The thesis
channel-pullback is a mean-reversion strategy with a trend filter. It buys pullbacks to the lower regression channel or volume support only in confirmed uptrends, then exits at the upper channel or resistance. It runs across a 24-name universe of large-cap US equities spanning tech (AAPL, MSFT, NVDA), financials (JPM, BAC, V, MA), healthcare (UNH, JNJ, PFE), and staples (PG, KO, WMT, COST). The design is intuitive: let the trend filter keep you on the right side of the market, and harvest the noise around it.
Recent activity
The strategy is live and trading on a scheduled cadence. Over the past week it executed cleanly with zero rejections — single-name entries on 2026-08-17 through 08-24, plus a busier three-trade session on 08-20. Notable fills include a BAC round-trip (buy 31 @ $64.07 on 08-13, sell 31 @ $64.52 on 08-14), a UNH position opened at $390.72, and repeated AAPL activity around the $311–$315 range.
One caveat worth flagging honestly: the account's total equity drifted down across the week, from $9,718.18 on 08-17 to $9,539.20 on 08-24. That is a soft patch, not a verdict, but it is consistent with a strategy still finding its footing in live conditions.
Backtest and validation
Over 451 days the backtest returned +7.62% (final equity $10,761.52, ~4.19% CAGR), with a Sharpe of 0.40 and a max drawdown of 14.83%. The win rate is a low 39.39% across 137 trades — acceptable for a system that presumably lets winners run past losers, but it leaves little margin for error. Turnover is very high at 2,311%, so this is an active strategy where fees and slippage matter.
The walk-forward validation is where I get cautious. It failed the gate. Three of four folds were positive, and fold 3 was excellent (+20.68%, Sharpe 3.86), but fold 2 was ugly: −11.42% at a −1.70 Sharpe with a 16% drawdown. That single bad regime dominates the risk picture. Out-of-sample results are modest but real: +3.63% at a 0.74 Sharpe.
The deflated statistics tell the story. The Probabilistic Sharpe Ratio sits at a respectable 0.702, but the Deflated Sharpe Ratio — which penalises for the 7 trials run — collapses to 0.196. In plain terms: once you account for how many variants were tried, the evidence that this edge is real, rather than lucky, is weak.
Verdict
channel-pullback is a coherent, honestly-constructed strategy with one genuinely strong regime and a plausible economic thesis. But the low win rate, sub-1 Sharpe, heavy turnover, and — above all — a failed validation with a near-zero DSR mean it has not yet earned conviction. I would keep it live on a small allocation for observation, watch whether the current live drawdown deepens, and resist scaling it until it survives more out-of-sample regimes than it has so far.