AI’s predictive capabilities and how they will affect sports gaming, was structured as another of the red zone-green zone debates that the trade show featured this year. The four-man panel was moderated by Rick van der Kleij. Emphasis at the outset was already heavily leaning one way; I counted dozens seated in the green zone, in favor of the proposition, and just nine in the red zone.
The panel itself was evenly split – two in favor, two opposed. Sky Betting Trading Director Andy Wright described himself as “unconvinced” of the value of AI, and said he was struggling to understand how AI can help in attracting customers and in measuring how social trends in communication over sports down the pub has an immediate influence over price points on bets.
This kind of innate understanding of humans, Wright said, might be tougher for machine to learn and replicate. He asked how AI can dream up innovations, like humans can, and said that operators are looking now for areas of innovation AI can’t manage, and the answer tends to be the real human stuff – creativity, imagination, social awareness, and empathy. In other words, he asked, can AI learn to understand humans like humans understand each other?
BestBet360 CEO Dolan Beuthin was also in the red zone. He said he, too, struggled with a few areas of AI.
“How will it understand changes like the weather?”, he asked. “Or movements by players? Is it going to absorb commentary that it hears? How can it update its analysis in real-time mid-match?”
Bethuin also wondered how operators would be able to set different odds from one another, if a purely quantitative analysis yields perfect odds for everything? “If human intuition and instinct no longer has a place in sports odds,” he said, “you might as well shut up shop.”.
Global Gaming Group Head of Sportsbook Christopher Naudi was in the green zone, but nevertheless had some questions. If you are feeding the machine data, he wondered, at what stage does data become irrelevant? AI works without human intervention, it is spitting out results and answers that nobody “understands” in terms of an explanation of how we got to that solution. Since the neural network is, essentially, a black box, humans can’t really analyse how it reaches its conclusions; we can only test and confirm their accuracy. “Above all,” Naudi said, “this makes me slightly hesitant to see how it could replace trade.”
Varum Sriram, Head of Data Science & Risk Management at SimpleBet, was firmly in AI’s corner, perhaps understandably. Sriram spoke in clear, strong terms about AI’s ability to reduce the human bias which exists in our thinking, and thus how AI can make optimal predictions by determining the correct weighting for different signals. When asked about the boundaries for how to define AI within the debate he spoke of a framework of statistical algorithms designed by “really smart humans who understand statistics.”
In play, he admitted, it might be difficult to account for all the different factors; you need a data scientist, he said, to really understand the right suite of data solutions to use. You might not have info that a player is tired, or that they slept badly that night, but you might be able to infer that from stats. If you can’t get your hands on all the data, there are techniques like regression, data forest technique, and others which allow you to infer the likelihood of factors like fatigue or poor performance in a specific game.
A lot of raw data is generated and collected today, and these data can be fed into a neural net to extract signals from. You do need a modern tech stack that can support the machine learning algorithm, and, as Sriram acknowledged, the stack must be fed quickly, or the human trader is still going to have an edge. But if you engineer the right signals, you can pass these through machine learning algorithms and learn how those signals can be weighed against one another, avoiding human bias to achieve more accuracy. Sriram drew significant parallels between the finance and the sports betting industry, saying that ten years ago traders were making qualitative decisions in finance. Now math PhDs are employed as quant traders and work with AI and machine learning.
“Humans,” Sriram said, “are still involved, though, at least.”
