⚠️ Personal research and trading journal — not investment advice. The author does not provide licensed advisory services.
The most debated question in any momentum trading system isn't when to buy. It's how long to hold.
My primary system uses an exit based on a trailing stop triggered by price action — when the stock closes below its 21-day moving average after an initial move. In practice, this often means holding for 30-90 days depending on how fast the stock moves. Some exits happen in 2 weeks. Some at 4 months.
A common alternative argument: just hold everything for 180 days (about six months) and let the trend run. The theory is that momentum stocks need time to play out, and actively exiting on short-term pullbacks costs you the big winners.
I tested it.
The Setup
Using the first_pullback system — buy after a stock's first pullback from a breakout, when it retests the breakout level with contracting volume — I ran two variants across a 6-window walk-forward test on US data (2020-2025):
Variant A (Standard): Exit on 21-day EMA cross after initial move, or on 5-7% loss from entry Variant B (180d hold): Hold for 180 calendar days unless stopped out by hard stop (5-7% loss); no trail exit
What the Data Showed
Variant B (180d hold) showed +60% more total profit than Variant A across the test period: +326R total versus +203R total.
That's a meaningful difference. If you held everything for six months instead of using a trailing exit, you captured significantly more of the upside.
But the walk-forward picture was more complicated:
- 4 of 7 windows: 180d hold was better
- 3 of 7 windows: Standard trailing exit was better
The 180d version doesn't uniformly win. It wins in trending years (2020, 2021, 2023) where momentum stocks extend far past typical MA21 exits. It loses or ties in choppy years (2022, early 2024) where stocks that triggered the system often reversed hard after the initial move.
Why I Didn't Change My System
I didn't adopt the 180d hold as the primary exit for three reasons:
1. The sample is regime-biased. The 2020-2025 period includes two significant bull runs (2020-2021 recovery and 2023). Momentum systems with longer hold periods naturally look better in long-trend environments. I can't know in real time which regime I'm in.
2. 4/7 windows is not convincing. If I changed a core exit rule and only 57% of windows supported it, I'd be making a configuration change on shaky ground. My walk-forward threshold for adopting a new rule is closer to 6/7.
3. The drawdown profile changes. When 180d hold loses, it loses harder. In 2022, holding momentum longs for 6 months meant riding significant pullbacks. The trailing MA21 exit limited damage. The improved returns in bull years came with larger individual losses in bear years.
What I did instead: tagged 180d hold as a soft observation — "CA soft-tag for 6-month forward observation." I'll run it as a paper candidate alongside the live system for the next 6 months and see how it performs on out-of-sample data from 2025-2026 before making any system change.
The Broader Principle
This is the "hold period trap" in backtest research: when you extend the hold period, you almost always get higher total returns because you capture more of the right-tail moves. But you also increase exposure to reversals, raise max drawdown, and reduce trade frequency (which can make the system more dependent on specific setups working out).
The right comparison isn't "total profit over the test period." It's "risk-adjusted return per window" — does the longer hold improve Sharpe consistently, or only in specific regime types?
In this case: specific regime types. Not consistently.
I'll revisit after 6 months of live observation. If the pattern holds on out-of-sample data from a market period I didn't use to form the hypothesis, I'll consider wiring it in.
Until then: the validated configuration is what I trade.
Track. Study. Wait. Strike.
Personal research and trading journal — not investment advice. The author does not provide licensed advisory services. — MOEasymmetry
Draft 2026-06-12. Source: O'Neil Round 4 walk-forward test, first_pullback system, US data 2020-2025, 6 windows (later extended to 7). +326R (180d) vs +203R (standard exit) over the full period. 4/7 windows favored 180d. Classification: soft observation pending 6-month forward tracking. System unchanged.