“The Brain,” computing and quantifying its quarter of a million data points, had pegged the Cincinnati Bengals as likely winners of the AFC Championship Game. Betting lines had the Kansas City Chiefs as consensus 1.5-point favorites at kickoff, however.
The Artificial Intelligence-driven, neural-network-utilizing, deep-learning-empowered predictive algorithm developed by Toronto-based Quarter4 ultimately had it wrong. But it started to get it right with in-game assessments of an air-tight game that had the Bengals tied for much of the fourth quarter. Harrison Butker kicked the winning field goal with three seconds left to send Kansas City to Super Bowl 57.
Most importantly, “The Brain” started to get it right early, with micro-bettors beginning to swing to the Bengals as quarterback Patrick Mahomes appeared to be running out of pain tolerance and healthy receivers. For bettors looking to exploit lines and the betting industry looking to write them, that matters. And being able to assess layers of information at high speed without the taint of gut feelings and emotion is crucial.
Such is the reason AI is inevitably beginning to leach into both sides of the sports betting equation.
Gaming Today spoke with Quarter4 founders Kelly Brooks and Danijela Covic about how their four-year-old company fits in the suddenly AI-conscious realm, and how the “The Brain” keeps up with all its new competitors, including ChatGPT.
“We did have the Bengals, which is against the market. But the reason we had the Bengals was obviously their historic meetings, which we know have been a lot more successful for the Bengals, which is a really relevant question because of the injuries.
“[Injured Chiefs tight end Travis] Kelce was a point of contention. There’s a lot going on with Mahomes. But we only had the Bengals at a 1% differential.
“Within the first half, we changed it to 57%, the Chiefs and then we went 67, 75 and 80%, which is kind of crazy considering the fact it was only a three-point game, that was a close game always.”
“We process every one-to-two seconds and then Brain is now learning every one-to-two seconds. So what happened with the Chiefs and the Bengals game, we originally were calling the Bengals to win at the beginning of quarter one. So the AI is constantly running. It’s constantly running during every one-to-three seconds of every game that’s playing wide. That’s how powerful the system is.
“So as the game was going through, we also pick up when a player is out, when an injury happens. So the market started feeling that Kansas City was going to shift. The market was going to shift. The bettors were going to change their bets in-game. The markets actually started really showing it around the end of quarter two, quarter three.
But we started really understanding toward the beginning of quarter two, before halftime, that there was going to be a shift. The machine started understanding that. So that’s very powerful, especially when you talk about either giving a bettor or a book a leg up. That in-game is the most powerful piece. But a human cannot do it at that level of depth and speed.
What’s the State of Artificial Intelligence in Sports Betting?
“The sports betting world is all about analysis and marketing and everything else. So when we’re talking about Quarter4, for example, and we’re talking about all of the modeling that we are doing, we’re looking at the whole picture, we’re looking at the whole season, we’re looking at the whole millisecond experience and game, and we’re looking at how that’s affecting the crowd, the player, the environment, the atmosphere.
“And we’re transferring that into data that’s readable by the human mind. But it’s being determined by an artificial intelligence that is processing that a million times faster than we can as a human being.
“[AI] has grown quite a bit, but I think what’s actually happened is it’s just becoming more noticed. It’s always been around, it’s just now more noticed. So people are like, ‘Oh, is this AI? Is that AI?’ And the answer, most likely, it is. It’s just now you actually care to know about it.
I think what’s actually happened is it’s just becoming more noticed. It’s always been around, it’s just now more noticed.
How does AI Resolve Factors That Defy Pattern-Finding?
AI excels at identifying patterns quickly. But what if Team A has beaten Team B 10 straight times, but Team A is entirely different (and terrible) in the eleventh meeting?
“I think us as humans, that’s where our bias would come in. ‘Well, OK, they’ve met 10 times and there’s obviously a massive overhaul here. There should be a no-brainer selection. Right?’
“The difference between what AI does, it essentially assigns a confidence to every single player, to every single stat, to the location and the travel. We consider elevation of when one is playing, how far they’ve traveled, who did they just play, are they on a hot streak or are they in a losing streak and why are they doing this?
“Is it the teams? Is it the players? And then it looks at all of those combinations and it does take into consideration, OK, the last 10 meetings, Team A was a lot more successful than Team B, but what was the reasoning behind that?
“Were the players all the same, who was injured, who wasn’t injured? Where were they playing? When were they playing? And then what was the pattern behind that?
“So it looks at all of their historic matchups, but it also looks at the current players there, their seasons. And it really takes the whole season into consideration and beyond their careers getting up to that season to make a decision for that singular game.”
How Does the Brain Assess In-Game Elements?
“It’s automatic. It’s called feature engineering and basically, there’s weight adjustments too. What the engineers are constantly doing is understanding what is relevant and what is not.
“The actual artificial intelligence will determine what it thinks is relevant or not. So, if player number 6 broke their leg, is it relevant, right? Because based on past performances that player has had against the other team, or with the current roster against the opposing roster, what is the actual impact? The machine might spit it out and say, ‘doesn’t matter.’
“We have proprietary technology called the injury effect. It’s basically when a player is injured, how much impact does that player have on every scenario within the game? We actually run that. There are times when it doesn’t have an impact at all, and there are times when it impacts it by a certain percentage point.
“There are times when there’s headlining players that really do impact that probability by 3-5% or even greater.
“So again, the machine determines how relevant is based on all of those, those previous patterns and predictable patterns.”
The Brain keeps learning
A few years ago, Brooks and Covic described The Brain as a teenager. So how old is it now?
“I think she’s an oldie in the NBA. I think in the NBA she’s about 42, and she’s a very seasoned corporate executive.”
“She’s in her 30s, for sure, for hockey and football, if not even older, I think probably. But I think in WNBA she’s about to hit 19. So we’re thinking this season she’s going to be hitting her twenties. She’s getting real quick.”