Strategy Overview
Bollinger Reversion is a classic technical mean-reversion system: buy when price closes below the lower Bollinger Band, sell when it closes above the upper band. The universe spans 24 large-cap names across tech (AAPL, MSFT, GOOGL, NVDA), financials (JPM, BAC, V, MA), healthcare (JNJ, UNH, PFE, ABBV), consumer staples and discretionary (PG, KO, WMT, COST, MCD, NKE, HD, DIS), and energy/industrials (XOM, CVX, CAT, HON). The breadth is intentional — Bollinger signals are more reliable in mean-reverting, liquid names, and this roster leans heavily toward that profile.
Backtest Performance
Across 451 trading days and 76 completed round-trips, the strategy returned +17.55% (9.46% annualised CAGR) on a $10,000 starting equity, ending at $11,755. The win rate of 63.9% is respectable for a mean-reversion system — more than six in ten trades closed profitably.
The shadow is the risk profile. A 20.6% maximum drawdown is steep for a strategy whose whole premise is buying quality names at oversold levels. The Sharpe of 0.66 reflects a return stream that is positive but noisy, and the turnover figure of 1,721% signals frequent repositioning — relevant when scaling beyond paper trading, where friction compounds quickly.
Walk-Forward Validation: Where It Gets Complicated
The four-fold walk-forward tells a story of sharp decay:
| Fold | Period | Return | Sharpe | Max DD |
|---|---|---|---|---|
| 1 | Aug 2024 – Jan 2025 | +7.67% | 1.78 | 7.1% |
| 2 | Jan 2025 – Jul 2025 | +1.42% | 0.25 | 20.6% |
| 3 | Jul 2025 – Dec 2025 | +0.77% | 0.20 | 6.7% |
| 4 | Dec 2025 – May 2026 | +0.40% | 0.14 | 11.2% |
Fold 1 is genuinely strong — a 1.78 Sharpe with a sub-7% drawdown is the kind of performance a strategy should be built to replicate. Folds 2–4 show monotonically declining returns and risk-adjusted quality. The out-of-sample return lands at just +0.4% with a Sharpe of 0.14, and the Deflated Sharpe Ratio of 0.342 (against 6 trials) confirms what the raw numbers suggest: after correcting for multiple testing, there is not yet statistical confidence that the in-sample edge is real rather than fitted. Validation status: not passed.
All four folds are positive — which is mildly encouraging and rules out outright reversal — but positive-but-near-zero is not the bar.
Recent Activity
The strategy has been quiet since mid-June. The last executed trades — buys in WMT, COST, PG, and GOOGL on 31 May and 1 June, a sell of PG on 9 June, and a buy in DIS on 10 June — suggest the system found actionable oversold setups in the consumer and tech names during that window. Since 15 June, every scheduled run has produced zero executions, with 1–3 signals rejected daily. The portfolio sat at $10,001 on 22 June, essentially flat cash-equivalent, suggesting open positions have given back most of the mid-June gains.
Risks to Watch
- Regime sensitivity: Fold 1 captured a period well-suited to mean reversion. Trending or high-volatility regimes (see Fold 2's 20.6% drawdown) punish the strategy hard.
- Edge decay: The monotonic decline across folds is the key watch item. Another poor out-of-sample period would make the case for retiring or reparameterising the strategy.
- Turnover friction: At 1,721% annual turnover, live trading costs would meaningfully compress returns from the already-thin recent folds.
Bottom Line
Bollinger Reversion demonstrates a real, if fragile, mean-reversion thesis. The backtest headline is encouraging; the walk-forward is an honest corrective. The strategy stays live for continued data collection — one more fold of strong performance would materially shift the PSR and DSR scores. Until then, treat it as a monitored experiment rather than a capital-deployment candidate.