Interpretations of Probability

Probability is the mathematics of uncertainty: a precise language for reasoning about events that have not happened yet. Before building the machinery, it is worth being exact about what the number itself means, because the same value can be read in three different ways, and each reading is the right tool in a different situation.

Where the subject comes from

The theory began in the 1650s with a correspondence between Blaise Pascal and Pierre de Fermat about how to divide the stakes of an interrupted game of chance. From that exchange grew the first systematic rules for computing with uncertainty, and for two centuries the subject developed as the mathematics of games: dice, cards, and lotteries, where outcomes are symmetric and can be counted. The modern subject is far broader, but games of chance remain its cleanest laboratory, which is why dice and cards still dominate probability questions today.

The classical interpretation

The classical view powers most interview questions: a die has six equally likely faces, so the probability of rolling a five is . No data is required, only counting, which is why the next module of this course is devoted entirely to counting well.

The definition has a built-in limitation worth noticing. It only applies when the outcomes are equally likely, and "equally likely" is itself a probabilistic idea, so the definition is circular as a foundation. In practice this never causes trouble: symmetry tells you when outcomes are interchangeable. A die is symmetric under relabelling its faces, so no face can be more likely than another. Whenever you invoke the classical definition, the symmetry argument is what justifies it.

The frequentist interpretation

The frequentist view is how probabilities are measured from data. If a coin of unknown fairness lands heads on 5,300 of 10,000 flips, the frequency is an estimate of its heads probability. The word estimate matters: in a short run the observed frequency can sit far from the true value, and quantifying how far is the central problem of statistics. The connection between long-run frequency and true probability is made precise by the law of large numbers, which appears in Probability II.

The subjective interpretation

Some events cannot be repeated and have no symmetry to exploit: an election, a product launch, tomorrow's weather. A probability attached to such an event is a calibrated degree of belief. The calibration standard is betting: to say an event has probability is to say you would regard 3-to-1 odds against it as exactly fair. This reading is the foundation of Bayesian inference, where beliefs are revised by Bayes' theorem as new evidence arrives; the mechanics arrive later in this course.

It is useful to be fluent in converting between probabilities and odds. Odds of to against an event correspond to probability ; probability corresponds to odds to against. So "4 to 1 against" means probability , and probability means odds of 2 to 3 against, or equivalently 3 to 2 in favour.

Why the interpretations coexist

The three readings are not rivals; they are different routes to a number that then obeys one shared set of rules. You count when outcomes are symmetric, you measure frequencies when you have repeatable data, and you calibrate beliefs when you have neither. What makes probability a single subject is that the axioms of the next lessons apply identically to all three, so every theorem proved once is available everywhere.