How grades work

What goes into it (16 metrics)

The field is the qualified pool — players with 300 minutes or more per 82 games that season, which is 434 players in 2025-26. They are not split by position. Compare guards only against guards and “a good rebounder for a guard” starts to look like “a good rebounder”.

Where “players graded” is below 434, players short of the attempts cutoff are shown as “—”.

Shooting

On-ball

Off-ball

Rebounds

Defense

“A 1 : B 2” is the ratio the parts are weighted in. Each part is turned into a rank first, then added in that ratio — add the raw values together and whichever part has the larger units decides the order by itself.

Each metric name leads to a page about that metric alone, with the number of qualified players by season and 9 seasons of leaders.

From rank to grade

For each metric the qualified players are ranked, and the grade comes from the percentile alone. Standard deviations from the mean (z-scores) are not used. The shape of the distribution differs wildly from metric to metric, and a rank does not care about shape.

The bands are symmetrical: S and G are the same width, and so are A and F.

What this site will not do

Metrics measured as rates are unreliable while attempts are few. This site handles that in two steps.

• Shrinkage — the fewer the attempts, the harder the value is pulled toward the league average. How hard is worked back from measuring the same metric across 9 seasons and asking what share of its variation comes from skill
• An attempts cutoff — anything still short of it is left unranked. “—” is not below G; it means there were too few attempts to rank. Metrics that measure volume itself carry no cutoff, since doing little already shows up correctly as a low mark

The cutoff and the shrinkage are one setting, not two. Move only one and the bar for being graded no longer matches how hard values are pulled back.

What these grades cannot measure

Offense and defense can be measured in very different detail. On offense, who did what is almost all on record; on defense, much of the work that matters never appears in public data — rotating correctly, navigating a screen, and above all deterring an opponent from going that way at all. Defensive grades are an approximation built from the acts that can be counted. It is the same reason the tags lean toward offense.

• On-ball defense is built from steals, contests and suppression — but disciplined containment leaves no trace in the box score, so defenders who work through positioning alone barely register here
• The opponent FG% behind rim protection is credited to whoever was nearest, so the team's defense rides along with it. Read it next to how much they defended
• Passing is measured as AST% — that is, how many of a team's baskets you had a hand in, not how well you see the floor. A great pass that ends in a miss counts for nothing; an easy one that drops counts as an assist
• DREB%, behind rebounding, depends on who else is on the floor: several strong rebounders alongside you and your share falls
• The matchup data behind suppressing the other team's best player is estimated from tracking, not an assignment a coach made. Switches and help change who gets credited. Only summed over a whole season does it become a usable tendency

Grades during the season

Early in a season the numbers move: with few attempts the estimate leans on the league average, and it settles as games are played. A “—” marks an item whose attempts have not reached the cutoff, which scales with the team's games played.

Grades move while a season is running. Even when a player does not change, a rank shifts as new players qualify around them. They settle once the regular season ends; past seasons no longer move.