The thesis
Bollinger-reversion is a classic mean-reversion play: buy when price closes below the lower Bollinger band, sell when it pushes above the upper band. The idea is that price extremes are temporary and tend to snap back toward a moving-average center. The agent runs this logic across a 24-name large-cap universe spanning tech (AAPL, MSFT, NVDA, GOOGL), financials (JPM, BAC, V, MA), staples (PG, KO, WMT, COST) and healthcare (JNJ, UNH, PFE, ABBV). It is currently flagged live.
Headline backtest
On paper, the numbers are inviting. Over 451 days the strategy returned 17.55% (9.46% CAGR), ending at $11,755 equity from a $10,000 base, with a 63.89% win rate across 76 trades. That high hit rate is the signature of mean reversion — many small wins, punctuated by the occasional deep loser.
The cracks show in the risk profile. Sharpe is a modest 0.66, and max drawdown reaches 20.57% — you had to stomach a one-fifth equity dip to collect that return. Turnover is a striking 1,720%, meaning the book churns roughly seventeen times over; fees stay low here ($76 total) but that intensity would bite harder under real spreads and slippage.
Where it breaks: validation
This is the part that matters, and the verdict is blunt: validation failed. All four walk-forward folds are positive, which is genuinely encouraging — the edge is directionally consistent, not a single lucky window. But the trend inside those folds is the problem:
- Fold 1 (Aug 2024–Jan 2025): +7.67%, Sharpe 1.78
- Fold 2: +1.42%, Sharpe 0.25
- Fold 3: +0.77%, Sharpe 0.20
- Fold 4 (Dec 2025–May 2026): +0.40%, Sharpe 0.14
Performance decays monotonically toward zero. The out-of-sample return lands at just 0.40% with an OOS Sharpe of 0.14 — statistically indistinguishable from flat. The Probabilistic Sharpe Ratio (0.814) looks fine in isolation, but the Deflated Sharpe Ratio — which penalizes for the 6 trials run — collapses to 0.342. Once you account for how many variants were tried, the surviving edge is thin.
Live behavior
The recent tape reinforces the caution. Scheduled runs from Aug 11–17 executed zero trades, with one or two signals rejected each session; equity has drifted sideways near $10,000 on $2,752 idle cash. The last real fills were back on Aug 7 (CAT buy, DIS round-trip). The strategy is effectively parked, waiting for band-breaks that aren't coming in a calmer market.
Verdict
Strengths: a coherent, well-understood thesis, four consecutive positive folds, and a strong win rate. Risks: front-loaded returns that fade to noise out-of-sample, a punishing drawdown, and a deflated Sharpe that flunks the overfitting screen. This is a strategy that looks live but hasn't earned promotion — best treated as a monitored candidate, not deployed capital.