Big claims are often made for the way in which the trading rooms of all the big bookmakers have become automated with algorithms now employed across every organisation to provide the odds, in particular with in-play.
But according to one company beavering away in the data analytics arena, we really haven’t seen anything yet. In betting, the rise of the machines is very real and it promises to transform the betting industry, or so says the founder of Stratagem, a big data company that is currently building a machine learning product which it claims will be able to consistently beat the markets
“Stratagem is a data science predictive analytics company the treats sportsbook trading in the same way as a financial trading product,” says Andreas Koukorinis. “We have made a big investment in machine learning and artificial intelligence. We have some relationships with leading universities and have been awarded funds through some government schemes.”
These institutions include Imperial College London, the University of Liverpool and the University of Bath. Moreover, the company is also already working with the betting industry–notables include betting exchange providers Matchbook and Betdaq and the bespoke betting broker Eastbridge.
Before we get to the implications of what Stratagem is promising, GamCrowd asked Koukorinis about the science behind the product.
He said there are two types of data involved in the Stratagem process. First comes the data you can receive on a daily basis from sportsbooks and from sports-data collection agencies.Then there is the unstructured data, from videos, which needs to be translated and eventually made synchronous with the betting lines.
“It’s a big task; you need to develop a common language,” he says. “That is step one. It involves a description of what happened during an event and a characterisation of what happened.”
At that point the betting predictions come into the process, both before and during a game. “We’re teaching machines to do that,” he adds.
The prediction engine will have to learn about the state of the game and understand the state of the betting markets at the same time. At that point, Stratagem hopes the machine will be able to decide what to do for itself.
“These are all established ideas in finance,” says Koukorinis. “I don’t know what it hasn’t been so prevalent in sports-betting. Maybe the data wasn’t easily available. But I do know that getting the data to talk to each other is really hard.”
What Stratagem has been able to achieve is to “teach” its machine on historical data. “It’s learnt offline. It’s like one of those robot vacuum cleaners, which learns by trial and error the routes around a room to get into all the corners so that after a while it knows the optimum pathway. That’s what we are doing right now. The next task is to get it to learn while in running.”
Koukorinis says it’s about “incremental gains.” It has taken Stratagem two-and- a-half years to get to this point and he says the ‘Holy Grail’ is still a while away just yet.
What is that Holy Grail? “A machine that that when a game is about to start, can assess how it is progressing from the real-time data feeding into the algorithm. Football is the hardest to model. Tennis is easier. Scaling is an issue”
In old-fashioned terms it sounds like a ‘bookie buster’ but Koukorinis and Stratagem insist the product - as and when it is made available - is aimed at both sides of the bookmaker/punter divide.
Stratagem claims that in any given betting market, its predictions can be up to 12% betting than the odds. “When there is an evens market, say 50/50, we currently are about 2.5-3% better and we are aiming to get 4x as much, which is 12%. The natural limit (theoretically) is about 76%.”
Despite what would seem obvious – that a machine-learning product with sufficient accuracy has the potential to put traditional bookmakers out of business – Koukorinis says he doesn’t
see Strategem’s position as being “adversarial” to the bookmakers. “We are making the market more efficient, giving a better understanding and a better performance.”
The company already has an advisory service available via Stratabet (http://www.stratabet.com), alongside its betting exchange partners.
“Our goal is to make the market more transparent,” he adds. “More knowledge and better understanding. It works in finance. Everybody gains. History shows that better predictions can reduce uncertainty. When people don’t have enough information, they are less involved. Hopefully our product leads to the betting markets to be more like the financial markets.”