Stars and Stakes

Or, was the 2025-26 NBA season actually bad?

In the 2025-26 NBA discourse, everything’s amazing and nobody’s happy. I enjoyed Andrew Sharp’s roundup¹, but the short version is that there’s currently a tension between robust financials for the league and a sense that the quality of the product on the court is in decline. Personally, I’ve watched fewer NBA games this season than in years past, and I don’t feel like I missed anything. But my contrarian side is likes to test conventional wisdom. Is the product really getting worse or is something else causing our collective dissatisfaction?

I’m skeptical of a lot of narratives. Many believe that pervasive tanking, where teams intentionally lose in order to improve their odds in the NBA draft, is the primary problem. But the league is zero sum. A third of it will always be uncompetitive. I don’t believe that matchups including the worst teams have ever been compelling. Further, there are more options competing for our finite attention these days, so perhaps we have just raised our expectations. The NBA product could be objectively better and still lose fan enthusiasm if it’s competing with the infinite choice of Netflix instead of Seinfeld reruns on a random Tuesday night. Or maybe the subjective product quality actually depends on the celebrity status of the players. The league is transitioning out of the Lebron/Steph era and we don’t yet have new players transcending the sport into broader culture.

My own ambivalence towards the NBA regular season stems from feeling that 1) star players rarely play and 2) that games are rarely close. It’s a huge letdown when you expect to see Nikola Jokic battle Shai Gilgeous-Alexander and instead watch DaRon Holmes trade baskets with Cason Wallace in a 20 point blowout. This is in stark contrast to how thrilling it was to watch a healthy Steph Curry go nuclear against a healthy Kawhi Leonard in the play-in tournament.

The Setup

Like that play-in contest, compelling NBA games generally have two things: stars and stakes! Are those criteria rarer today? I pointed a Jupyter notebook at the unofficial NBA.com API to see if I could confirm or deny my suspicion that they were. I pulled data for three regular seasons:

• 1996-97: The year after the Chicago Bulls’ 72-10 record²
• 2015-16: The year the Golden State Warriors went 73-9 and lost to Lebron James’s Cleveland Cavaliers in the NBA Finals.
• 2025-26: This year!

I chose the first two seasons because I remember them as high water marks for NBA quality and popularity. I made plans with friends around Warriors/Cavs/Thunder matchups in 2016. However, my life circumstances have changed since then; maybe I’m just busier now. I wanted to know if those seasons looked different in the data.

I made an important assumption: most games in the NBA season do not matter. The subjective quality of the product is largely driven by marquee matchups. I define this as matchups between the top N teams in the league. I pulled data for matchups between the top 10 teams from each season. And I pulled game logs for the top 3 players on each team in those matchups. Ultimately, I restricted the analysis to matchups between the top 6 teams, focusing on the top 2 players for each team. The stakes of the regular season are unavoidably lower, but watching All-NBA caliber players on championship contenders battle is a good approximation of the games that actually matter in the NBA regular season.

The criteria for what makes a player a star were rigorously crafted using the most advanced statistical methods available to mankind…

Actually I just built a different Jupyter notebook to let me manually select the teams and players. My choices are subjective, though I imagine they would match general consensus from NBA fans.

Star Appearances

Do NBA stars miss more games in 2025-26 than in previous years? The answer is yes. Stars missed about a quarter of their games in marquee matchups this season. That number was lower in in 2015-16, and in 1996-97, the top 4 players in marquee matchups only missed 2 games total!

The per-player breakdown adds useful context. Blake Griffin, Jayson Tatum, and Jalen Williams (who account for the majority of missed games) had long term injuries. This runs counter to the narrative that stars just take nights off. There’s more gray-per-player in 2025-26, but honestly, I expected more absences this year. Perhaps this chart explains why:

The majority of marquee matchups in 2025-26 had missing stars. That feeling of “eh, this game doesn’t really matter because the lineups will be different in the playoffs” is real. Visually there’s a downward trend, but I’d be cautious about drawing a conclusion without adding additional years to the analysis. Either of our comparison years could be anamolous. Still, this season compares unfavorably to two of our subjectively great seasons.

Competitive Stakes

Maruqee matchups generate anticipation for the average fan, but even these can be boring if the game isn’t competitive. My intuition is that NBA games are less competitive, but the data don’t support that. There’s no obvious trend in our the point differential histogram below, and the summary statistics are similar across seasons. There was a mix of close games and blowouts in all 3 seasons I looked at.

But what are the chances I’ll see a competitive game in a marquee matchup? Well, this season actually looks pretty good in that regard! I looked two definitions of “competitive”:

1. The game was within 10 points sometime during the 4th quarter. This says that the losing team had a decent chance to win late in the game.
2. The game was a clutch game: the score was within 5 points within the last 5 minutes.

Games between championship contenders are overwhelmingly likely to be at least somewhat close, across all 3 seasons I looked at. This runs counter to my intuition, and it’s why we check the numbers before shipping our 🌶️ takes.

Stars and Stakes

What really matters, though, for the average fan, is the combination of these criteria. “Did I get to see a close game between the best players on the NBA’s best teams?” Our missing players chart already hints that the answer is often “no”, but game quality looks even worse when you factor in competitiveness. Only 15% of marquee matchups in 2025-26 were competitive with all 4 of the top players playing. That is 6 total games in the season! One per month! 1997-98 had 22 such games, about one per week.

So, we’ve broken Betteridge’s Law: The 2025-26 NBA season really was objectively bad compared to previously great seasons. In plain language, in 1997 I could be pretty sure that I’d see a great game with the best players on Sunday’s NBA on NBC matchup. That wasn’t true this season.³

What To Do

A problem well-stated is a problem half-solved

- Charles Kettering

We’re breaking rules of thumb like crazy today. Because while our analysis surfaces a problem clearly, solutions are elusive. For example, the NBA’s Player Participation Policy is purpose built to increase the quantity of high quality games, but it has not been effective. Today’s game is genuinely more physically demanding, and injuries are more likely. It’s not just about tweaking player incentives. And the tradeoffs for reducing the physical demands are not cheap. Shortening the season, the most obvious solution, would negatively impact the league’s revenue. Other options require radically tinkering with how the game and season are structured. It’s not as though the NBA isn’t trying, with the in-season tournament and play-in tournaments as the most visible examples.

I’m writing this during the 2026 NBA playoffs, and I’ve watched as many games in the past two weeks as I have in the previous 3 months. The formula is pretty simple. The postseason has higher stakes, and stars are more likely to participate. Here’s hoping the league finds a way to give us more of both in the regular season. Maybe they should see if Theo Epstein has any spare time in his calendar.