How the rankings work

Rankings aren't standings

Standings sort teams by how often they win. The rankings sort them by how much they win by, and against whom. Most of the time the two agree. When they don't, it's usually one of these:

• A team that wins big can rank above a team with a slightly better record that keeps winning close games. Close games are close to coin flips, so they say less about who's better.
• A tougher schedule lifts a team; a soft one holds it back.
• Two teams a few tenths apart are tied for any practical purpose. That's less than one free throw.

Why margin? Over a season, point differential predicts future results better than win-loss record does. It's the idea behind Bill James's Pythagorean expectation and the power ratings sports sites publish, such as Sports-Reference's Simple Rating System.

Team rating

Every finished game gives one equation:

home score − away score = home team's rating − away team's rating + home-court edge

With every game of the season together there's one set of ratings that fits all the results best (a least-squares fit, the same idea as the Simple Rating System). The home-court edge comes out of the same fit, from the season's own games. Ratings average zero, so +10 means 10 points better than an average team on a neutral floor.

Two guards keep it fair:

• Blowouts count up to 25 points. Winning by 50 counts the same as winning by 25, so running up the score doesn't pay.
• Every team starts with 2 imaginary games against an average team. Early in the season that pulls ratings toward zero; after a dozen real games it barely matters.

The rating, in parts you can add up

The fit sounds like a black box, but it isn't. Each team's rating works out to exactly:

rating = (average margin + schedule − home games) × games ÷ (games + 2)

• Average margin: points scored minus allowed per game, with blowouts capped at 25.
• Schedule: the average rating of the opponents it played. Positive means a tougher-than-average schedule.
• Home games: the home-court edge times (home games − away games) ÷ games. Teams with extra home games give a little back.
• Early-season pull: the 2 imaginary games, shrinking the total a little toward average.

That's what "See the math" shows for every team on the Rankings page, and the site checks that the parts add up to the rating every morning before it publishes.

Worked example

A team has played 20 games, 11 at home and 9 away. It wins by 9.0 a game after capping blowouts, and its opponents average +1.5. The season's home-court edge is 3.0 points, so the home adjustment is 3.0 × (11 − 9) ÷ 20 = 0.3.

(9.0 + 1.5 − 0.3) × 20 ÷ 22 = 10.2 × 0.909 = +9.3

Player rating

Player ratings arrive with this season's box scores. The formula is set now so it can be checked before the first number is published.

Freeman Rating = production + team impact

Production: Game Score per 40 minutes

Game Score is John Hollinger's one-number box-score summary (Basketball-Reference glossary). Misses and turnovers count against, so volume without efficiency doesn't win:

PTS + 0.4 FGM − 0.7 FGA − 0.4 (FTA − FTM) + 0.7 OREB + 0.3 DREB + STL + 0.7 AST + 0.7 BLK − 0.4 PF − TOV

Scaled to 40 minutes, a full game, so starters and reserves can be compared.

Team impact

Good players make good teams. Each player gets a share of her team's rating, sized by how much she plays: one of five on the floor, for her share of the game.

team rating × (her minutes per game ÷ 40) ÷ 5

Worked example

A guard scores 20 on 8-of-15 shooting and 3-of-4 free throws, with 2 offensive and 5 defensive rebounds, 1 steal, 4 assists, 1 block, 2 fouls and 3 turnovers. That line's Game Score is 15.9. If that's her average in 30 minutes a game, production is 15.9 × 40 ÷ 30 = 21.2. Her team is +20, so team impact is 20 × (30 ÷ 40) ÷ 5 = +3.0, and her Freeman Rating is 24.2.

A player needs 100 minutes before she's ranked; below that she's listed without a rank.

What the ratings don't know

• Injuries and lineups. The rating sees results, not who played. A team that just got its star back is better than its rating says. That's what the Film Room videos are for.
• Neutral sites. Our data doesn't mark them, so tournament games count the listed home team as home.
• Defense in the box score. Steals and blocks are only part of defense; great positional defenders are underrated by production and partly rescued by team impact.
• Small samples. Early-season ratings move a lot. That's why they start near average.

Every formula here is tested in code against worked examples like the ones above before any number goes live. Ratings are estimates, not facts; they're for information and entertainment, not betting (see the terms).