Momentum Exit & Position-Sizing System
A complete, mechanical, evidence-aligned system for running a high-CAGR long-only momentum book — built from four deep-research sweeps (academic exit theory, extension risk, US Investing Championship exit rules, and championship concentration). The companion note Momentum Exits — Research Companion holds the citations and findings behind every rule here.
The one idea everything rests on
Extension never makes you sell. It makes your leash shorter.
You never fade genuine strength (the data says that caps winners, which is fatal to momentum’s fat-tail math). As a stock gets extended, you tighten the trailing stop so a real reversal exits you near the top. Exits are always a break, never a distance. That single principle reconciles “let winners run” with “protect the gain” — and it’s what the academic literature and the championship traders both converge on.
The system (the rulebook)
1. Entry & initial risk — risk ~1%, always
- Buy only at a low-risk spot (pullback to / breakout from a base). Never chase an extended entry.
- Initial hard stop = ~1.5× ATR or 7–8%, whichever is closer. This is the one universal, no-exception rule across O’Neil, Ryan, and Minervini.
- Size so that hitting that stop = ~1% of capital (a 5% starter on a 7–8% stop ≈ 0.35% portfolio risk — David Ryan’s exact figure).
2. Build the position — never open at full size (progressive exposure)
- Enter at a pilot ~5% (½ of target).
- Add the second tranche only as it works — on a fresh continuation move, not just any +5%. Each add is smaller (1R → ½R → ¼R) and financed by the existing paper profit (Minervini “free-rolling”).
- After the add, raise the stop to ~breakeven (blended cost) so the position is free.
- Press when winning, cut when losing — largest exposure in the easiest tape, smallest in the hardest.
- Hard cap: no single name > ~15–20% of the book regardless of how tight the stop is (caps gap damage).
3. Get to “free” fast (the cushion)
- At ~1.5R profit, move the stop to breakeven. Open risk → zero.
- Never let a 20%+ gain turn into a loss (Minervini’s “protect the line”).
- This is what makes size safe: by the time the position is large, it sits on house money, not your cost — so the inevitable earnings gap eats profit, not principal.
4. The exit engine — MA-stack state machine (two-speed trail)
The 10/20 EMA relationship tells you which regime you’re in; the 50 DMA is the actual exit line. (This is Kell’s 10/20 stack + the O’Neil-disciple “Seven-Week Rule.”)
| State | Meaning | Action |
|---|---|---|
| 10 > 20 > 50, all rising | Strong uptrend, stock obeying its fast MAs | Hold/aggressive. Trail the 10 DMA. Only trim on a vertical climax. |
| 10 < 20, but price > 50 | Pullback / basing — a warning, not a break | Caution. Tighten or trim a partial; the 50 DMA is still the line. Most healthy pullbacks live here. |
| Price closes below 50 DMA | Intermediate trend broken | Full exit. |
- Exit only on a confirmed violation: close below the line, then a next-day move below that day’s intraday low. (Two-step confirmation kills the single-stab whipsaw.)
- The operative stop is always the higher of {hard stop, the relevant MA} — the hard stop governs early (before the 50 DMA has risen above your cost), the MA trail governs once established.
5. The only case you trim into strength
- A vertical, news-less climax — O’Neil: 25–50% surge in 2–3 weeks after a long advance; Kell: exhaustion far above the 10-day MA. Sell ~half into the spike, ride the rest on the trail.
- A stock making orderly new highs is not this. Leave it alone — the data says fading genuine multi-month strength is counterproductive.
6. Regime switch — when to play big
- Index above its 50 DMA → full system, full size, wider book (toward 10–12 names).
- Index below → half size or cash, tighter book (toward 4–6 names).
- Concentration is a dynamic risk lever, not a constant — wide in good tape, tight in bad (Kell: 6–7 core in good environments, 1–2 in poor).
7. Portfolio shape
- 8–12 names in good tape; fewer in weak tape. This is Minervini’s large-account band — the most diversified any champion runs, and more concentrated than a 20-stock book.
- Full positions land around ~25% of equity, reached via progressive exposure, never opened at full size.
- The build-in rule + 15–20% cap size the book naturally; don’t force a number.
The trim-vs-hold decision (time-value math)
When a winner is extended, “trim now vs. hold through a dip to the 50 DMA and recover” is a velocity-of-capital problem. Measure four numbers from your own trade history, bucketed by extension severity:
- d = % giveback from here to the 50 DMA (observable now)
- p = probability it dips to the 50 DMA and recovers to a new high (vs. breaks)
- u = conditional further upside if it recovers
- T = round-trip weeks (extension → 50 DMA → new high)
Expected gain from holding ≈ p·u − (1−p)·d, as a rate g_hold = (p·u − (1−p)·d) / T.
Hold only if g_hold > g_book (what redeployed capital earns elsewhere). The clean threshold — the recovery probability at which you’re indifferent:
$$p^* = \frac{d + g_{book}\cdot T}{u + d}$$
- p > p* → hold · p < p* → book · p ≈ p* → trim half (you’re near indifference; hedge).
The math reproduces the extension rule: the more extended the stock, the larger d, which raises p* — making holding harder to justify, so you trim more. Mildly extended → small d → low p* → hold. The extension level isn’t an arbitrary trigger; it’s the d term moving your breakeven probability.
Why this balances gains and risk
| Risk it kills | The rule that kills it |
|---|---|
| Big single-trade loss | 1% risk + tight 7–8% initial stop |
| Gap through the stop | Build-in + breakeven stop + 15–20% cap → big size sits on house money |
| Capping winners | Break-based two-speed trail, never a distance-sell |
| Giving back a big gain | Extension shortens the leash → exit near the top |
| Whipsaw / churn | Close-based, two-step-confirmed exits; loose leash early |
| Regime blow-ups | Index-50DMA switch to cash + dynamic concentration |
In one line: small fixed risk in → earn your size by being right → let strength run on a leash that tightens as it extends → leave only on a confirmed break → run wider in good tape, tighter in bad.
What to backtest / tune on your own universe
- The extension trigger — at what level (e.g., % above the 50 DMA, or N ATRs above the 20 DMA) does the leash switch from 50→20 DMA / does the climax-trim fire?
- The p, u, d, T table by extension bucket → derive your own p* thresholds (turns trim/hold/book into a lookup table).
- 20×5% vs 10–12×~10% on identical entries/exits — compare CAGR, max drawdown, and median annual return (not just mean), since momentum is right-skewed.
Hard caveats (don’t forget these)
- Most academic Sharpe/return figures are gross of transaction costs; net-of-cost evidence is the literature’s biggest gap. Watch turnover and re-entry.
- The championship headline returns (+334%, +941%, ~29,000%) used ~2:1 margin in hot tape over short audited windows — not repeatable unlevered process. Zanger’s +29,000% is inseparable from the −75% that followed.
- O’Neil’s oft-cited “sell at 20–25% profit” rule failed verification — don’t build around it. The defensive rules (7–8% stop, sell into climax, 50-day line) are the solid ones.
- This is a long-only plan. It deliberately drops the short-side machinery (the source of academic “momentum crashes”), which doesn’t apply to you.