The Mallory Group

Do supermax contracts make NBA players worse?

No. The numbers show a decline, but the contract does not cause it. Age and mean reversion explain all of it.

Research framing

The question

NBA teams give supermax contracts to a few of their best players. It is a deal only a player's own team can offer, worth up to 35 percent of the salary cap, more than any rival is allowed to pay him. Only veteran players who have earned honors like All-NBA, MVP, or Defensive Player of the Year qualify, and only with the team that holds their rights. The fear is that a guaranteed maximum payday changes a player. Maybe he coasts. Maybe the pressure hurts him. Either way, the worry is that performance drops after the signature.

The short answer

There is no supermax penalty. Supermax players do post worse numbers after they sign. But they are elite veterans, and elite veterans decline for two ordinary reasons. They get older. And they sign right after a peak season, so they regress toward their normal level. Control for age and prior performance, and the supermax effect disappears.

How I tested it

I compared team rosters season to season, from 2017 to 2024, to see who stayed, who changed teams, and who signed a supermax. I kept players who logged real minutes each season, so the sample is rotation players, not every name on a roster. For each one I measured on-court impact before and after, using Box Plus-Minus, a public metric from Basketball Reference. Then I split players into three groups by what they did: stayed, changed teams, or signed a supermax deal. I compared how each group's impact shifted.

What the data shows

Three findings came out of it:

  • Performance regresses to the mean, strongly. Top players decline. Struggling players bounce back.
  • The supermax penalty is a mirage. It disappears once you account for age and prior performance.
  • Players who change teams decline more than players who stay. That runs against the idea of a fresh start. Some of it is selection, though. Teams often move players who are already slipping.

The full numbers, with sample sizes and model details, are on the findings page. The code, the data, and the full method are public in the GitHub repo.