The fallibility of human athletes keeps human oddsmakers employed in yet another business where artificial intelligence is ascendant.
The advance of technology and a growing appetite — among sportsbooks, but presumably, eventually American bettors — for fast-paced in-game betting markets has almost expunged humans from the process already. Even sports with paces of play more conducive to in-game wagering — like golf and baseball — leave few stagnant moments to synthesize data from the fields of play into lines that bettors have time to consume.
But these artificial intelligence systems still require human oversight, said Max Wright, chief commercial officer of IMG Arena, because while they excel at assessing data points, they struggle with boiling points and low points.
“Trying to understand things like momentum is really hard for a computer algorithm to determine.”
“I think that the point at which you have a solution that can be left entirely on its own with no manual oversight or intervention is still a long way off,” Wright told Gaming Today. “Trying to understand things like momentum is really hard for a computer algorithm to determine. The data points will tell them that Player X is still in the lead, but Player X may have just double-bogeyed the previous hole, slid back into the field, had a temper tantrum, may look dejected. The caddy hasn’t been able to calm them down.
“How do you input that data point so that the model can process that? I don’t think it can, at least not currently anyway.”
Wright surmised that even for the next 10 years, what novices conceptualize as all-knowing AI will still require a human supervisor to account for human elements. For now, the system and the trader will keep to their own “layer.”
“You’ve got one [layer], what the maths tells [sic] you should be the outcome,” Wright explained, “based upon years and years and years worth of historical data about what happened in similar situations. And that will create the true probability. And the betting markets are offering true probability.
“But then there’s pricing in a margin and an outcome into all the different solutions.”
Humans Move Into Supervisor Roles in Sportsbooks
So for now, humans remain in the room. But doing much less.
“Live betting has always needed a huge degree of reliance upon maths and algorithms and automation in order to deliver its product,” Wright said. “If you rewind to retail betting experiences from 30 years ago, what you had was, it was a very manual process: people pricing up horse races through experience and knowledge and mathematics on paper, and then publishing those prices.”
“Today, we are almost entirely automated in the way in which we produce the live betting experience. You have betting algorithms that are set up to process live information coming from the course and then to spit out live odds. And that is in the modern context, is a reasonably basic form of AI.”
Wright used a PGA Tour moment from the Arnold Palmer Invitational in March as an example.
“So even though it might be overwhelmingly likely that [Kurt] Kitayama will win from the position that he’s in, as we saw at Bay Hill [earlier this month] when the drive went left on — was it on 10? — that has a big effect and that’s something that a trading team will look at, and they will start to price in, and it will affect the outcome of different players.”