UFC betting is a challenging issue for any oddsmaker. Fights occur on a relatively rare basis, styles can lead to unexpected matchups, injuries are not always transparent and a single clean punch can change a number of rounds. While AI has the potential to analyze a wealth of data that a human bettor simply could not, this does not necessarily translate to better pricing of a fight.

Those prices ultimately translate to simple moneylines and prop markets for legal sports betting in New Jersey customers. However, behind the scenes, sportsbooks could be analyzing fighter records, strike statistics, takedown rates, age, reach, recent performances, and betting activity. That could be a process made quicker with the help of AI, but the UFC is still a very tricky sport to model.

AI Can Process More Variables at Once

Machine learning, of course, is all about scale. While a human oddsmaker can learn from previous fights and grasp the difference in style, an algorithm can analyze thousands of variables at once. It can check for accuracy, defensive percentage, takedowns, submission attempts, opponent quality and round-by-round performances.

It can come in handy when sportsbooks have to value an entire fight card. AI can also detect connections that are not readily apparent by hand. Maybe fighters with a certain level of reach and a certain level of takedown defense do well against certain styles. Machine-learning systems can scan historical data to look for those patterns.

The question is if those patterns are still applicable when the next fighter enters the Octagon.

UFC Data Is Relatively Limited

Football clubs may compete in 40 or 50 matches during a season. Baseball teams have 162 regular season games. The UFC fighters fight only 2 or 3 times a year.

That gives models much less data. Fighters can also drastically change from one outing to the next. They may shift their weight classes, change coaches, get injured, or find a whole new way of playing.

A model that has been trained extensively on the previous year’s performances might thus be analyzing a version of the fighter that no longer exists.

That is the basic difference between the pricing of a UFC and any modeling of a sport, where there is new data from teams and players every week.

Styles Create Problems for Pure Statistics

Fight records can also be misleading. A competitor may have a great number of takedowns due to the poor wrestling defense of their opponents. The numbers could be far less significant with an elite grappler.

Human oddsmakers can intuit those stylistic relationships in ways that purely statistical models might not be able to.

The same problem is present for striking. Two fighters can have the same accuracy, and fight in totally different ways. One can throw large quantities from distance, and another can counterattack. The number alone is not as important as the context.

For example, this is because the most effective AI systems are likely to enhance, not entirely replace, human traders.

Markets Provide Information Too

Sportsbooks don’t just calculate the price of fights on their own. After odds are posted, there is information from betting activity. When there’s a significant influx of funds on one fighter, operators might tweak the price to control risk or simply because the market seems to be saying something.

This holds particular importance for legal sports betting in New Jersey, where licensed bookmakers will be able to handle a large amount of wagering traffic in the lead-up to and during key UFC events.

AI can keep track of those movements in real time. It can spot irregular betting activity and cross-check activity across markets much faster than a human team could manually. This makes AI not just for setting an opening price, but also for managing odds after the market opens.

Injuries Remain a Major Blind Spot

A fighter could be in a fight with an injury he or she may not have disclosed to the public, but to his or her training staff. Another may have had a particularly challenging weight loss. Both scenarios may have a significant impact on performance. That’s one of the biggest challenges in UFC betting.

History might be saying one fighter should be favored and private information is saying the other way.

Artificial intelligence can’t fill that information void automatically. Indeed, even a very sophisticated model might generate a very accurate-appearing price on the basis of imperfect assumptions. One of the reasons human judgment is still important is that.

AI May Be Better at Finding the Starting Price

The most likely future is likely a hybrid approach. AI can analyze vast amounts of data and give an initial probability for each fighter. Those prices can then be adjusted by human traders based on style, recent news, injuries and other news that is hard to quantify.

After betting starts, prices can be adjusted frequently using automated systems that track market movement.

That may prove to be more useful than selecting between humans and machines, at least for sportsbooks that provide UFC bets.

It is not necessary for AI to tell you what is going to happen in order to make it better for betting on the UFC. It should just make it easier for sportsbooks to calculate odds.

Thus, the human oddsmaker might be able to hold onto his job in the face of the AI revolution. The distinction is that the next generation of traders will be accompanied by an algorithm.