Hong Life CalculatorsUnderstanding the winning streak probability calculator — concept and principle
How surprising is a five-game winning streak? This calculator gives two different answers: the chance that a run of at least k straight wins appears somewhere in a stretch of n games, and the chance that the next k games are all wins. Use the first to judge whether a streak in a season is unusual.
How it is calculated
- Next k all wins = p^k, where p is the win probability per game.
- Streak somewhere in n games is found exactly by dynamic programming. Track the probability of being on a current run of 0, 1, …, k − 1 wins. Each game, a win moves you from run j to run j + 1 (with probability p) and a loss sends you back to 0 (with probability 1 − p). When a win would take you to k, that probability is added to the "streak reached" total and removed from play.
Games are assumed independent with the same p.
The principle
A long sequence offers many starting points for a streak. There are 26 possible windows of five games in 30 games, so even though any single window succeeds with a probability of only 7.8%, the chance that at least one does is much larger. The windows overlap, so you cannot just multiply or add them; the dynamic programming method handles the overlap by counting each sequence once, at the first moment a streak is completed. A classic result (Schilling) is that the longest run in n games is typically close to log base 1/p of n(1 − p).
Worked example
- Inputs: p = 60%, k = 5, n = 30.
- Next five all wins: 0.6^5 = 0.07776 = 7.78%.
- Running the run-length recursion over 30 games gives a total probability of 0.64428, so the chance of at least one five-game winning streak is 64.43%.
- For comparison, a 50% team over the same 30 games has about 36.8%, and a 60% team over only 10 games about 23.3%.
| Win probability per game (%) | P(streak of k somewhere in n) |
|---|---|
| 15 | 0.17% |
| 30 | 4.43% |
| 45 | 24.95% |
| 60 | 64.43% |
| 75 | 94.43% |
| 90 | 99.96% |
| 120 | 103.97% |
How to read the result
If the "somewhere" probability is high, a streak of that length is normal for a team of that quality and says little about a change in form. If it is low (for example under 5%), the streak is genuinely unusual under the independence assumption, which may hint that p was underestimated. The rule-of-thumb longest run for p = 0.6 and n = 30 is log(12)/log(1/0.6) ≈ 4.9, matching the result that a five-game streak is slightly more likely than not.
Common mistakes
- Quoting p^k as the chance of a streak "this season"; it covers only one specific window.
- Treating p as fixed when opponents and venues vary.
- Forgetting that "at least k" includes longer streaks.
Frequently asked questions
How do I get the chance of a losing streak?
Enter the loss probability (100 − win %) as p; the same maths applies.
Is the "hot hand" real?
Studies disagree; streak probabilities like these are the baseline any hot-hand claim must beat.
Why is the result not just n × p^k?
That would double-count overlapping windows and can exceed 100% for long seasons.
Sources: Mark F. Schilling, "The Longest Run of Heads", The College Mathematics Journal 21(3), 1990 · William Feller, An Introduction to Probability Theory and Its Applications, Vol. 1, Wiley (recurrent events and runs)