Regression to the Mean in Sports: Why Great Seasons Are Followed by Good Ones
Why an outstanding season is usually followed by a merely good one, how much of a record is luck in different sports, and how to read a career year.
Any sport · Calculator
Bill James noticed that a baseball team’s winning percentage tracks the ratio of runs scored to runs allowed, and wrote the relationship in a form that resembles the Pythagorean theorem. The same idea has been applied to other sports with exponents fitted to their scoring.
Its common use is as a check on a record. A team that wins well beyond what its scoring margin implies is usually read as having overperformed, and MLB’s glossary uses exactly that reading to explain why such a team should not be expected to repeat its record. Enter the real wins too, and the calculator shows the gap.
By Reign Creative LLCPublished More in the sports hubs
Use totals from the same games: runs for baseball, points for basketball, goals for hockey.
Expected winning percentage.561
Expected winning percentage
win% = scoredx ÷ (scoredx + allowedx)
Expected record
expected wins = win% × games played
The exponent x sets how strongly the scoring margin translates into wins. Bill James’s original used 2. Later work fitted exponents to each sport’s results; the presets in the calculator are the ones a cited source publishes:
| Preset | Exponent | Source |
|---|---|---|
| Baseball — Bill James’s original (2) | 2 | MLB.com glossary — Bill James’s original formula |
| Baseball — Baseball-Reference (1.83) | 1.83 | MLB.com glossary — the exponent Baseball-Reference uses |
| Basketball — Basketball-Reference (14) | 14 | Basketball-Reference glossary — its Pythagorean wins formula |
| Hockey — James’s formula as applied by Ryder (2) | 2 | Ryder (2004), Hockey Analytics — James’s formula interpreted for hockey |
There is no American football preset: we could not open a published source for the exponent usually quoted for it, so if your reference gives one, enter it as a custom exponent.
Two illustrative seasons — not real teams: a baseball club that scored 800 runs and allowed 700 over 162 games, and a basketball team that averaged 112 points and allowed 108 over 82 games (9,184 and 8,856 in total). The hockey row is Alan Ryder’s own example: four goals for and three against.
| Preset | Scored – allowed | Expected win% | Expected wins |
|---|---|---|---|
| Baseball — Bill James’s original (2) | 800 – 700 | .566 | 91.8 of 162 |
| Baseball — Baseball-Reference (1.83) | 800 – 700 | .561 | 90.8 of 162 |
| Basketball — Basketball-Reference (14) | 9,184 – 8,856 | .625 | 51.2 of 82 |
| Hockey — James’s formula as applied by Ryder (2) | 4 – 3 | .640 | — |
The basketball team outscores its opponents by less than 4% yet projects to win well over half its games: with an exponent of 14, small margins carry a lot of weight. In baseball, the 100-run margin is worth about ten wins above an even 81–81 record with either exponent.
For how simulators turn ratings into results — and why a good team still loses — read understanding probability in sports simulations.
162-0 Baseball Draft & GM Team asks a related question from the other direction: it projects the record a roster you draft would post over 162 games, from roster quality, position fit, hitting, pitching, speed, defence and coaching.
Reign Creative LLC makes 162-0 Baseball Draft & GM Team, so we have a commercial interest in it. It is not affiliated with, or endorsed by, any league, team, players’ association or athlete.
Good to know
League, governing-body and team names appear here only to identify the formulas and rules being used. Reign Creative LLC's sports games and calculators are not affiliated with, endorsed by or licensed by any league, team, players' association or athlete, and all names and marks belong to their owners.
Why an outstanding season is usually followed by a merely good one, how much of a record is luck in different sports, and how to read a career year.
Why a better roster still loses, how season length compounds small edges, and how many runs you actually need before a simulated result means anything.
Inside a sports sim: how ratings become probabilities, where randomness enters, the two main engine designs, and what a simulated record does and does not mean.
What ERA, WHIP and WAR are telling you on a draft card, what they leave out, and how to read a projected 162-game record without over-trusting it.
The difference between tearing a roster down and reshaping it, the honest tests for which your team needs, and why the worst outcome is doing neither.
Baseball
The full slash line, isolated power and BABIP from one hitter’s counting stats.
Open the calculatorBasketball
True shooting, effective field-goal and the plain shooting percentages from one box-score line.
Open the calculator