Pythagorean expectation in baseball

Pythagorean expectation estimates the winning percentage a team should have, based only on the runs it has scored and the runs it has allowed. Bill James invented it in the early 1980s, and the name comes from the formula’s resemblance to the Pythagorean theorem — squares over a sum of squares — not from anything to do with triangles on a baseball field.

The idea behind it is almost obvious once stated: over a long enough stretch, teams that outscore their opponents win games. If you know how many runs a team pushed across and how many it gave up, you can estimate its record without looking at the record at all.

What makes it useful is precisely that it ignores the actual won-lost column. That independence is what lets it act as a check on it.

The original version squares both numbers:

Win% = RS² / (RS² + RA²)

where RS is runs scored and RA is runs allowed. A team that scored 750 and allowed 700 comes out at 750² / (750² + 700²) = 0.534, or about 86 wins in a 162-game season.

Analysts later found that a slightly smaller exponent fits real baseball better. Most modern versions use roughly 1.83 instead of 2, and the more refined Pythagenpat method sets the exponent from the game’s run environment rather than fixing it. The differences are small — usually under a win across a full season — so the simple squared version is perfectly good for a back-of-envelope read.

The formula also generalizes. The same shape, with different exponents, estimates expected records in basketball, football, and hockey.

The interesting number isn’t the expectation itself — it’s the difference between a team’s real record and its Pythagorean one.

A team winning more than expected has usually done it by winning close games and losing the blowouts. That distribution is mostly luck: it depends on how a season’s runs happened to be bunched, and bunching does not repeat reliably. The same is true in reverse for teams underperforming their expectation.

A gap of four or five wins in either direction is common and doesn’t need explaining. A gap of eight or ten is a genuine flag: this team’s record is telling a different story than its run scoring, and the run scoring is usually the more honest witness.

The practical use is prediction. A team five games over its expectation is a decent bet to slide; a team five games under is a decent bet to improve.

Pythagorean expectation is a good rule, not a law, and it has known blind spots.

• Bullpens bend it. A team with an excellent closer and a poor long-relief corps really can win more one-run games than its run differential implies, because its best pitching is deployed exactly where it moves win probability most. That’s a repeatable skill, not luck.
• Small samples are noisy. Over 20 games the formula tells you almost nothing. It needs most of a season to settle.
• A single blowout distorts it. One 20–2 loss inflates runs allowed in a way that says little about the team’s typical performance. Some analysts cap extreme margins for this reason.

Treat a large gap as a question worth asking, not an answer on its own.

Working this out by hand every week is a chore, which is why TwentySeven computes it for you. The season view lists Pythagorean over- and under-performers across the league, so you can see at a glance which contenders are being carried by close-game luck and which teams are better than their record looks.

It pairs naturally with the rest of the season picture — run differential, power rankings with heat trends, and playoff odds that fold all of it into a single probability.