Strategy Overview
Bollinger-reversion follows one of the oldest mean-reversion playbooks in technical analysis: buy when price closes below the lower Bollinger band, sell when it breaches the upper band. The universe is a 24-stock basket of liquid US large-caps spanning technology (AAPL, MSFT, GOOGL, NVDA), financials (JPM, BAC, V, MA), healthcare (JNJ, UNH, PFE, ABBV), consumer staples (PG, KO, WMT, COST), and select cyclicals (MCD, NKE, HD, XOM, CVX, CAT, HON, DIS).
The thesis is sensible: in a basket of high-quality names, extreme short-term dislocations tend to resolve. You're not betting on trend — you're betting on gravity.
Full-Period Backtest
Over 451 days of paper trading, the strategy logged:
| Metric | Value |
|---|---|
| Total Return | 17.55% |
| CAGR | 9.46% |
| Sharpe Ratio | 0.66 |
| Max Drawdown | 20.57% |
| Win Rate | 63.89% |
| Trades | 76 |
| Fees | $76 (flat) |
A 63.9% win rate and positive CAGR are genuinely encouraging. The strategy did what it said it would do in aggregate. The Sharpe of 0.66, however, sits below the 1.0 threshold typically associated with strategies worth serious capital allocation, and a 20.57% peak drawdown is steep for a basket of investment-grade names.
Walk-Forward Validation: The Real Story
The validation run across four sequential folds is where the picture gets complicated.
| Fold | Period | Return | Sharpe | Max DD | Trades |
|---|---|---|---|---|---|
| 1 | Aug 2024 – Jan 2025 | +7.67% | 1.78 | 7.06% | 24 |
| 2 | Jan 2025 – Jul 2025 | +1.42% | 0.25 | 20.56% | 11 |
| 3 | Jul 2025 – Dec 2025 | +0.77% | 0.20 | 6.66% | 18 |
| 4 | Dec 2025 – May 2026 | +0.40% | 0.14 | 11.19% | 20 |
Fold 1 was exceptional — a Sharpe of 1.78 and only a 7% drawdown. Every subsequent fold deteriorated monotonically. By Fold 4, the out-of-sample return had compressed to 0.40% with a Sharpe of 0.14. The Probabilistic Sharpe Ratio (PSR) lands at 0.814, and the Deflated Sharpe Ratio (DSR) — which penalizes for the number of trials tested — falls to 0.342. Against 6 strategy trials, the validation gate correctly returned passed: false.
The degradation pattern is consistent with a strategy that overfit to 2024's volatility regime and has not adapted as conditions evolved.
Recent Activity
The past week has been quiet. Between June 1–8, the strategy executed trades only on June 1 (GOOGL at $377.90 × 6 shares; PG at $140.03 × 17 shares) after earlier entries in WMT and COST on May 31. The four subsequent daily runs executed zero trades, with one rejection surfacing on multiple days — likely a position-sizing or risk-limit check blocking a signal that didn't quite clear the threshold.
Portfolio equity fluctuated between $9,870 (June 2) and $10,206 (June 5), with the June 8 reading at $10,132. Cash is parked at $1,008.70, suggesting the strategy is mostly deployed.
Strengths and Risks
Strengths: Positive return across all four validation folds is notable — the strategy never lost money out-of-sample, even as edge compressed. The high-quality universe limits catastrophic single-name risk, and the flat-fee structure keeps costs transparent.
Risks: The monotonic Sharpe decline is the dominant concern. A DSR of 0.34 means the observed edge is likely not statistically significant once trial inflation is accounted for. The strategy may also be regime-sensitive: Fold 1's strong performance coincided with a period of elevated mean-reversion opportunity that has since faded. A 20.57% maximum drawdown for a mean-reversion strategy on blue chips is also higher than ideal.
Outlook
Bollinger-reversion is a competent baseline that demonstrates the platform's paper-trading infrastructure works end-to-end. But the walk-forward evidence argues against live capital deployment in its current form. The next logical step is parameter recalibration — band width, lookback period, position sizing — followed by a fresh walk-forward run before revisiting the validation gate.