The thesis
channel-pullback is a buy-the-dip-in-an-uptrend strategy: it waits for a confirmed uptrend, then buys pullbacks to the lower regression channel or volume support, exiting at the upper channel or resistance. It is a hybrid by design — mean-reversion entries riding a trend-following backbone — and runs across a 24-name large-cap universe spanning tech, financials, healthcare, staples, and energy (AAPL, MSFT, NVDA, JPM, V, JNJ, PG, XOM, and others). The strategy is currently live.
Recent activity
The scheduled runs have been quiet and orderly. Over the last six sessions the book executed a handful of trades — 1 on 2026-10-27, 2 on 10-22, 1 on 10-21 — with zero rejected orders throughout, suggesting the signal and sizing logic are agreeing with available cash. Total paper equity sits at $11,082.66 as of the 10-27 run, up gently from $11,010.59 a week earlier.
The trade tape tells a consistent story of small, fast rotations. NVDA was bought at $220.45 (10-14) and sold at $220.92 (10-21) — essentially flat — then re-bought at $211.74. AAPL was bought at $346.80 (10-22) and sold at $347.14 (10-27), another near-scratch. PFE round-tripped from $27.83 to $28.17. These are thin edges, consistent with a strategy that takes many shots and relies on the occasional larger winner.
Backtest and validation
The full backtest returned 7.62% (final equity $10,761.52, CAGR 4.19%) over 451 days across 137 trades, with fees of $137 and no FX cost. The risk profile is unremarkable: Sharpe 0.40, max drawdown 14.83%, and a 39.39% win rate. That low hit rate is not itself damning — it is the expected signature of asymmetric payoffs — but the mediocre Sharpe says the winners are not large enough to make the ride comfortable. Turnover is a striking 2,311%, so this strategy lives or dies on execution costs.
Walk-forward validation is where caution sharpens. Three of four folds were positive, but the dispersion is enormous: fold 3 returned +20.68% (Sharpe 3.86) while fold 2 lost -11.42% (Sharpe -1.70, 16% drawdown). Out-of-sample return averaged just 3.63% at a 0.74 Sharpe. Critically, validation did not pass: with 7 trials, the probabilistic Sharpe ratio is 0.702 but the deflated Sharpe ratio is only 0.196 — well short of what we require to trust the edge survives the search that found it.
Verdict
channel-pullback is behaving exactly as a well-governed live experiment should: trading cleanly, bleeding nothing to rejections, grinding out a small positive line. But it is regime-dependent — that -11% fold is a real risk, not noise — and the deflated Sharpe tells us the historical edge is not yet distinguishable from luck. Keep it live as paper, keep the position sizes modest, and do not promote it on a backtest it hasn't earned.