The thesis
Channel-pullback is a trend-following-meets-mean-reversion idea: within a confirmed uptrend, it buys pullbacks to the lower regression channel or volume support, then exits at the upper channel or resistance. In plain terms, it tries to buy dips in names that are still going up and sell into strength. The universe is 24 large-cap US stocks — mega-cap tech (AAPL, MSFT, NVDA, GOOGL), financials (JPM, V, MA, BAC), plus defensives and staples (JNJ, PG, KO, WMT). It is currently running live as a paper account.
Recent activity
The strategy trades on a scheduled daily run, and the last week shows it working the book actively but selectively. Between 22 and 29 September it executed roughly eight fills against nine rejections — days like 2026-09-29 and 2026-09-22 closed with zero executed and two rejected, which suggests entry conditions frequently aren't met or sizing/cash checks block the order.
The fills themselves are coherent with the thesis. On 25 September it bought NVDA at $224.16, then sold those 8 shares on 28 September at $230.06 — a clean, small pullback-to-strength round trip. It also rotated into staples and payments (KO at $87.44, MA at $567.49) while trimming MSFT and BAC. The live account total sat near $11,014 on 29 September, up modestly from ~$10,960 mid-week.
Backtest and validation
Here the picture gets more sober. Over 451 days the backtest returned +7.62% (CAGR 4.19%), with a Sharpe of 0.40, a 14.83% max drawdown, and a 39.39% win rate across 137 trades. A sub-40% win rate is fine if winners are larger than losers, but paired with a Sharpe below 0.5 it signals a thin edge. Turnover of 2,311% is very high — this strategy churns, and although modelled fees are only $137 total here, in a real book that churn is a cost drag to watch.
Most importantly, cross-validation did not pass. Three of four folds were positive, but the spread is alarming: Fold 3 returned +20.68% (Sharpe 3.86) while Fold 2 lost −11.42% (Sharpe −1.70). That single bad regime is doing real damage to the risk-adjusted profile. Out-of-sample return was +3.63% at a 0.74 Sharpe — better than the full-sample Sharpe, which is mildly reassuring, but the deflated Sharpe ratio of 0.196 (across 7 trials) is the number that fails the gate. It says that once you account for how many variants were tried, the edge is hard to distinguish from luck.
Verdict
Strengths: a sensible, explainable thesis, disciplined round-trips, and positive out-of-sample return. Risks: a weak win rate, punishing turnover, one deeply negative fold, and a deflated Sharpe well below any confidence threshold. The live run is worth continuing as an observation, but on the current evidence channel-pullback has not earned a promotion beyond paper.