The thesis
Channel-pullback is a classic pullback-in-an-uptrend system: it buys dips toward the lower regression channel or volume support when a name is in a confirmed uptrend, then exits at the upper channel or overhead resistance. The universe is deliberately conservative — 24 mega- and large-cap names spanning tech (AAPL, MSFT, NVDA), financials (JPM, BAC, V, MA), healthcare (JNJ, UNH), staples (PG, KO, WMT, COST) and industrials (CAT, HON). This is a strategy designed to fade noise inside durable trends, not to chase breakouts.
Backtest: a small edge, honestly priced
Over 451 days the system returned 7.62% (CAGR 4.19%), turning $10,000 into $10,761. That headline hides the texture. The win rate is only 39.39% across 137 trades, so the edge comes from asymmetry — winners larger than the frequent small losers — rather than hit rate. The Sharpe of 0.40 is modest, and max drawdown of 14.83% is meaningful for a mean-reversion book. Turnover is a striking 2,311%, meaning capital is recycled roughly 23 times; at $1 per trade the $137 fee bill is trivial today, but that churn would bite harder at scale or with wider spreads.
Validation: it did not pass
The walk-forward test is where the caution flags go up. Across four folds the strategy was positive in three, but fold 2 (Jan–Jul 2025) lost 11.42% at a −1.70 Sharpe with a 16% drawdown — evidence the edge disappears, or inverts, in the wrong regime. The result leans heavily on fold 3 (Jul–Dec 2025): +20.68% at a 3.86 Sharpe. Strip that window and the record looks pedestrian. Out-of-sample return was 3.63% (Sharpe 0.74), respectable but well below the full-sample figure.
The deflated statistics tell the story cleanly: PSR of 0.702 is decent, but the deflated Sharpe ratio of 0.196 — adjusted for the 7 trials run — sits far below any reasonable confidence threshold. The validation gate returned false, and that verdict looks correct.
Recent live activity
Despite failing validation, the strategy is running live on a paper book. Recent scheduled runs have been quiet — several days with zero executions — punctuated by short round-trips: a BAC position bought at $62.98 and sold the next day at $63.12, a CAT pair, and a PG lot. Live equity has drifted from about $10,034 down to $9,794, so the live paper account is currently fractionally underwater versus its $10,000 start, with roughly $4,471 sitting in cash. The low activity is consistent with a system that only fires on qualifying pullbacks.
Verdict
Channel-pullback is a coherent, low-turnover-per-signal idea with a plausible economic rationale, and its backtest edge is real if unspectacular. But the fragility is genuine: a negative fold, a DSR near zero, and early live results below water all argue for treating this as a monitored experiment, not a funded conviction. The next fold of live data — especially any regime resembling that painful 2025 stretch — will tell us far more than the backtest already has.