What is the Pythagorean expectation? Using runs and points to spot lucky teamsGuides · Baseball · Stats explained

What is the Pythagorean expectation? Using runs and points to spot lucky teams

Key points

  • Uses total points scored and allowed to estimate the winning percentage a team "should" have
  • The exponent is about 1.83 for baseball and about 13.91–14 for basketball
  • Records built on one-run wins are more likely to come back down

1In one sentence

The Pythagorean expectation uses a team's total runs (or points) scored and allowed to estimate the winning percentage it "should" have. The baseball statistician Bill James came up with the idea, and it's named for the formula's resemblance to the Pythagorean theorem.

Its value is that runs scored and allowed are more stable than wins and losses. A team that wins lots of one-run games will have a good-looking win total but a small run differential; the Pythagorean expectation warns you that a record like that is more likely to come back down.

2The formula

Expected winning percentage = scored^x ÷ (scored^x + allowed^x)

x is the exponent. The original version used 2; later research found about 1.83 fits baseball better (the figure Baseball-Reference uses). Basketball has much more scoring and relatively small margins per game, so the exponent is far larger; the common value is 13.91 (the figure worked out by Daryl Morey). Basketball-Reference's glossary uses a rounded 14, so the numbers you see on different sites differ slightly. Different studies give slightly different exponents, but that's the order of magnitude.

Multiply the expected winning percentage by games played to get "expected wins". Actual wins minus expected wins is often used as a "luck" figure.

3A baseball example

Say a team has played 162 games, scoring 780 runs and allowing 700, with x = 1.83:

  • 780 to the power of 1.83 is about 196,100; 700 to the power of 1.83 is about 160,900 (rounded approximations).
  • Expected winning percentage ≈ 196,100 ÷ (196,100 + 160,900) ≈ 0.549.
  • Expected wins ≈ 0.549 × 162 ≈ 89 wins.

If the team actually won 95, that's about 6 wins above expectation. Common reasons are winning an unusually large share of one-run games or an especially good bullpen in high-leverage innings. That isn't necessarily fake strength, but historically this kind of "excess wins" usually shrinks the following year.

A real example is the 2018 Seattle Mariners. As MLB.com points out, they went 89–73, which looks like a decent record, but they allowed 34 more runs than they scored over the season, and their Pythagorean record was only 77–85. The gap of 12 wins between the record and the run differential is the classic case of "the record looks better than the team".

4How it works in basketball

An NBA season is 82 games with more than a hundred points per game, and the expected winning percentage from the 13.91 exponent is very sensitive to point differential. A team outscoring opponents by about 5 points a game on average has an expected winning percentage a little over six in ten; a margin of 10 points or more is usually a record near the top of the league.

So in the NBA, average point differential can almost be used directly as a measure of strength. A team with a high winning percentage but an ordinary differential has been scraping through a lot of games; a team with an ordinary winning percentage but an impressive differential has mostly lost close games, and its record has room to rise.

5How to use it when watching

  1. Open MLB standings or the NBA standings and find teams whose winning percentage doesn't seem to match their runs or points.
  2. Actual wins clearly above expectation: it doesn't mean they'll collapse, but check whether they've relied on lots of one-run or extra-innings wins.
  3. Actual wins clearly below expectation: often the bullpen or late-game moments have gone wrong, and the record tends to recover.

It's the same thinking as football's xG: use a more stable number to check how much of a record is luck. It isn't a prediction tool, and it doesn't account for injuries, trades or strength of schedule; it works best as a "health check" on a record.

6FAQ

13.91, 14 or 16.5: which should I use?

13.91 is Daryl Morey's figure, Basketball-Reference uses a rounded 14, and some (John Hollinger) use 16.5. Basketball-Reference notes that 16.5 is another common choice but has a larger error. The differences are small; what matters is using the same exponent when comparing teams.

Is there much difference between an exponent of 2 and 1.83 in baseball?

Not much. With the example above (780 runs scored, 700 allowed), an exponent of 2 gives an expected winning percentage of about 0.554 and about 89.7 expected wins; 1.83 gives about 89.0. The difference is less than 1 win, so the original version is fine for a quick estimate.

Can it be used for other leagues?

Yes. The formula is the same, but the exponent changes. In Basketball-Reference's glossary, the NBA uses 14 and the WNBA 10. Each sport has a different scoring environment, so the exponent that fits best differs too.

Is there something more precise than 1.83?

Yes. A more advanced method called Pythagenpat adjusts the exponent according to the two teams' total runs per game: the higher the scoring environment, the larger the exponent. For everyday watching, a fixed exponent is enough.

7Sources

Illustrated portrait of A-He

About the author

A-He

Editor-in-chief

Editorial pen name

A-He is the newsroom's editor-in-chief, covering cross-sport topics, tools, rules explainers and esports. She starts with source quality and the current rulebook, then asks which numbers actually answer the reader's question, making definitions, limits and exceptions easy to compare across sports.

View profile and articles ›

‹ Back to guides · Glossary