Position Sizing: The Skill That Matters More Than Picking Winners

A few years ago I put about $100 into NVIDIA in a small “YOLO” account, alongside a couple of other tech names. The position ran up roughly tenfold, and I sold half, convinced it was richly valued and probably overvalued. The half I kept went on to do another tenfold, so my original stake ended up returning more than a hundred times. It is a fun story to tell. It is also close to meaningless: the gains were a few thousand dollars, immaterial next to the rest of my portfolio.

Being right barely helped me, because the bet was never sized to matter. A $100 stake can become a cocktail-party story; only a position large enough to feel a little uncomfortable can change a portfolio. And the same size that makes a winner matter also makes a loser matter. That tension is what this guide is about: a speculative position should be small enough to survive and large enough to matter, and most people get one side of that wrong.

Disclosure: I am not a stock picker; I hold diversified, evidence-based funds (AVUV and DFSV) for serious money. This guide is educational and is not individualized investment advice.

Being right is rare, and being right at size is rarer

Start with how hard picking winners actually is. Hendrik Bessembinder studied every US common stock since 1926 and found that only 42.6% beat one-month Treasury bills over their full lifetimes, so about 57.4%, slightly more than four of every seven, underperformed cash.¹ The market’s entire net wealth creation traced to the best-performing 4% of companies, about 1,092 firms; the other 96% collectively only matched Treasury bills. Owning the giant winners matters enormously, and finding them in advance is extremely hard. That is the paradox a stock picker is up against before fees, taxes, or behavior enter the picture.

The behavioral evidence is unkind to active trading

The people who trade the most tend to do the worst. Barber and Odean tracked 66,465 households at a discount broker and found the most active fifth earned 11.4% a year while the market returned 17.9% over the same window.² In later work they showed individual investors are net buyers of attention-grabbing stocks, the ones in the news or with unusual volume, because buying means searching among thousands of candidates while selling is usually limited to the few names you already hold.³

My NVIDIA trim is a textbook case of a third bias. Odean documented the disposition effect: investors realize gains far more readily than losses, and the winners they sell go on to beat the losers they keep by about 3.4% over the following year.⁴ Selling half of a tenfold winner felt prudent, but it was a reflex, not a plan. The useful question was one I should have answered in advance: what rule will I follow if this becomes a 5x, a 10x, or a 50x position? A pre-committed rule beats an in-the-moment judgment that is mostly the disposition effect wearing a suit.

Sizing is harder than direction, even with a crystal ball

The sharpest evidence that sizing matters more than picking comes from Elm Wealth’s “Crystal Ball Challenge,” which the Economist covered in June 2026 under a fitting subtitle: more difficult than knowing what to buy is knowing how much.⁵ Elm gave 118 participants, most of them graduate finance students, $50 each and a literal look at the next day’s Wall Street Journal front page before they bet on the S&P 500 and 30-year Treasuries, with leverage up to 50x. Even with tomorrow’s news in hand, 45% lost money, one in six went bust, and the average pot finished at $51.62, a 3.2% gain.⁶ They guessed direction barely better than a coin flip and sized their bets badly.

The follow-up with AI models told the same story. Given ten runs each and an imaginary $1 million, only two of the four finished ahead: Claude at about $2.6 million and ChatGPT at about $1.5 million, with Grok and Gemini below the starting stake. Their direction accuracy topped out barely above 60%, yet they ran leverage of roughly six to twelve times, which courted a wipeout.⁷ The five expert macro traders Elm invited to the original game were the only group that clearly succeeded, finishing up an average of 130% (a median of 60%). Their direction accuracy was only modestly better at 63%. What set them apart was sizing: they sat out roughly a third of the opportunities and bet big only when confident. The edge came from how much they bet, not from a better crystal ball.

The poker lens

Good poker players do not just ask whether they probably hold the best hand. They ask about pot odds, the size of their edge, the variance of the outcome, how deep their bankroll is, and whether they can survive a bad run. The investing translation is direct: what is my expected edge, how big is the downside, how correlated is this bet with the rest of my portfolio and my career, and how much regret can I tolerate if it goes to zero or if it twentyfolds right after I trim. Forecast accuracy and valuation opinions are only inputs. Sizing is what decides whether an edge survives or ruins you.

The Kelly criterion, briefly and mathematically

The clean mathematical version of this idea is the Kelly criterion, from a 1956 paper by John Kelly that links an informational edge to the bet size that maximizes the long-run growth rate of wealth.⁸ For a bet that wins with probability $p$ (and loses with $q = 1 - p$) at net odds of $b$ to one, the growth-optimal fraction of your bankroll to wager is

For an even-money bet, where $b = 1$, this simplifies to $f^{*} = 2p - 1$. So a 60% edge on a coin-flip payoff implies betting 20% of your bankroll; a 55% chance at two-to-one odds implies about 32.5%. For a continuously held investment with excess return $\mu$ (the return above the risk-free rate) and variance $\sigma^{2}$, the analogous fraction is

Two cautions keep this honest. Full Kelly maximizes long-run growth but rides brutal drawdowns, and because real edges are estimated with error, overbetting past $f^{*}$ is punished far more than the same amount of underbetting. That is why practitioners use fractional Kelly: betting half of $f^{*}$ keeps roughly three-quarters of the growth at about half the volatility, in the Gaussian approximation.⁹ For a single stock you almost never know the true probabilities, so Kelly is not a number to compute for NVIDIA. It is a discipline: never size a bet without weighing it against your bankroll, the uncertainty in your edge, and the risk of ruin.

A practical framework: core and explore

For serious wealth, keep 95% to 100% in a diversified, low-cost, rules-based portfolio. Vanguard’s principles, clear goals, broad diversification, low cost, and discipline, remain the right default for most households.¹⁰ If you want to speculate, carve out an optional 0% to 5% “sin sleeve” and treat it as entertainment or education, sized to the kind of investor you are.

Inside that sleeve, size each position against your total portfolio, not the fun account alone. A 20% position inside a 2% sleeve is only 0.4% of your wealth, which is psychologically exciting and financially contained. A workable per-position range:

The calculator below makes the tradeoff concrete. Set a position size and your own odds, and it shows whether the upside is large enough to move your portfolio and whether the worst case is small enough to survive.

So was my NVIDIA non-bet rational?

Yes. With a large portfolio, a $100 position is pure entertainment, and even a hundredfold return does not change my net worth in any meaningful way. A position big enough to matter would now be large enough that gambling it without a durable edge is uncompensated concentration risk, which is a perfectly rational thing to decline. I did not “miss” the lesson by failing to size NVIDIA bigger. The humbler reading is this: if I lack the conviction to make a position material, I should not expect it to matter, and if I make it material, I have to accept being wrong in a way that hurts. For most people, most of the time, that math argues for skipping the bet.

Who this is for, and who it is not

A small speculative sleeve can be reasonable if you:

• Have a high savings rate and a long horizon.
• Are already diversified and will size from total wealth.
• Use written rules and no borrowed money, and can tolerate a total loss of the sleeve.
• Treat it as entertainment, not a path to retirement.

It is a poor idea if you:

• Would be tempted to size up after a win or chase losses.
• Need the gains for the plan to work.
• Borrow to amplify bets, or trade on attention and headlines.
• Cannot stomach the position going to zero.

See a speculative position in your whole portfolio

A position only means what it weighs against everything else you own. The same $5,000 bet is a rounding error in one portfolio and a real risk in another, and a “small” single-stock position can quietly overlap an index fund you already hold. Seeing the position’s true weight, its correlation with the rest of your holdings, and how it behaves in a drawdown turns “this feels fine” into a number you can check.

Frequently Asked Questions

How much should I put in a single stock?

For a pure speculative bet, a common range is 0.1% to 0.5% of your total portfolio, rising to maybe 1% for a high-conviction pick. Size from total wealth, not from your fun account. The test is whether the upside is large enough to matter and the worst case, a total loss, is small enough to survive.

What is the Kelly criterion in simple terms?

It is the bet size that maximizes long-run wealth growth given your edge. For a win probability p at net odds b to one, the fraction is p minus q over b. In practice investors bet a fraction of that (often half), because edges are uncertain and overbetting is severely punished. For a single stock you cannot really estimate the inputs, so treat Kelly as a discipline about ruin, not a precise number.

Is it ever OK to gamble on individual stocks?

For entertainment, with money you can lose, in a sleeve small enough that a total loss does not derail your plan, yes. As a strategy to build wealth, the evidence is against it: most stocks underperform Treasury bills over their lives, and the most active traders earn the lowest returns.

Why did selling half my winner feel wrong?

Because trimming a winner often is the disposition effect, the tendency to realize gains too readily while holding losers. It feels prudent, yet on average later returns do not justify it. The fix is to set a selling rule while you are calm, before the position moves, so the decision follows the rule instead of the reflex.

How big does a position need to be to matter?

Enough that even a big win changes your net worth by a percentage point or more. Below that, being right is a story rather than a financial outcome. Above that, being wrong becomes a real risk, which is exactly why a material speculative bet needs a genuine edge and a loss budget.

Key Takeaways

• Being right is not enough. A win only compounds if the position was sized to matter.
• Size beats direction. Even players with tomorrow’s news blew up through leverage and bad sizing.
• Size from total wealth. A 20% position in a 2% sleeve is 0.4% of your portfolio.
• Kelly is a discipline, not a formula for NVIDIA. Bet against your bankroll, your edge uncertainty, and the risk of ruin.
• Small enough to survive, large enough to matter. If a bet cannot be both, the rational move is usually to skip it.

Related Guides

• Is Factor Investing Dead?: evidence-based tilts versus discretionary stock picking.
• Concentration Risk: when a single position quietly dominates your outcome.
• An Index Fund Is Not a Financial Plan: the diversified core a sin sleeve sits on top of.

Sources

1. Hendrik Bessembinder, “Do Stocks Outperform Treasury Bills?” Journal of Financial Economics 129(3), 2018 (42.6% of stocks beat T-bills over their lives; the top ~4%, 1,092 firms, account for all net wealth creation). ssrn.com.
2. Brad Barber and Terrance Odean, “Trading Is Hazardous to Your Wealth,” Journal of Finance 55(2), 2000 (most-active households earned 11.4% vs 17.9% for the market). ssrn.com.
3. Brad Barber and Terrance Odean, “All That Glitters: The Effect of Attention and News on the Buying Behavior of Individual and Institutional Investors,” Review of Financial Studies 21(2), 2008. academic.oup.com.
4. Terrance Odean, “Are Investors Reluctant to Realize Their Losses?” Journal of Finance 53(5), 1998 (the disposition effect; winners sold beat losers kept by ~3.4% over the next year). wiley.com.
5. The Economist, “Why Macro Trading Is Hard” (June 23, 2026). economist.com.
6. Elm Wealth, “When a Crystal Ball Isn’t Enough to Make You Rich” (118 participants, $50 stake, next-day WSJ front page; 45% lost money, 16% went bust, average gain 3.2%). elmwealth.com.
7. Elm Wealth, “Do AIs Make Good Traders?” (Claude ~$2.6M and ChatGPT ~$1.5M finished ahead; expert traders averaged +130%, median +60%, 63% direction accuracy, by varying position size). elmwealth.com.
8. J. L. Kelly Jr., “A New Interpretation of Information Rate,” Bell System Technical Journal 35(4), 1956. princeton.edu.
9. Leonard MacLean, Edward Thorp, and William Ziemba, The Kelly Capital Growth Investment Criterion (World Scientific, 2011); and Thorp, “The Kelly Criterion in Blackjack, Sports Betting, and the Stock Market.” stat.berkeley.edu.
10. Vanguard, “Vanguard’s Principles for Investing Success.” vanguard.com. Figures are point-in-time and were verified against journal, SSRN, and primary sources.

Disclosure: the author holds diversified funds (AVUV and DFSV) for serious money and any individual-stock holdings are immaterial to the portfolio. This article is educational and is not financial advice. Speculative positions can lose all of their value; size them only with money you can afford to lose.