Method
Why a hot streak proves less than it looks
Random variation alone produces long runs of success in sports. A calculation of how often streaks of a given length should appear by chance, so you can judge whether a run is unusual or simply expected.
By Journal Info Editorial Team · Published 2026-11-15
Every season produces a hitter on a hot streak and a team that cannot lose. The streak is real — the hits happened — but the inference usually drawn from it is not. Long runs appear in random sequences more often than instinct suggests, and the arithmetic to check that claim is short enough to do on paper.
This article works through the probability of streaks under a simple model, shows why the model is the right first test, and explains what a streak does and does not tell you.
A simple model
Start with the simplest possible assumption: every at-bat is independent, and the batter has a fixed probability of getting a hit each time. If his season batting average is .300 (hits divided by at-bats, per MLB Glossary — Batting Average), treat each at-bat as a coin with a 0.3 chance of a hit.
Independence is an assumption, not a fact, and it is the thing a streak might disprove. That is exactly why it is the right baseline: if streaks appear at the rate the independent model predicts, they provide no evidence of anything beyond random variation.
Under this model, the chance of three hits in a row is 0.3³ = 0.027, or about 2.7%. That looks rare, and for one specific batter over one specific set of three at-bats, it is. The error is in the next step, where that small number is treated as the chance of seeing any streak at all.
Counting opportunities, not just outcomes
A hitter with 500 at-bats has 498 overlapping three-at-bat windows, not one. Overlapping windows are not independent (a window starting at at-bat 1 shares two at-bats with the window starting at at-bat 2), so the naive multiplication 498 × 0.027 = 13.4 overcounts. The exact figure can be computed by tracking the current run of hits through all 500 at-bats. For a .300 hitter, the chance of at least one run of three or more hits in a row is about 99.996%, which rounds to 100.0%. A streak that looks improbable per attempt becomes unremarkable across a season.
The same reasoning scales to longer streaks. Four hits in a row has probability 0.3⁴ = 0.0081, about 0.8%, and five in a row 0.3⁵ = 0.00243, about 0.24%. Those numbers keep shrinking, and the number of windows does not: a 500-at-bat season contains 497 four-at-bat windows and 496 five-at-bat windows. The exact chances for one season are in the table below.
| Run length (consecutive hits) | Chance of one specific run | Chance of at least one such run in 150 at-bats | in 500 at-bats |
|---|---|---|---|
| 3 | 2.7% | 95.1% | 100.0% |
| 4 | 0.81% | 57.6% | 94.5% |
| 5 | 0.243% | 22.2% | 57.4% |
| 6 | 0.0729% | 7.2% | 22.4% |
| 8 | 0.0066% | 0.7% | 2.2% |
| 10 | 0.0006% | 0.06% | 0.2% |
The table assumes every at-bat is an independent 0.3 chance of a hit. A five-hit run, which sounds like a notable stretch, occurs somewhere in a 500-at-bat season for a .300 hitter more often than not (57.4%). An eight-hit run is rare (2.2%), which is where a streak starts to become a real surprise under this model. These are exact probabilities for a model with a constant .300 chance, not claims about any real player.
The arithmetic that resets intuition
The clean way to get a feel for it is to think about what happens when you relax the definition of a streak. If “hot streak” means eight hits in ten at-bats rather than eight in a row, there are more ways to qualify. For one fixed ten-at-bat window the chance of eight or more hits is only 0.16%, but a season has hundreds of overlapping windows, so the chance of at least one grows well beyond that. The moment you allow a streak to be described after the fact — a run of games, a stretch of at-bats, a hot month — you are no longer testing a prediction. You are choosing the window that fits.
That is the mechanism behind most streak stories. The event is selected after it happened, from a set of possible events that was never specified in advance. A prediction made before the fact and then confirmed is evidence. A pattern noticed afterwards is a description.
What would count as evidence
The distinction gives a practical test. A streak is evidence of a real change in skill if it survives three conditions:
- It was predicted in advance. The threshold was stated before the at-bats happened, not chosen to match them.
- It repeats. A genuine shift in ability shows up again in the next comparable sample, not once.
- It comes with a mechanism. A change in approach, health or role that would explain the change in outcomes.
Without those, the streak is an observation about the past and nothing more. It may still be a good story, but it is not evidence about the next game. These three conditions are our suggestion for how to read a streak, not a published test.
Where the model does break
The independent model is a baseline, not a truth. Real batters face different pitchers, in different parks, in different game states, and their true ability varies across a season. Those factors make the model approximate. The factors can cut both ways: some could make streaks more likely than the model predicts, and some less likely. The independent baseline is the right place to start, and any claim that a streak is meaningful has to show how far it departs from that baseline.
Applying this to other sports
The same structure applies wherever you see a run: a team winning eight straight, a goalkeeper with a run of clean sheets, a shooter who cannot miss. In each case the questions are identical. What is the baseline rate per attempt? How many attempts were there? Was the streak defined before or after it happened? Answer those three and most “impossible” runs turn out to be ordinary.
Method
The probabilities come from a short script that tracks the current run of hits across 150 or 500 at-bats with a constant 0.3 chance of a hit in each at-bat, and from the binomial formula for the ten-at-bat window. The .300 rate is an illustration. Nothing here uses real player data. The batting-average definition was read on 2026-10-31.
Sources
- MLB Glossary — Batting Average (AVG): hits divided by at-bats, the rate used as the per-at-bat probability in the model.
AI assistance recorded as: AI draft, not edited