A basketball team is ahead by 14 points. It has arguably its best player, the opponent is dead on their feet and all of the numbers indicate a good game. Next, the coach does something different. He drops his superstar to the bench and pits him there.
This can be a bit confusing for those who follow the betting odds. Why take your best player when there are another 10 minutes to go, and it could have been a great win without him!
After all, basketball coaches don’t seek out algorithms to impress. They are desperately trying to win games, develop players and endure an exhausting season. And sometimes, those priorities create decisions and choices that help no one except for the one making them and even then, not always.
The odds of winning by 20 points are not always better than the odds of winning by 5 points
Suppose that an NBA coach has a team up by 12 points with six minutes left to play. Suppose that an NBA coach has a team up by 12 points with 6 minutes to go. His opening centre has only been playing for 34 minutes and is tired. Will he be able to keep him on the ground?
It could be that a statistical model finds that the team has a higher probability of winning at the center than at any other location. The coach, however, has different knowledge. His team has to face another game tomorrow, and his center is having a physical issue.
Now it’s more important to defend than to extend.
The difference is significant, especially when it comes to betting on basketball games. Point spreads are based on point-by-point victories, not just wins. An underwhelming performance doesn’t mean a team failed to deliver what its coach was looking for.
Not that the coach is sleeping because someone missed the mark, that is. Nor is the coach losing sleep because someone didn’t predict correctly, that is.
The Utah Jazz were an interesting case study
The NBA levied a $500,000 penalty on the Utah Jazz for actions that occurred with Lauri Markkanen and Jaren Jackson Jr. throughout February 2026.
The league notes Utah took both players out of the game against Orlando and Miami in the fourth quarter. Neither returned, they were there to play on but the result was still up for grabs.
The NBA deemed those decisions to be detrimental to “competitive integrity. Commissioner Adam Silver singled out attitude that went beyond winning to draft position.
This is an unusual one, but an interesting one. Teams may have goals that go beyond the outcome for a specific game.
It is hard to know what those objectives are for prediction models. A franchise’s past throwing performance doesn’t necessarily tell you what it will try to emphasize tonight.
Rotations are just moving targets
The algorithms that make some people good at basketball thrive on routine. Minutes, lineups, and shooting opportunities that are always predictable, these all give them a reliable set of tools for their tasks.
Unfortunately, for the maths, coaches like to experiment.
For a young guard, he can get 25 minutes instead of his usual 10. A less skilled shooter might be able to be wheeled in as a defensive specialist in case the other team has an exceptionally dangerous solution.
All of these changes impact on the number of rebounds, assists, scoring efficiency and more.
Think Oklahoma City coach Mark Daigneault. Preparing for the 2025-26 season, he said that his staff had spoken with each player separately about their workload, which took into account the individual needs of the players for recovery.
That’s logical when it comes to playing basketball. However, certain situations are not always apparent in public data.
An algorithm can identify that the player typically plays for 32 minutes. It can’t automatically determine if his coach is going to give him 24 tonight.
The entertainment industry has also some surprises
There is a similarity between predicting basketball games and digital games. They are both mathematical systems used to describe uncertain results, but the process involved is very different.
A person strolling through the Nolimit City Slots games provided by BobCasino, for instance, will discover the games shaped on pre-programmed probabilities and unpredictable person outcomes. The human element of the game comes next with basketball – a choice that can alter the circumstances.
The mathematical rules of a slot do not change while playing. The basketball coach may switch his plans midway through the third quarter.
So let’s take a look at what sports are all about.
Rest Days can make a difference!
In 2023, the NBA implemented a new Player Participation Policy to prevent teams from squandering their most important players.
Some absences are limited by the rules, such as a few with multiple stars or those in nationally televised games. However, there is still a way to get proper injury management and proper rest.
Cleveland earned a $100,000 penalty in November 2025 for not playing Donovan Mitchell and Evan Mobley against Miami. The league found that it did not have the right to make such a decision because it interferes with its participation policy.
These are just a few examples of the importance of availability.
The NBA mandates teams to post injury numbers before games, and there are extra reporting dates for back-to-back contests. However, having knowledge that someone is there doesn’t mean that he will play for a set amount of time.
A player could start a game, do a great job and then sit quietly next to the assistant coaches the rest of the game.
The final minutes can tell a different story
Another difficulty occurs when making late game decisions.
If a coach guards a narrow lead they may slow down the offense, increase the chance of offensive players being forced into their backs, and decrease the amount of fast break possession. Another coach who has already got a good head-hold may adopt unskilled substitutes.
Both moves can have a significant effect on the game tempo.
This is why predicting the winner and predicting the exact winning margin are two different issues. However, even if a team is of a certain quality, there remains a degree of uncertainty with regard to the tactical options.
Lineup combinations, fatigue indicators, and historical coaching tendencies can be included in modern statistical systems. They aren’t helpless. But they still rely on information that could be incomplete.
Basketball is still wonderfully human
The fun part is that the coaches don’t always have to be unpredictable. They have to manage competing demands.
One night, the well-being of a player is the priority. The night of another promising rookie is being developed. The same coach may drastically reduce his rotation in the playoffs.
Algorithms can be trained on these patterns, but not all substitutions can be predicted by the algorithm.