive. The caveats are in the texture: a 20.57% max drawdown against a Sharpe of only 0.66, and eye-watering 1,721% turnover. This is a strategy that trades constantly and earns its return through many small edges, not a few big ones.
Why validation rejected it
Here is the honest part: our walk-forward gate failed this strategy, and the fold data shows why. All four out-of-sample folds were positive — but they decay monotonically: fold 1 returned 7.67% at a 1.78 Sharpe, then 1.42%, then 0.77%, and the most recent fold just 0.40% at a 0.14 Sharpe. In other words, nearly the entire 17.55% headline was earned in the first five months of the test; the last year and a half is barely distinguishable from flat.
The risk-adjusted statistics agree. The deflated Sharpe ratio (DSR) lands at 0.342 — once you account for the 6 trials behind this configuration, there is not enough evidence the edge is real rather than lucky. PSR of 0.814 is respectable but not conclusive.
Verdict
bollinger-reversion is not broken — the recent AAPL and MSFT trades are textbook, and four positive folds is not nothing. But the trend is unmistakable: the edge was strongest early and has been bleeding out, consistent with a well-known, heavily-arbitraged pattern. We are keeping it live as a benchmark, but the failed gate is the right call. Watch the Sharpe and drawdown, not the 17.55%.