Position sizing:
Position Sizing and the Art of the Nibble: How Much Is Too Much in One Idea
"Diversification is a protection against ignorance. It makes very little sense if you know what you are doing."
— Warren Buffett, 1993 Berkshire Hathaway Shareholder Letter
Every other lesson in this series has been about being right — finding the quality compounder, sizing up the megatrend, spotting the moat that's actually widening. This one is about what happens after you're right, or wrong, and why the single number that determines the outcome is one almost nobody spends enough time on: how big the position was in the first place.
Position sizing is where risk management stops being theory and becomes arithmetic. A brilliant analysis behind a 2% position and a mediocre analysis behind a 25% position can produce the same portfolio-level outcome — for entirely different reasons, and only one of those outcomes was actually earned. This lesson builds the sizing discipline: what the research says about how many positions are enough, why professional bettors size below what looks mathematically optimal, and the case for concentrating hard on the rare idea worth it.
The Question Every Position-Sizing Rule Is Trying to Answer
There are two well-argued, opposite instincts about how much to put in one idea, and both come from very good investors. The diversification instinct says: you cannot know in advance which of your analyses is wrong, so spread the bets and let the good ones outweigh the bad ones over time. The concentration instinct, associated most with Buffett and Charlie Munger, says the opposite: if you have genuinely done the work and found a rare, well-understood opportunity, diluting it with your twentieth-best idea just to look diversified is a tax on your best thinking, not a form of safety.
Both instincts are correct, applied to different situations — the mistake is picking one as a permanent identity rather than a rule matched to the specific case. What follows is not an argument for one extreme; it's the research and the frameworks that tell you where, for a given idea and a given level of genuine conviction, the sensible range actually sits.
What the Diversification Research Actually Shows
The academic starting point is Evans and Archer's 1968 study, which found that the reduction in a portfolio's unsystematic (business-specific) risk tapered off sharply by around 10 to 15 stocks — adding an eleventh or sixteenth holding barely moved the needle further. Later research, using different time periods and methodologies, has generally revised that threshold upward, placing it closer to 20 to 40 holdings depending on how the portfolio is constructed and which risk measure is used. The consistent finding across nearly sixty years of revisiting this question is the shape of the curve, not the exact number: diversifiable risk falls very fast over the first ten to twenty names, and then flattens hard — the fortieth stock in a portfolio is doing almost nothing that the twentieth wasn't already doing.
The Kelly Criterion: A Ceiling, Not a Target
The Kelly Criterion is a formula, developed originally for gambling and information theory, for the fraction of a bankroll to wager on a favorable bet in order to maximize long-run growth. For a simple win/lose bet it is f* = (bp − q) / b, where p is the probability of winning, q is the probability of losing, and b is the net odds received. Run the numbers on a bet with a 60% win rate, an average win of $1,500 and an average loss of $1,000, and the formula says to risk about a third of the bankroll on that single bet.
Almost no serious practitioner — in poker, sports betting, or investing — actually bets full Kelly, and the reason matters more than the formula itself. Kelly assumes you know the true probability and payoff with certainty. In investing you never do; every input into that formula is an estimate, and an overconfident estimate turns "optimal" sizing into a way to blow up a portfolio on a single bad read. This is why professionals size at a fraction of Kelly — commonly half or a quarter — and the trade-off is favorable precisely because the relationship is not linear.
The practical use of Kelly for an equity investor is not to compute f* to two decimal places — the required inputs (a precise win probability, a precise payoff ratio) simply are not knowable for a stock the way they are for a coin flip. Its use is as a mental ceiling: it formalizes the idea that position size should scale with the strength and reliability of your edge, that sizing at your maximum theoretical conviction is never the right target, and that the cost of overbetting is asymmetric — it grows much faster than the benefit of getting slightly closer to "optimal."
Concentration With Conviction: The Munger and Buffett Case
At a 1994 USC commencement address, Charlie Munger proposed a thought experiment: imagine you were handed a punch card with only twenty slots for your entire lifetime of investment decisions, and once all twenty were punched, you could never invest again. Munger's point was not that twenty is a magic number — it was that almost nobody actually invests as though their decisions were scarce, and that scarcity is exactly the discipline that should be governing every position. "To me, it's obvious that the winner has to bet very selectively," he said. Buffett made the same point differently at the University of Florida in 1998: "very few people have gotten rich on their seventh best idea."
The case for concentration is not a case against risk management — it's a case that spreading capital thinly across mediocre ideas to "be safe" is itself a risk, just a quieter one: the risk of diluting genuine, well-researched conviction down to a market-like return while still carrying all the effort of active research. The honest resolution between this and the diversification research above is not a contradiction. It is that the diversification research describes a portfolio built from ideas of roughly similar, moderate conviction — for that portfolio, 15 to 30 names captures nearly all the available risk reduction. Concentration becomes rational specifically when conviction is not uniform — when one or two ideas are genuinely, demonstrably better understood than the rest of the list, and diluting them equally with idea seventeen is giving up expected return for a diversification benefit that, per the research above, barely exists past that point anyway.
Two Playbooks: Adding on Weakness vs. Adding on Strength
"Add on confirmation" means different things depending on what kind of edge is being traded, and conflating the two is a common sizing mistake. A quality-compounder thesis is a business-fundamentals bet — confirmation there usually means the business is executing as expected, and a lower price without a change in the fundamentals is more attractive, not less, which is why Muffett's own approach treats a decline in a name already owned as an invitation to add. That is fundamentally different from the pyramid-buying approach central to William O'Neil's CANSLIM system (covered in the earlier Stage Analysis lesson), which is a momentum and technical-pattern bet — there, confirmation means the stock is proving the pattern correct by moving up from a defined buy point, and O'Neil's discipline adds in tranches on strength while cutting any position that falls 7–8% below the purchase price, no exceptions and no story.
| Sizing Question | Quality-Compounder / Value Approach | Momentum / CANSLIM Approach |
|---|---|---|
| What counts as "confirmation"? | Business fundamentals executing as expected, often at a lower price | Price breaking out and moving up from a defined buy point |
| When do you add? | On weakness, if the thesis is intact | On strength, in tranches, typically within a few percent of the prior buy |
| What triggers a cut? | A change in the business thesis itself, not the price | A fixed, mechanical stop — commonly 7–8% below cost, regardless of story |
| What is the time horizon? | Years — the position is sized to be held through volatility | Weeks to months — the position is sized to be exited quickly if wrong |
Neither approach is more "correct" as a sizing philosophy — they are matched to different kinds of evidence and different holding periods. The mistake to avoid is borrowing the emotional comfort of one system to justify the other: using "the thesis is still intact" (a value-investing standard) to avoid taking a mechanical stop-loss on what was actually a momentum trade, or panic-selling a long-term compounder because the price did something a CANSLIM trader would have treated as a sell signal. Decide which playbook a position belongs to before buying it, and apply that playbook's add-and-cut rules, not whichever one is more comfortable in the moment.
A Practical Framework: Sizing by Conviction Tier
A simple, three-tier structure translates all of the above into a rule that can actually be followed under pressure, rather than re-litigated every time a position moves:
| Tier | Starter Size | Cap at Cost | What Qualifies |
|---|---|---|---|
| Core conviction | 2–3% | 10–15% | Deep research, a clear moat, management incentives checked, valuation disciplined — a genuine "punch card" idea |
| Standard position | 1–2% | 5–8% | Solid thesis, meets the checklist from the quality-compounders lesson, but not yet at the highest conviction tier |
| Starter / speculative | 0.5–1% | 2–3% | Interesting but unproven — a name being researched in real time, sized to survive being wrong |
The cap at cost matters more than the starter size: it is the number that actually protects the portfolio, because it defines the maximum damage a single broken thesis can do regardless of how much conviction felt justified in the moment of buying more. A position that has grown past its cap through appreciation — a genuine win, not a sizing mistake — is a different problem addressed below, not a reason to loosen the cap on the next new idea.
Rebalancing: The Good Problem That Still Needs a Rule
The prior lesson on quality compounders made the case that doing nothing is very often the highest-value action once a good business is owned — and that is still true here. But a winning position that triples in value has, without a single new decision being made, silently become three times as concentrated a bet as it was on the day it was bought. That is not a reason to reflexively trim a compounder — cutting winners early is exactly the disposition-effect mistake covered in the behavioral-risk lesson — but it is a reason to have decided, in advance, at what size a position's continued growth becomes a concentration risk large enough to warrant a partial trim regardless of how good the business still is. A pre-committed rule (for example, trimming back to the tier cap only when a position exceeds it by some meaningful margin, rather than at the first sign of outperformance) captures the difference between prudent risk control and the same loss-aversion instinct that causes investors to sell winners too early.
Try It Yourself: The Concentration Cost Simulator
This is pure arithmetic, not a forecast: if one position in the portfolio were to fail completely, how much of the total portfolio would that single failure take with it, at different levels of concentration? Enter the number of equal-weighted positions in two hypothetical portfolios and an assumed loss on the one that fails, to see how directly the number of positions determines the damage a single mistake can do.
Building a Position-Sizing Discipline
- Decide which playbook a position belongs to — fundamental/value or momentum/technical — before buying, and apply that playbook's add-and-cut rules only.
- Write a starter size and a cap-at-cost for every position before the first purchase, using a conviction tier rather than a gut feel in the moment.
- Treat the cap at cost as the number that actually protects the portfolio — it bounds the damage a single broken thesis can do, regardless of how convinced you feel later.
- Separate "adding on confirmation" (thesis validated, price now more attractive) from "averaging down on hope" (thesis broken, price just lower) — they look identical on a brokerage statement and are opposite decisions.
- Decide a rebalancing trigger for winners that have grown past their tier cap, in advance, so a genuine success doesn't quietly become an unmanaged concentration risk.
- Remember that Kelly-style sizing logic argues for betting below your calculated conviction, not up to it — the cost of overbetting is asymmetric to the benefit of precision.
- Reserve your largest sizes for the rare idea that would make Munger's twenty-slot punch card — most ideas are not that, and sizing them as if they were dilutes the ones that are.
- Write down, before buying, the maximum drawdown in this one name you could tolerate without changing your behavior elsewhere in the portfolio — if you can't state that number, the position is not yet sized.
Sizing Is the Discipline That Makes Every Other Judgment Matter
Every lesson in this series before this one has been about being right — recognizing a quality compounder, understanding what actually compounds, protecting against inflation, recognizing your own worst behavioral instincts. Sizing is the multiplier applied to all of that judgment, and it is a multiplier that can turn a correct analysis into a rounding error or an incorrect one into a portfolio-ending mistake. The research on diversification, the logic of Kelly, and the case for concentrated conviction are not competing schools of thought to pick a side of — they are three different ways of arriving at the same conclusion: decide, in writing, before you are emotionally invested in the outcome, how big a single idea is allowed to become.
