Mathematics
Bayesian vs. Frequentist Interpretations of Probability
Quick fact
The frequentist definition of probability was formalized by Richard von Mises in the early 20th century, but the Bayesian approach actually came first: Bayes' theorem was published in 1763.
Why this is interesting
If a meteorologist says there’s a 70% chance of rain tomorrow, do they mean it will rain 7 out of every 10 days like this? Or are they 70% confident it will rain?
Read the full explanation
Understanding Bayesian vs. Frequentist Interpretations of Probability
Say you have a coin. A frequentist says the probability of heads is the long-run proportion of heads if you flip it many times—if you flip it 1000 times and get 500 heads, that’s your evidence. For a Bayesian, probability is a measure of uncertainty or belief. Before you flip the coin, you might believe it’s fair (prior belief). After seeing some flips, you update that belief (posterior belief). This difference shows up in practice. A frequentist asks: 'If I repeat this experiment many times, how often do I get a result this extreme?' A Bayesian asks: 'Given my prior and this data, how likely is my hypothesis?' Neither is 'correct' in all cases; they are different ways to answer 'What is the probability of an event?'
A deeper explanation
The core difference lies in what probability is: a physical property (frequency) or a mental state (belief). The frequentist view works well for repeatable events like coin tosses or rolling dice, where you can imagine an infinite series of trials. But it struggles with unique events—like the chance that a specific candidate wins an election—because you can’t repeat that election many times. The Bayesian view handles unique events gracefully: probability expresses how confident you are, and you update that confidence using Bayes' theorem: P(H|D) = P(D|H) × P(H) / P(D). Here, P(H) is your prior, P(D|H) is the likelihood of the data given the hypothesis, and P(H|D) is the posterior. This formalizes how evidence changes belief. Why it matters: In medicine, a frequentist might report a p-value from a clinical trial, asking 'If the drug had no effect, how often would we see results this extreme?' A Bayesian might combine prior knowledge with trial data to estimate the probability the drug works. This leads to different interpretations of the same result and fuels debates about how to make decisions under uncertainty.