Random does not mean evenly mixed
Ask someone to invent a “random” series of coin flips and they will usually produce something that alternates a lot and rarely repeats more than twice. Real random flips look different. They contain runs of the same result, clusters of one outcome and quiet stretches of another. Those features are not a sign of a fault. They are what independence looks like.
Why people expect too few streaks
People tend to judge a sequence by whether it “looks” balanced. A pattern like H T H H T H T T feels random. A pattern like H H H H H H feels suspicious. Yet any exact sequence of 6 fair flips has the same probability, 1/64. The six-in-a-row one only feels special because we notice the pattern.
The second reason is that a streak of a given length has many chances to start. A run of five heads has a 1 in 32 chance of starting on any particular flip, but in a long sequence there are many possible starting points, so the chance of it appearing somewhere is much larger.
A checkable example: three heads in ten flips
How likely is at least one run of three or more heads somewhere in 10 fair flips? Counting is manageable. The number of 10-flip sequences with no run of three heads in a row can be built up step by step, and it comes to 504 out of 1,024. So the chance of at least one run of three heads is (1,024 - 504) / 1,024 = 520 / 1,024, about 51%.
Just ten flips, and it is roughly a coin flip whether a run of three or more heads shows up. Most people would guess far lower.
Longer sequences, longer streaks
As the sequence grows, the longest run grows, but slowly. A rough guide for fair coin flips is that the longest run is typically in the region of the base-2 logarithm of the number of flips.
- Around 100 flips: a longest run of about six or seven is typical, and runs of five or more are very common.
- Around 1,000 flips: a longest run of about nine or ten is typical.
These are approximate and individual sequences vary, but they carry the message. A run of eight heads in a thousand flips is unremarkable. It would be more surprising if the longest run were only three.
The same idea explains why coin toss streaks of heads appear in any long record, and why a series of reds in a roulette log, or the same face repeating on a die, does not mean something unusual is happening.
Streaks versus patterns we imagine
Streaks and clusters also fool us in other settings.
- Clusters in time or space, such as several unusual events close together, often arise by chance. Lottery coincidences are a good example, as are many coincidence myths.
- In sport and games, a player who makes several shots in a row may be described as “hot”. Streaky sequences also appear in purely random data, which is why the hot hand idea needs careful statistical testing before being believed.
- Even a good sample can look streaky when you choose where to start counting, a selection effect.
A related effect is that random sequences, despite their runs, still average out in proportion terms over time, in line with the law of large numbers.
What a streak does not tell you
A streak in a fair, independent process gives no information about the next outcome.
- After five heads, the chance of heads on the next flip is still 50% for a fair coin.
- A run of losses does not mean a win is “due”, and a run of wins does not mean luck has run out. Believing otherwise is the gambler’s fallacy.
- A long streak can raise a fair question about whether the process is truly independent or fair, but you would need a lot of evidence to conclude that. A single streak is weak evidence.
Real devices can deviate from the ideal. Biased coins, worn dice or faulty random number generators exist, so a very unusual pattern is sometimes worth testing. The test is statistical, not a hunch from a run you happened to notice.
Why it matters for anyone who plays games of chance
Noticing a pattern in recent results can feel like insight, and it can encourage someone to keep playing, raise stakes or chase losses on the belief that the pattern will continue or reverse. Neither follows from the maths. Games with a house edge cost the player on average whatever the recent sequence looked like, and outcomes remain uncertain.
If gambling feels hard to control, a local support service can offer confidential help.