Volatility Rules: Why Position Sizing Isn't a Decision
Position sizing isn't a decision — it's a consequence. How the volatility of each stock governs both where the stop goes and how many shares you can buy.
What really separates traders who survive from those who don't isn't being right more often. It's not breaking when they're wrong.
Almost everyone pours their energy into the same question: what to buy? It's the exciting part, the one that fills forums and videos. In our first article we made a concrete case: buy stocks breaking to new all-time highs. But there's a second question that decides who is still trading five years from now, and it's far less glamorous: how much can one afford to lose in a single trade? At ATH Scanner we believe this second question matters just as much as the first — and that the answer isn't the trader's to choose. It's dictated by the volatility of each stock.
Why 2%, not 1%, not 5%
The most cited rule in risk management is simple: never risk more than 2% of capital on a single trade. Alexander Elder formalized it in Come Into My Trading Room (2002). It's not a magic number or a law of physics — it's a round, robust rule — but it rests on three pillars that reinforce each other.
The first is the math of ruin. Risking a small, fixed percentage of capital on each trade keeps the probability of blowing up the account close to zero, as long as the system has positive expectancy. Pushing that percentage to 5% or 10% doesn't increase returns proportionally: it explodes the probability of a losing streak wiping out the account. It's the old gambler's ruin problem applied to markets. The Kelly criterion — the formula linking statistical edge to optimal bet size — points in the same direction: in practice, it's common to operate at a fraction of Kelly, half or less, because the drawdowns generated by full Kelly are unbearable. Risking small per trade isn't timidity: it's what the people who've been doing this for decades actually do.
The second pillar is even more relentless because it's pure arithmetic: recovery is asymmetric. A loss isn't offset by a gain of the same size.
| If you lose… | …you need to recover |
|---|---|
| −10% | +11% |
| −20% | +25% |
| −33% | +50% |
| −50% | +100% |
| −75% | +300% |
| −90% | +900% |
The curve is deceptive at first and brutal later: past 30-40% it shoots toward infinity. That's why protecting capital from large losses is what keeps you in the game long enough for the big winners to show up. Someone risking 2% per trade stays on the left side of that table, where a bad streak is still recoverable in reasonable time. Someone risking 20% gambles on crossing to the side where there's no practical way back.
The third pillar is psychological — and far from being a soft detail, it's the most rational of all. In 1979, Daniel Kahneman and Amos Tversky published prospect theory and measured something that's been replicated thousands of times since: losses weigh roughly 2.25 times more than equivalent gains. Kahneman received the Nobel Prize in Economics for this work in 2002. The operational consequence is direct: the pain of losing isn't linear, and that changes how decisions are made under pressure. Risking 2% of a $15,000 account isn't the same as risking 2% of a million. The relative figure is identical; the emotional weight isn't. A scared trader makes bad decisions — moves stops, averages down, abandons the system. The psychological component of risk isn't an externality to be ignored: it's part of the math.
The mistake of using the same stop on everything
This is where most traders trip. They internalize the 2% and apply it wrong: they set the stop 2% (or 5%, or 7%) below the entry price, the same for every stock. It's comfortable, and it's a disaster.
A quiet stock moves 1-2% a day. Another, in the middle of a euphoric run, moves 20% in a month with nothing unusual happening: that's its normal breathing. Setting a stop at 2% of the price on a stock that swings 20% monthly is guaranteeing the stop will hit before the thesis has been invalidated. It doesn't trigger because the idea was bad; it triggers because it was too tight for the asset it was meant to protect. And the reverse: a wide stop on a calm stock leaves risk on the table that didn't need to be taken.
A fixed percentage ignores the only thing that matters when placing a stop: how much that specific stock actually moves. A stop only makes sense if it's calibrated to the volatility of the asset it protects.
Volatility isn't constant — it comes in waves
There's a second, subtler problem. The volatility of a stock isn't a fixed number you check once. It changes over time, and it does so in bursts.
This isn't trader intuition: it's one of the most solid facts in quantitative finance. Back in 1963, Benoît Mandelbrot observed that in markets, large changes tend to be followed by large changes, and small ones by small ones. It's what we now call volatility clustering: turbulent periods bunch together, and so do calm ones. In 1982, Robert Engle turned that observation into a formal model — the ARCH model — that captures how volatility varies over time; Tim Bollerslev generalized it in 1986 with GARCH, the current standard. Engle received the Nobel Prize in Economics for this work in 2003.
There's a symmetry we like here. In our first article we anchored trend-following in the 2013 Nobel awarded to Fama, Hansen, and Shiller. Risk calibration rests on the 2003 Nobel to Engle. The same discipline — letting the data lead — supports both halves of a system: when to enter and how much to risk.
The practical consequence is humble: stops can only be calibrated with what has already happened and what is happening now; the future can shift regime without warning. A stop fitted to recent volatility is the best we have, but it's not a prediction. It won't protect against a sudden jump that multiplies volatility overnight. Accepting that is part of operating honestly.
Volatility sets the stop, and the stop sets the size
And here comes the idea that flips the whole perspective.
The usual mental sequence is: "I'll buy 100 shares of this company." It's backwards. The correct sequence starts from risk and ends at the share count, not the other way around.
First, fix how much loss is acceptable: 2% of capital. Next, the volatility of the stock dictates where the stop must go to avoid being hit by noise. And those two things together — the maximum risk and the distance to the stop — determine how many shares can be bought. Position sizing isn't a trader's decision: it's a mathematical consequence of how much the stock moves.
Put another way: the trader doesn't choose the position size. The market, through the volatility of each stock, dictates how many shares fit inside that 2%. A highly volatile stock demands a wider stop, and therefore a smaller position to stay within the risk limit. A calm stock allows a tighter stop and a larger position at the same risk. Size stops being an opinion about conviction and becomes the output of a subtraction.
Once inside: the stop never moves down. Ever.
Calibrating the stop well on entry is half the job. The other half is the discipline of never moving it the wrong way.
The rule is absolute: the stop loss never goes down. Moving it down "to give it a little more room" is the most common — and most expensive — way to turn a small, planned loss into a large, improvised one. The stop is set with a cool head before entering, precisely so that decisions don't have to be made with a hot head when the price moves against you.
When the price moves in favor, yes, the stop moves — but only upward. The trailing stop follows the climb, protecting part of the gain, and it's calibrated exactly like the initial stop: to the volatility of the asset, not to an arbitrary percentage. We flagged this in the first article: even inside a healthy trend, a volatility spike can hit the stop before the move resumes. It's accepted as part of the game. The goal isn't to capture the entire move — it's to capture the bulk of it, without exposing the gains to a reversal that would take them away.
Diversification has a limit
With more capital, a new lever appears: diversification. More money allows risk to be spread across more simultaneous positions, so that none of them weighs too much. It's healthy. But it has a limit, and it's worth saying so clearly, because popular intuition pushes in the wrong direction.
More stocks don't mean more safety. Past a certain point, each additional position doesn't diversify: it just adds noise, dilution, and tracking work. The great managers know this. Warren Buffett has held Apple at 40-50% of Berkshire Hathaway's public equity portfolio at its peak (2023-2024). Stanley Druckenmiller built his track record by concentrating his best ideas in a few high-conviction positions, not by spreading across hundreds. The goal isn't to own many stocks — it's to own enough not to depend on any single one, and not one more.
An honest caveat: they concentrate in businesses, with conviction and a horizon of years, in a framework different from a system with stops. But the underlying lesson translates: diversification protects up to a point, and then it gets in the way.
Where exactly that point lies, and how the individual 2% risk combines with a portfolio-level risk cap, is a topic of its own. Elder, in fact, pairs the 2% rule with a monthly 6% cap for the entire portfolio — but that belongs in the next article, where we'll develop sizing and portfolio-level risk.
From the idea to the tool
The conclusion is uncomfortable precisely because it's so unglamorous: the survival of a trading system is decided at the per-trade risk level, and that risk is only managed well when the stop is calibrated to the actual volatility of each stock. Risk management is boring. It's also the only thing that separates those who keep trading from those who got left behind.
The problem is that calibrating a stop to the volatility of each stock can't be done by eye — and certainly not stock by stock when there are thousands of candidates. That's why, at ATHScanner, every stock that passes the filter comes with its entry price and stop already calculated, adjusted to its own volatility. It's not a cosmetic detail: it's what turns the principle of this article into something executable every day, without bias and without opinion. The tool does the math; the decision of what to trade still belongs to the trader.
Sources
- Elder, A. (2002). Come Into My Trading Room. John Wiley & Sons. — The 2% rule per trade.
- Kahneman, D. & Tversky, A. (1979). Prospect Theory: An Analysis of Decision under Risk. Econometrica, 47(2), 263–291. — Nobel Prize in Economics awarded to Kahneman, 2002.
- Tversky, A. & Kahneman, D. (1992). Advances in Prospect Theory: Cumulative Representation of Uncertainty. — Loss aversion coefficient λ ≈ 2.25.
- Mandelbrot, B. (1963). The Variation of Certain Speculative Prices. Journal of Business, 36(4). — Volatility clustering.
- Engle, R. (1982). Autoregressive Conditional Heteroscedasticity (ARCH). Econometrica, 50(4). — Nobel Prize in Economics 2003.
- Bollerslev, T. (1986). Generalized Autoregressive Conditional Heteroskedasticity (GARCH). Journal of Econometrics, 31(3).
- Berkshire Hathaway — Apple stake, 13F filings (2023–2024 and beyond).