Thesis
Bollinger-reversion bets on statistical mean-reversion across a universe of 24 liquid large-caps spanning technology, financials, healthcare, consumer staples, and energy. The rule is deliberately simple: enter long when price closes below the lower Bollinger band — a stretched, oversold condition — and exit when price recovers above the upper band. No leverage, no shorting, no sentiment filters. The appeal is tractability: every signal and exit is mechanically auditable.
Backtest Performance
Over 451 days the strategy returned 17.55% on a starting equity of $10,000, compounding at a 9.46% CAGR. The win rate of 63.89% across 76 trades is genuinely competitive for a passive, signal-driven approach. Maximum drawdown came in at 20.57%, tolerable in absolute terms but worth monitoring given the portfolio concentration in single-name equities.
One eyebrow-raiser: portfolio turnover printed at 1,720%, implying the strategy cycles its notional capital roughly 17 times over the period. In a live account that figure would materially erode returns through commissions and market impact; in this paper-trading environment the $76 in modelled fees is essentially a rounding error.
Walk-Forward Validation: The Cautionary Story
The four-fold walk-forward test passes on one narrow criterion — every fold was profitable — but the trajectory is the concern:
| Fold | Period | Return | Sharpe | Max DD |
|---|---|---|---|---|
| 1 | Aug 2024 – Jan 2025 | 7.67% | 1.78 | 7.06% |
| 2 | Jan 2025 – Jul 2025 | 1.42% | 0.25 | 20.56% |
| 3 | Jul 2025 – Dec 2025 | 0.77% | 0.20 | 6.66% |
| 4 | Dec 2025 – May 2026 | 0.40% | 0.14 | 11.19% |
Fold 1 drove virtually the entire headline return. By Fold 4, the out-of-sample Sharpe has collapsed to 0.14 and the out-of-sample return to 0.40% — barely above cash. The Deflated Sharpe Ratio of 0.342 (well below conventional thresholds) and a Probabilistic Sharpe Ratio of 0.814 confirm that the platform's own validator has flagged this run as not passing. With six trials in the parameter search, there is meaningful probability that the attractive Fold 1 metrics reflect lucky parameter alignment rather than a robust signal.
Recent Activity: The Strategy Goes Quiet
Live operation tells a similar story. The last executed trades cluster around late May and early June — buying WMT, COST, GOOGL, and PG, then closing the PG position nine days later for a modest gain. Since June 19, the strategy has run daily and rejected every candidate signal (2–3 per day), executing zero trades. Portfolio value has drifted from $10,172 on June 19 to $9,879 on June 26, reflecting open-position mark-to-market rather than new activity.
The silence is not necessarily alarming — a mean-reversion strategy should sit on its hands when prices are not at extremes — but the combination of rising rejection counts and a declining portfolio total suggests current holdings are giving back some of the earlier gains.
Strengths and Risks
Strengths: High win rate, full transparency of signal logic, a universe limited to liquid blue-chips, and no fold that lost money in absolute terms.
Risks: The monotonic Sharpe decay across folds is the dominant concern. A strategy that earns 1.78 Sharpe in its first period and 0.14 in its most recent period has either encountered a regime change or was overfit to early-period conditions. High turnover also creates execution sensitivity: slippage assumptions that look benign in paper trading may not survive real fills on 17× annual turnover. Until a fresh out-of-sample period shows stabilisation in risk-adjusted returns, treating the full-period 17.55% headline as forward-looking expectation would be premature.