Signal vs Noise

Your Record Is What Your Points Say It Is

Point differential and win percentage are the same fact in different clothes. When they disagree, the points are telling the truth, and next season usually collects the debt, at a rate of about 37 points per win.

Strip a season down and the standings are just the scoreboard added up. Within any season since 1999, the correlation between a team's point differential and its win percentage runs between 0.87 and 0.95, averaging 0.911 across 861 team-seasons. In an average season, 92 percent of teams sit on the side of .500 their point differential points to, and that sign agreement has held every single year, through rule changes, scoring eras, and the 17-game schedule. The record and the points are the same fact wearing different clothes.

Which makes the interesting teams the ones where the clothes do not match.

The exchange rate

First, the conversion table. Wins trade for points at a stable rate of roughly 35 to 40 points of differential per win off .500. Predict a team's win total as half its games plus its point differential divided by 37 and you land within two wins of the truth for 89 percent of team-seasons, and for at least three quarters of teams in every individual season. The fine print: the measured season-by-season rate ranges from about 30 to 45 points per win, so 37 is a good working number, not a constant of nature. Outscore opponents by 37 over a season and you have bought yourself about one win above average.

Luck regresses, hard

Now the teams whose record outran their points, the 9-8 team with the differential of a 6-11 team. Call the gap luck, because that is what the data says it is: a team's next-season change almost never continues in the direction of its current luck. Across the span, only about 32 percent of teams moved the way their luck pointed, below a coin flip in all 26 consecutive pairs of seasons. Wins the points did not earn get given back.

At the extremes the regression is brutal. Teams two or more wins ahead of their points improved the next season 13 percent of the time, 9 of 68, and declined by roughly three wins on average. Teams two or more wins behind their points improved 84 percent of the time, 51 of 61, by roughly three wins. If you want one screen for next season's risers and fallers, this is the one the record endorses most strongly.

A team two wins ahead of its points has improved the following season 9 times in 68. The debt gets collected.

Where the luck actually lives

The residual has an address: one-score games. Teams well ahead of their points-based expectation hold a winning record in games decided by one to eight points essentially always, against a league base near 45 percent, and the luck residual correlates at +0.89 with one-score win percentage versus +0.42 with everything else. Part of that is arithmetic, narrow wins are how a record outruns its points, but it tells you what a lucky season physically looks like: a pile of field-goal finishes that broke one way. Close-game records are where luck hides, and close-game records do not persist.

The caveats, printed

Two wrinkles. The sign-agreement rate bottomed at 0.69 in 2016, the one season the identity got genuinely loose, and rebounded immediately. And the rank-space version of this rule, that when a team's rank in wins and rank in points disagree by eight or more spots next season follows the points, has hit at 72 percent over the span but went just 2 for 7 in the 2022 through 2024 cohorts. Small recent samples wobble; the quarter-century direction has not.

How to use it: every August, look up who won more games than their points earned. Those teams' fans are being set up. The team that went 10-7 outscoring nobody is a 7-win roster holding a lottery ticket that already paid out.