Cognitive Biases · CB-25
Truly random data almost always contains apparent clusters and streaks — and the human mind reliably mistakes this normal randomness for a meaningful pattern.
The tendency to perceive meaningful patterns, clusters, or streaks within genuinely random data, because true randomness — counterintuitively — produces clumps and runs far more often than an evenly-spaced, 'balanced-looking' sequence would.
Studied extensively in the psychology of randomness perception, including foundational work by Thomas Gilovich, whose research (alongside the hot-hand studies) demonstrated people's poor intuitive model of what genuinely random sequences actually look like.
The Mechanism
Genuinely random dots cluster more than intuition expects
A genuinely random process reliably produces visible clumps and gaps — an evenly-spaced, non-clumpy pattern is actually what would be statistically unusual for true randomness, yet people consistently interpret the normal clumpiness of real random data as evidence of an underlying cause.
01 · TRUE RANDOMNESS LOOKS CLUMPIER THAN PEOPLE EXPECT, NOT MORE EVEN
This is a specific, well-documented failure of intuition about randomness
When asked to generate a 'random-looking' sequence by hand, people reliably produce something too evenly distributed compared to genuine randomness — and conversely, when shown a truly random sequence, they perceive its natural clustering as suspiciously non-random.
02 · IT DRIVES FALSE ALARMS IN EPIDEMIOLOGY AND PUBLIC PERCEPTION
'Disease clusters' are a recurring, high-stakes real-world case
Apparent geographic clusters of illness are frequently reported as alarming patterns requiring a causal explanation, when careful statistical analysis often shows the observed clustering falls well within the range expected from purely random distribution of cases across a population — a distinction epidemiologists have to carefully communicate.
03 · IT'S CLOSELY RELATED TO, BUT DISTINCT FROM, THE TEXAS SHARPSHOOTER FALLACY
Perceiving the cluster vs. drawing the target around it afterward
The clustering illusion concerns misperceiving random clumping as meaningful; the related Texas sharpshooter fallacy specifically concerns retroactively defining a pattern's boundary after seeing where the data happened to fall — related failure modes, but the clustering illusion is the more basic perceptual error underlying both.
Where It Fails / Inversion
Where it fails / inversion
Some apparent clusters are genuinely meaningful and worth investigating — the corrective isn't to dismiss all clustering as noise, but to apply an actual statistical test for whether the observed clustering exceeds what pure chance would produce, rather than relying on visual impression alone in either direction.
How To Use It
Worked example · evaluating an apparent sales or performance 'hot streak' by region
Before concluding that one sales region's recent outperformance reflects something real (a better strategy, a stronger market), calculate whether that level of variation across regions falls within the range expected from pure random noise given the number of regions and time periods being compared — with enough regions and enough time periods, some region will show an impressive-looking streak purely by chance.
How to use it
Before treating an observed cluster or streak as evidence of a real underlying cause, ask whether that level of clustering is actually unusual for a genuinely random process of this size, or whether it falls within the range chance alone would produce — genuine randomness clusters far more than intuition expects.
See Also