Why Do Random Sequences Produce Streaks That Look Meaningful?

If a fair coin produces four heads in a row during twenty tosses, many people begin to suspect the coin. Our intuitive picture of randomness often looks neatly mixed, with heads and tails alternating. A genuinely independent random process has no obligation to make every short segment look balanced. With enough trials, runs are not merely possible; they are inevitable.

Research on judgements of randomness finds that when people invent random sequences, they usually include more alternation than a random process would. When judging sequences supplied by someone else, they may also rate excessive alternation as more random. One reason is that we shrink a long-run property into a local demand: because heads and tails approach equal frequencies over many tosses, we expect each short stretch to display that balance immediately. Repeated outcomes are also easy to compress into a description such as “four heads in a row”, which makes them feel structured; a less memorable mixture can feel more accidental.

This tendency is sometimes called the clustering illusion: a run or cluster is treated as evidence of a stable mechanism. Yet a streak alone proves neither that a pattern exists nor that it does not. Sporting performance, market prices and disease cases may contain genuine dependence, changing conditions or shared causes. They should not automatically be modelled as coin tosses.

A better judgement neither forces results to look even nor tells a story whenever a run appears. It starts with a comparison model. If the events were independent, how often would a streak of this length occur? Does the observed series cluster systematically more than the model predicts? Sample size, selective attention and after-the-fact selection also matter. Randomness can draw patterns that look intentional; distinguishing a picture from a process requires comparison, not appearance alone.

https://srri.umass.edu/publications/falk-1997msr/
https://pmc.ncbi.nlm.nih.gov/articles/PMC8500893/