Record against run differential

The question

The first move in any "are they for real" conversation is to look past the record to the run differential, on the grounds that how much a team outscores its opponents carries forward more reliably than how its close games fell. That is the Pythagorean tradition, and it is usually right. This study asked whether it held in DataBaller's own data, sport by sport, and which version of the margin actually predicted the rest of a season.

What was tested

The claim, written before the run: a team's season-to-date scoring margin predicts its rest-of-season win rate better than its season-to-date record does. Tested leave-one-season-out at a quarter, a half and three quarters of each team's schedule, on five MLB seasons (2021 through 2025), six NBA seasons (2020-21 through 2025-26) and six NFL seasons (2020 through 2025), judged by a majority of seasons and by pooled error. Pythagorean record and a capped per-game margin were run as checks. Later rounds changed the target to the one a rating actually serves, pricing individual games, and tried a margin rebuilt from run ingredients.

What turned up

In MLB the record won. It beat raw run differential in four of five seasons, with the differential 2.3% worse pooled. Pythagorean record at the standard exponent was 2.7% worse, and a margin capped at four runs a game 1.0% worse. A mid-season MLB record predicted the rest of the season slightly better than its run differential did.

In the NBA the differential won clearly, in five of six seasons and 1.9% better, with Pythagorean record 2.7% better. In the NFL it won three of six, short of a majority; capping each game at 14 points, a two-score lead, took it to four of six and 1.9% better, a form chosen after the fact.

What beats the record depends on the target. For rest-of-season wins, the MLB record is nearly unbeatable. For pricing individual games, the record plus a run differential rebuilt from the ingredients, singles, doubles, triples, home runs and walks on both sides, won in every MLB season: rest-of-season margins 2.6% better in four of four, 4.0% on the 2025 holdout and 3.9% on 2026 in progress, and game forecasts 0.27% better with the holdout confirming at 0.29%. Rebuilding expected runs from components strips out when the hits happened to bunch, and that is where the signal was.

What it means when you ask

In MLB a raw run-differential gap is weaker evidence than the habit assumes. DataBaller's Power Rating for baseball therefore reads the record and a component run differential, and says whether a team's strength is its run creation or its run prevention in those terms. In the NBA the margin does the work it is famous for. In every sport, teams far ahead of their margin-implied record still fell back and teams far behind still recovered, in every season tested, which is the claim behind Team Sustainability.

What it does not mean

"Record beats differential" does not mean margins do not matter in baseball; the extreme gaps reverted every season. The MLB result does not carry to the NBA. The capped NFL form is not a validated finding, and the football rating's own claim is game pricing only, on the thinnest margin of the three sports.

The record

Measured 23 August 2026 with backtest_team_rating.py, with the exploration rounds written up in the team decision-metrics design. The MLB and NFL 2025 holdouts were consulted three times across those rounds and are no longer pristine, so the binding tests are the first unseen seasons: the 2026 NFL season and the 2027 MLB season. The full portfolio and its testing rules are on the decision metrics page.