Investors are routinely told to size positions by probability. There are two kinds of probability in play, they are not interchangeable, and almost nobody sorts a position into the right one before committing capital.

What follows sets out the distinction, why the two get blurred, and what changes once a position is classified correctly.

Class Probability and Case Probability

The first kind of bet has enough history behind it that a real pattern can be read. A broad index fund. A diversified SIP. A category with enough years on the clock that the odds can be reasonably priced.

The second is a one-off. A small-cap riding a story nobody has stress-tested. An IPO priced on a narrative rather than a balance sheet. A sector re-rating that has never quite played out this way in this market.

Ludwig von Mises named these in Human Action: case probability, the probability of something singular and unrepeatable, against class probability, which belongs to a whole category of similar cases. Frank Knight had drawn a related line in 1921 between measurable risk and genuine uncertainty. The distinction is old, and it is not a technicality - the two require different evidence and resolve on different timescales.

The Blur Is Comfortable, Which Is Why It Persists

Investors merge the two because merging them is easier than sitting with the alternative. It is more comfortable to expect a one-off to behave like the typical case than to accept it might not behave like anything at all.

So the return expectation built for the class gets borrowed and pinned onto the individual bet, and the hope becomes that the very uniqueness of the business is what will pay off. Both halves of that cannot be true at once: a position cannot be priced off a reference class and simultaneously justified by being unlike everything in it.

The Market Does Not Correct the Error on Schedule

The mistake is not self-limiting, because the market sometimes pays anyway. One good IPO listing gain or one fortunate derivatives trade says almost nothing about skill - the sample is too small and the noise too large.

Confusing a bull-market tailwind for personal judgement is how most trading accounts eventually return what they made. And because the payoff often arrives before the mistake does, the mistake gets buried, waiting to resurface at the next redemption.

What should move confidence instead is a long, consistent record:

Only One of the Two Can Be Waited Out

A class-probability bet reveals whether it was right relatively quickly, because it is one instance among many and the law of large numbers does the work of proving it out.

A case-probability bet has no such luxury. There is no larger sample to average into. Its uncertainty resolves only by the actual sequence of events playing out, whether that takes one earnings cycle or five. Waiting is not a virtue bolted onto the decision - it is the only mechanism by which a unique bet is ever confirmed or disproved.

That is what separates a good outcome from a good decision. Not any single moment, but the willingness to wait while the uncertainty resolves - which is a rate, not a personality trait: the rate at which an investor discounts five years of compounding against the discomfort of this month’s statement.

What This Means for Allocators

Classify the position before sizing it. Whether a reference class exists is the first question, and it determines whether any historical base rate applies at all.

Do not apply class expectations to case bets. If the thesis is that this one is different, the reference class has been forfeited and the expected return cannot be borrowed back from it.

And set the holding period from the bet type. A case bet resolves on events, not on averages - which means the exit rule has to be written before the position is opened, or it will be written by the drawdown.

Back to Writing