Thesis and Mechanics
Channel-pullback is a classic trend-continuation play with a disciplined mean-reversion entry. The strategy targets confirmed uptrends across a 24-name large-cap universe — spanning technology, financials, healthcare, consumer staples, and energy — and enters long when price retreats to the lower regression channel or a volume-support level. Exits target the upper channel or a nearby resistance zone.
The logic is intuitive: buy the dip inside a trend, not against it. By anchoring entries to regression channels rather than arbitrary percentage moves, the strategy tries to distinguish structural pullbacks from trend breaks.
First Week Live
Deployed on June 1, 2026 with a $10,000 paper allocation, the strategy executed five opening trades immediately, seeding positions in JPM, COST, and others. Over the following four trading days it added positions in UNH, WMT, MSFT, and DIS while rotating out of CAT, JPM (re-entered), UNH (re-entered at a lower price), and a large DIS block. By June 4 the portfolio had reached $10,380 — roughly a 3.8% gain in the opening week.
Signal activity has been selective: June 3 and June 5 produced zero executions, with one to two rejections each day, suggesting the strategy is finding few qualifying pullbacks in the current environment rather than over-trading.
Backtest Performance
Over 451 days (roughly 137 trades), the full backtest returned +7.62% with a 4.19% CAGR and a Sharpe of 0.40. Those headline numbers are modest. The more telling figure is the 39.4% win rate — the strategy wins on fewer than two in five trades, which means it is structurally dependent on letting winners run significantly farther than losers. Whether that plays out consistently is the central question.
Maximum drawdown hit 14.83%, and annualised turnover was an eye-catching 2,311% — the portfolio turns over roughly 23× per year, creating meaningful friction even at $1-per-trade flat fees.
Cross-Validation: Wide Variance, Failed Gate
The four-fold walk-forward validation is where the picture becomes complicated.
| Fold | Period | Return | Sharpe | Max DD |
|---|---|---|---|---|
| 1 | Aug 2024 – Jan 2025 | +6.53% | 1.25 | 6.16% |
| 2 | Jan 2025 – Jul 2025 | −11.42% | −1.70 | 16.09% |
| 3 | Jul 2025 – Dec 2025 | +20.68% | 3.86 | 3.26% |
| 4 | Dec 2025 – May 2026 | +3.63% | 0.74 | 8.44% |
Three of four folds are positive, and fold 4 — the most recent out-of-sample window — produced a respectable 0.74 Sharpe on a 3.63% return. But fold 2 was a significant loss event, with a 16% drawdown and deeply negative Sharpe. The swing from fold 2 to fold 3 (+32 percentage points of return, Sharpe swinging from −1.70 to +3.86) signals high regime sensitivity.
The statistical validation did not pass. The Deflated Sharpe Ratio stands at 0.196, well below the threshold needed to establish that the backtest Sharpe is unlikely to be noise given the number of parameter trials (7). The Probabilistic Sharpe Ratio of 0.702 is better but still leaves meaningful doubt. Put plainly: with 7 trials and the observed fold variance, the backtest cannot yet rule out overfitting.
Strengths and Risks
What works: The thesis is grounded in well-documented market microstructure — regression channels and volume support are not exotic signals. Out-of-sample performance in fold 4 is the closest proxy to real future behavior, and that fold looks reasonable. The first live week reinforces that picture.
What to watch: The win rate below 40% creates fragility; a string of moderate losers with no outsized winner can produce extended drawdowns quickly. The high turnover amplifies both edge and friction. Most importantly, the DSR failure means the strategy needs more live trade history before its edge can be considered validated — 137 backtest trades and 7 parameter trials is a thin base for confidence.
Verdict
Channel-pullback earns a watchful eye, not a strong conviction. Its regime sensitivity is real, its statistical validation incomplete, and its risk-reward relies on a skew profile that must be demonstrated over a longer live sample. The early live data is encouraging, but one strong week does not move the needle. It remains on monitored probation — a live candidate that needs time to prove its fold-3 performance was the rule, not the exception.