Thesis
Channel-pullback is a trend-following mean-reversion hybrid: it buys pullbacks to the lower regression channel or volume support within confirmed uptrends, then exits at the upper channel or resistance. In plain terms, it wants to buy dips in things that are already going up — a sensible framing that avoids the classic trap of catching falling knives in downtrends. The tradeable universe is 24 large-cap US names spanning tech, financials, healthcare, and staples.
Recent activity
The live book has been quietly busy. Over the past week the strategy logged one execution per scheduled run (with a flat, zero-trade day on 2026-08-28), concentrating heavily in AAPL. The recent tape reads as a series of tight round-trips: buy AAPL at 308.56, sell at 314.66; buy at 315.28, sell at 315.09. A UNH position (bought at 390.72) and a JPM entry (352.955) round out the exposure, alongside a CAT exit at 837.16.
The uncomfortable detail: total account value sits around $9,503 — below the $10,000 starting stake. Live paper trading is currently underwater, even though the backtest ended at $10,761.52. That gap between simulation and live is the number to watch.
Backtest and validation
On paper the strategy returned 7.62% over 451 days (CAGR ~4.19%), with a Sharpe of 0.40 and a max drawdown of 14.83%. The win rate is a low 39.39% across 137 trades — but combined with positive total return, that tells a healthy story: winners are meaningfully larger than losers, exactly the asymmetric payoff a pullback-to-resistance design should produce. Turnover, however, is extreme at 2,311%, and fees totalled $137 — friction that matters at this thin an edge.
The validation verdict is the headline caveat: it did not pass. Walk-forward across four folds was wildly inconsistent — fold 3 delivered a stellar +20.68% (Sharpe 3.86), but fold 2 lost 11.42% (Sharpe −1.70). Three of four folds were positive, and out-of-sample return held at a modest +3.63% (OOS Sharpe 0.74). The statistics temper the enthusiasm: PSR of 0.70 is respectable, but the deflated Sharpe ratio of just 0.196 — adjusted for 7 trials — signals real overfitting risk. The edge may be partly a product of the search, not the market.
Verdict
Channel-pullback has a coherent, defensible thesis and the right payoff shape: low hit rate, positive expectancy. But the evidence for a durable edge is weak. The failed validation, the fold-2 blowup, the punishing turnover, and a live book currently below water all argue for patience over scaling. This is a strategy to keep on a short leash — worth its live allocation as an experiment, not yet worth conviction. The next few folds of live data should settle whether fold 3 was signal or luck.