Thesis: Buy the Dip Inside the Trend
Channel-pullback follows a classic technical principle: wait for a confirmed uptrend, then buy when price retraces to the lower regression channel or volume-supported floor, and exit at the upper channel or a defined resistance level. Applied to a 24-stock universe of large-cap U.S. equities spanning tech, financials, healthcare, consumer, energy, and industrials, the idea is sound — it avoids chasing breakouts and instead demands price to come to the strategy before a position is opened.
With a win rate of just 39.4%, the strategy is explicitly built around asymmetry: losing trades must be small, and winners must be large enough to carry the book. That asymmetry is the hypothesis worth stress-testing.
Recent Live Activity
The past two weeks have been selective. Between July 2 and July 6, the strategy ran daily but executed zero trades, rejecting at least one signal each session — a sign that the entry filter is working as designed and not forcing positions into weak setups.
July 1 broke the quiet with two executions: a buy of 19 shares of DIS at $96.32 and a sale of 7 WMT shares at $108.53. July 8 was active again — 7 AAPL shares were sold at $314.01 (having been purchased on June 25 at $279.39, a roughly 12% gain in under two weeks) and 6 WMT shares were re-entered at $112.99. The AAPL trade is a clean illustration of the thesis working: a pullback entry followed by a channel-top exit. The WMT re-entry at a higher price than the July 1 exit is worth monitoring to confirm the signal was driven by a fresh channel reading rather than noise.
Total paper portfolio value stands at $10,333 as of July 8, with $2,262 in undeployed cash.
Backtest Performance
Over 451 days and 137 trades, the strategy returned 7.62% in total (CAGR ~4.19%), with a Sharpe ratio of 0.40 and a maximum drawdown of 14.83%. Turnover is high at 2,311% annualized — roughly five full portfolio rotations per year — which makes fee sensitivity a real concern in live deployment beyond the flat $1-per-trade model used in backtesting.
Cross-Fold Validation: the Real Story
The four-fold walk-forward tells a more nuanced story than the headline return:
| Fold | Period | Return | Sharpe | Max DD |
|---|---|---|---|---|
| 1 | Aug 2024 – Jan 2025 | +6.53% | 1.25 | 6.16% |
| 2 | Jan 2025 – Jul 2025 | −11.42% | −1.70 | 16.09% |
| 3 | Jul 2025 – Dec 2025 | +20.68% | 3.86 | 3.26% |
| 4 | Dec 2025 – May 2026 (OOS) | +3.63% | 0.74 | 8.44% |
Three of four folds were profitable, but Fold 2 was severe — a 16% drawdown and a −1.70 Sharpe over six months. This coincides with the choppy, trend-reversing tape that characterized early-to-mid 2025, a market environment that punishes pullback buyers when trends fail to resume.
The out-of-sample fold (Fold 4) returned 3.63% with a Sharpe of 0.74 — modest but positive, and notably the OOS Sharpe exceeds the aggregate in-sample Sharpe of 0.40, a constructive sign that the strategy is not purely curve-fitted.
Formal validation has nonetheless failed: the Probabilistic Sharpe Ratio sits at 0.702 (acceptable), but the Deflated Sharpe Ratio is only 0.196 — well below the 0.5 threshold typically required to assert that performance survives multiple-testing correction across the 7 parameter trials run.
Strengths and Risks
Strengths: Clear entry/exit logic, disciplined rejection of weak signals in live operation, and an OOS fold that held up better than the in-sample average.
Risks: The strategy is highly regime-dependent — it can give back significant capital when the broader tape turns rangebound or reverses hard. High turnover compounds fee drag at scale. The failing DSR means there is insufficient statistical confidence to rule out overfitting, given the number of optimization trials.
Channel-pullback is worth watching, but not worth scaling until the DSR clears its hurdle across a broader out-of-sample window.