Position Sizing and Stop Losses: The Exits That Actually Matter

---

title: "Position Sizing and Stop Losses: The Exits That Actually Matter" description: "Entries are 20% of returns. Position sizing and exits are the other 80%. Here's the discipline Tessera bakes in: 7% max position, regime-aware trailing stops, and when not to sell." publishedAt: "2026-04-08" updatedAt: "2026-04-08" keywords: ["position sizing", "stop loss strategy", "trailing stop", "risk management", "portfolio construction"]

TL;DR. Tessera caps any single idea at 7% of portfolio notional. Diversification floor is 15 positions; the typical portfolio holds around 20. Trailing stops are regime-adjusted, ranging from 8% in CRISIS to 20% in STRONG_BULL, and they are evaluated on the daily close — never intraday. Competitive rotation respects the stop ladder, not the other way around. Everything below is the reasoning behind those numbers.

Why exits matter more than entries

Retail investors obsess over entries. What to buy, at what price, on what day. The research on portfolio attribution has been clear for decades that this is the wrong place to spend energy.

Gary Brinson's portfolio attribution work (and the many replications since) decomposed long-run return variance and found that security selection and market timing together explained roughly 10% of the variance. The dominant contributors were allocation policy, sizing, and how losses were handled. The headline number has been debated, but the direction is not: what you do after you buy dominates what you buy.

A good entry with a bad exit loses. A mediocre entry with disciplined sizing and exits survives, compounds, and eventually wins. This post is about the second half of that sentence.

The 7% max position

Tessera will not put more than 7% of portfolio notional into any single idea. The reasoning is arithmetic before it is philosophical: 1 / 15 is approximately 6.7%. If you commit to holding at least 15 names for diversification purposes, 7% is the ceiling that respects that floor.

What does 7% buy you? It buys you the ability to be completely, catastrophically wrong about one thing and survive it. If a holding goes to zero overnight — a fraud unwind, an accounting restatement that vaporizes the equity, a sudden regulatory action — you lose 7% of the portfolio. That hurts, but it does not end you. A 50% concentration in the same scenario does.

Backtests consistently show that concentrating above roughly 10% per name produces only marginal Sharpe improvement while dramatically increasing worst-case drawdown. The tails get fatter faster than the mean moves. For a retail-scale portfolio that cannot tolerate 40% drawdowns emotionally or logistically, 7% sits at a defensible sweet spot.

Why fixed percent, not volatility-scaled

There is an academically elegant alternative: size each position so that you risk, say, 1% of portfolio per trade based on the stock's ATR or realized volatility. Risk parity for single names. It is beautiful on paper.

In practice it whipsaws. When volatility jumps — which is exactly when good opportunities show up, because dislocation creates mispricing — vol-scaled sizing shrinks your positions. You end up with your smallest exposures at the moment the setup is most favorable, and your largest exposures at the moment volatility is compressed and expected returns are lowest.

Fixed percent sizing is simpler, more robust across regimes, and — importantly — matches the mental model a human investor can actually hold. You know what a 7% position looks like. You do not need to recompute ATR every morning to know whether you are appropriately sized.

Regime-adjusted trailing stops

Entries are one parameter. How much room you give a position to breathe is another. Tessera adjusts the trailing stop percentage based on the current regime:

| Regime | Trailing Stop | Reasoning | | ------------- | ----------------- | ---------------------------------------------- | | STRONG_BULL | 20% | Let winners run; breadth supports risk-taking | | BULL | 15% | Reasonable noise buffer for normal pullbacks | | NEUTRAL | 12% | Tighter; less forgiveness for drift | | BEAR | 10% | Protect capital; cut losers early | | CRISIS | 8% + hard cap | Survival mode; no tolerance for drawdown |

STRONG_BULL (20%). Breadth is wide, most stocks are advancing, and the biggest risk is getting shaken out of a winner by a normal 15% pullback. Giving winners room to compound is worth the occasional deeper giveback.

BULL (15%). The standard configuration. Wide enough to survive routine earnings reactions and sector rotations, tight enough that a broken name does not bleed for months.

NEUTRAL (12%). Market is undecided. Breadth has weakened. You no longer have the wind at your back, so you tighten. A 12% move against you is now more likely to mean something than to be noise.

BEAR (10%). Most things are going down. The asymmetry flips: holding a falling name costs you the next better entry. You cut earlier.

CRISIS (8% + hard cap). Liquidity is impaired, correlations have gone to one, and every day you are in a broken position is a day you are not in cash. The stop is tight and absolute — no trailing giveback, no discretion.

Why stops on CLOSE, not intraday

This is the subtlest rule and the one most investors get wrong. Tessera evaluates stops on the daily closing price, not intraday.

Intraday stops get hit by flash crashes, opening-auction imbalances, thin pre-market liquidity, and algorithmic sweeps that reverse within minutes. A stop triggered by a 9:30am spike that reverses by 10:15am locks you out of a position that is otherwise perfectly fine. You realized a 12% loss for a reason that had nothing to do with your thesis.

The tradeoff is real and worth being honest about: occasionally the market closes at the lows, and a position rides through a 15% intraday move before your stop fires at the close. You take a worse fill than an intraday stop would have given you on that one day. In exchange you avoid dozens of false exits over the course of a year. Net of churn and slippage, closes win.

When NOT to sell (the hardest discipline)

The hardest rule is the one that tells you to do nothing.

If the thesis still holds — the sector-relative P/E discount is still there, the quality grade is still A or B, the earnings quality is intact — and the trailing stop has not been hit, you do not sell on narrative.

Worked example. A holding reports earnings and misses consensus by 2%. The stock drops 8% in a single session. The NEUTRAL regime stop is 12%. The quality grade is unchanged. The RELATIVE P/E signal still ranks the name in the top decile of its sector. Do you sell?

No. You hold. The 8% move was inside the tolerance you already priced in when you sized the position. Selling here is selling on the feeling that comes from a red candle, not on any information that contradicts the reason you bought.

The discipline of not selling is the discipline of trusting the framework you built on a calm day to protect you on a loud one.

Common sizing mistakes

A short catalogue of errors Tessera's rules are designed to prevent:

• Concentrating in "highest-conviction" ideas. Research consistently shows that self-reported personal conviction is uncorrelated with hit rate, and if anything is mildly negatively correlated. Overconfidence is a reverse indicator more often than a forward one.
• Averaging down on a broken thesis. If the quality grade dropped from A to D, the original buy thesis no longer holds. Adding more capital is not "being contrarian" — it is doubling down on a bet you would not make fresh today.
• Pyramiding winners without re-testing the thesis. Adding to a name that is up 40% "because it is working" ignores the most important fact: at a 40%-higher price, the valuation case may be exhausted. Re-run the screen. If it still passes at the new price, fine. If not, do not add.
• Over-diversification. Past roughly 20 to 30 names, additional positions do not meaningfully reduce idiosyncratic risk — they just dilute your best ideas and turn the portfolio into a closet index fund.

Integration with rotation

A natural conflict: competitive rotation wants to reduce position X to make room for a stronger candidate. The signal-based exit logic says hold X — the thesis is intact, the stop has not been hit. Who wins?

Rotation respects the stop ladder, not the other way around. You do not rotate out of a still-valid thesis just because a marginally stronger candidate appeared. The stop ladder is the protective floor. Rotation operates above it: it swaps out the weakest incumbent that is not otherwise protected by a live, intact thesis, in favor of a meaningfully stronger candidate.

Put plainly: a position earns its seat in two ways. It can earn it by being the strongest available (rotation's test) or by being a thesis that is still working (the stop ladder's test). Lose both, and you are out. Hold either, and you stay.

Related reading

• Regime-Aware Investing
• Competitive Rotation
• Quality Screening Factors