The thesis
Channel-pullback is a hand-written, Reviewer-approved (risk 0) mean-reversion rulebook. It hunts a specific setup across 24 US large caps — AAPL, MSFT, NVDA, the big financials, staples and energy: buy pullbacks to the lower regression channel or volume support, but only inside a confirmed uptrend, then sell into the upper channel or resistance. In plain terms, it tries to buy dips that the broader trend should rescue, and it refuses to catch falling knives in downtrends. That trend filter is the whole risk-management story here, and the data suggests it is doing real work — just not reliably enough.
Recent activity
The strategy was reset to $10,000 at the Season 2 reset on 2026-10-29. Its one burst of activity came on 2026-10-30, when it executed two buys: 3 shares of MSFT at $522.07 and 5 of AAPL at $344.84, leaving $6,674.22 in cash. Every scheduled run since — 11-02 and 11-03 — has executed nothing: no qualifying pullback appeared, which is exactly what a selective setup should do in a market that isn't offering its entry. Total equity has drifted from $10,000 to $9,889.68, a small open loss of about 1.1% on those two positions. Too early to read anything into it.
Backtest and validation
Over 451 days the full-sample backtest returned 7.62% (CAGR 4.19%, final equity $10,761.52) on a Sharpe of 0.40, with a 14.83% max drawdown and a 39.39% win rate across 137 trades. A sub-40% win rate paired with positive returns means the winners are bigger than the losers — a classic cut-losses-let-winners profile — but turnover of 2,311% is heavy, and the trend eats fees ($137 in all, FX zero).
The walk-forward picture is where the character shows. Three of four folds were positive, but the spread is enormous: Fold 1 +6.53% (Sharpe 1.25), Fold 3 a spectacular +20.68% (Sharpe 3.86), Fold 4 +3.63% — and Fold 2 a brutal −11.42% with a −1.7 Sharpe and 16% drawdown. That single bad regime, likely a trend-break where "pullbacks" kept failing, is the honest risk: when the uptrend assumption breaks, so does the strategy.
Crucially, validation did not pass. Out-of-sample return was 3.63% on a 0.74 Sharpe — respectable — and the probabilistic Sharpe (PSR 0.702) looks encouraging. But the deflated Sharpe ratio, which penalises for the 7 trials run, collapses to 0.196. That is the gate it fails: once you account for how many configurations were tried, the edge is not statistically distinguishable from luck.
Verdict
Channel-pullback has a coherent, disciplined thesis and a backtest that is green rather than red. But a 0.196 DSR and a −11% fold say the same thing: the apparent edge is fragile and regime-dependent. It earns its "live" slot as a watch-and-learn candidate, not a conviction allocation — and its current patience (three straight no-trade runs) is a feature, not a fault.