What Are the Odds of Getting Heads Five Times in a Row?
A run of five heads is one specified five-flip sequence, H-H-H-H-H, in a model with 2^5 = 32 equally likely ordered sequences.
A worked example with numbers
Write all five flip positions, not just the count. H-H-H-H-H is one sequence; H-H-H-H-T is another. Each named sequence has probability (1/2) × (1/2) × (1/2) × (1/2) × (1/2) = 1/32. If a player asks for at least one run of five heads somewhere within a longer session, that is a different question with many overlapping possible starting positions; 1/32 is not the answer to that broader event.
How to set the rule before the result
Use the heads-or-tails series tool to record a chosen number of independent flips, or use the streak simulator to explore a stated trial count. Before the first flip, say whether you are watching a pre-named block of five or searching a long history for any streak. Preserve every flip in the planned sample; deleting a tail before a run begins changes what was observed.
Common mistakes that change the odds or the process
The gambler’s fallacy says a tail becomes more likely because heads have appeared repeatedly. In the independent fair model, P(Heads on flip 6 | five earlier heads) = 1/2. The opposite error is treating five heads as proof of a rigged coin. A 3.125% event is uncommon in one named five-flip block, but it is expected occasionally across many games and many players. Test a physical coin only with a preplanned, much larger method.
Where this method stops being appropriate
The calculation assumes independent flips with a stable heads probability of 1/2. It does not apply unchanged to a trick coin, a picker that removes an outcome after selection, or a report that was noticed only after many attempts. Do not use the calculation to make gambling claims or to infer a person’s skill from a short streak. Probability describes a model of repeated trials, not a guarantee about the next toss.
How the random source fits into the rule
NIST’s binomial-distribution reference defines the model used when a fixed number of independent two-outcome trials is counted. A browser flip can use the cryptographically strong values described by MDN’s Crypto.getRandomValues reference, but the formula still depends on the rule that each trial is a fresh 1/2 mapping. A physical coin needs an additional fairness assumption; the equation cannot inspect the coin.
Calculate a pre-named run step by step
For one named five-flip block, multiply five conditional probabilities: 1/2 × 1/2 × 1/2 × 1/2 × 1/2 = 1/32. As a percentage, divide 1 by 32 and multiply by 100: 3.125%. The same value applies to T-T-T-T-T and to every other exact five-symbol sequence. A run looks unusual because people group matching symbols visually, not because that one sequence was assigned less probability before the tosses.
Five in a row somewhere is another event
Imagine recording 30 flips and then looking for any five-head block. The possible starting positions overlap: flips 1–5, 2–6, through 26–30. Those events cannot simply be added because one long run can satisfy several positions. The chance is therefore larger than 1/32 but needs a calculation for the specified session length. State whether the question names the first five flips or searches a longer record before quoting a number.
Use conditional probability correctly
After H-H-H-H, there are only two equally likely next outcomes in the fair model: H or T. The chance that the completed pre-named five-flip block is all heads from the beginning remains 1/32; the chance of a head on the already reached fifth flip is 1/2. These are compatible statements because they condition on different information. A past run cannot exert a balancing force on a fresh independent toss.
When a streak is useful information
A streak can prompt a question about procedure: was the coin actually the same, were every toss and reroll recorded, and was the sample selected in advance? It cannot answer those questions alone. For a classroom demonstration, retain the whole ordered list. For a consequential random process, use its documented testing and audit procedure; five observations provide no sound basis for approval or accusation.
Put the percentage in a usable scale
A probability of 3.125% means about 1 named all-heads block in 32 independent five-flip blocks on average. It does not mean that every set of 32 blocks contains exactly one. One set can contain none, another can contain several. The reciprocal is useful for intuition only when the blocks are pre-specified and non-overlapping.
When someone notices five heads inside a long history, ask when they began watching and how many other patterns would have attracted attention. Recording that stopping rule protects the calculation from hindsight. The observed history then has a defined event rather than an after-the-fact story.
Name the direction of the run
Five matching outcomes of either kind has probability 2/32 = 6.25% in one named five-flip block, because HHHHH and TTTTT are two acceptable sequences. Five heads specifically is 1/32. Changing “heads” to “either all-heads or all-tails” doubles the event and must double the numerator.