The thesis
bollinger-reversion is a textbook mean-reversion play: buy when price slips below the lower Bollinger band, sell when it pushes above the upper band. The premise is that short-term dislocations in large, liquid names revert. The universe backs that up — 24 mega-caps spanning tech, financials, healthcare, staples and energy (AAPL, MSFT, NVDA, JPM, UNH, XOM, and peers). These are exactly the kind of well-covered names where band touches tend to be noise rather than regime change, which is the right habitat for this idea.
Recent activity
The strategy is live and has been trading, though lightly of late. Late October and early November saw real positioning: buys in GOOGL and CAT (29 Oct), MSFT and AAPL (2–3 Nov), then a notable round-trip in AAPL — bought at $329.82 on 3 Nov, sold at $352.27 the next day — alongside an MSFT trim and fresh buys in NVDA ($184.15) and XOM ($166.26). That AAPL flip is the thesis working as designed: a quick reversion capture.
Since then it has gone quiet. The last three scheduled runs (9–11 Nov) executed nothing, with one rejection on 9 Nov. Portfolio total sits at $9,994.82 against a $10,000 start — essentially flat, and fractionally underwater. Cash has been drawn down to $1,055.32, so the book is now mostly invested and waiting.
Backtest vs. validation
On paper the backtest flatters the strategy: +17.55% total return over 451 days, a 9.46% CAGR, a 63.89% win rate across 76 trades, and final equity of $11,755. Sharpe is a modest 0.66, and max drawdown of 20.57% is already a warning that the ride is not smooth. Turnover of 1,721% is high — this is an active strategy whose edge has to clear its own churn.
The validation tells a harder story, and it is the part worth dwelling on. Validation did not pass. All four walk-forward folds are positive, which sounds reassuring until you read them in order: +7.67% (Sharpe 1.78), then +1.42%, +0.77%, and +0.40% (Sharpe 0.14). That is a near-monotonic decay toward nothing. The out-of-sample return is just 0.4% with an OOS Sharpe of 0.14 — the edge essentially lives in the earliest, most in-sample window.
The deflated metrics confirm the skew. PSR of 0.814 looks healthy, but the Deflated Sharpe Ratio — which penalises for the 6 trials run — is only 0.342, below the 0.5 line you would want before trusting the result. In plain terms: the headline return is probably more a product of a favourable early period and selection across trials than a durable, repeatable edge.
Verdict
Strengths are real — a coherent thesis, a sensible universe, a positive win rate, and no negative fold. But the risks dominate: fading out-of-sample performance, a failed validation gate, a low DSR, and a 20% drawdown. Treat bollinger-reversion as a candidate under observation, not a conviction allocation, until later folds stop shrinking.