Machine Learning In Sports Betting
Artificial intelligence has been the buzzphrase echoing throughout scientific communities for some time, but it has only recently stepped across the precipice into the wider, mainstream consciousness.
Throw the term ‘AI’ at anyone who grew up the 80s and 90s, and you’ll quickly stir up memories of a menacing, human-hunting Terminator, but the more realistic applications will come from harnessing a computer’s ability to learn, correct and evolve.
Welcome to the world of machine learning.
Artificial intelligence has actually been around for a long, long time. Whilst tools such as Siri, Amazon Alexa and even autonomous vacuum cleaners fall under the AI bracket, they have no capacity to carry out tasks beyond the remit of their static programming.
Machine learning is very much the next step towards a future form of general intelligence that could mimic human thinking, problem solving and creativity – but do so at an incredibly heightened speed.
What Has All Of This Got To Do With The Betting Industry?
Consider for a moment that electrical circuits can process information around one million times faster than biological circuits. This means that an advanced AI tipping system would be capable of carrying out more than 2,500 years of human deliberation and study, over a single race, football match or even political market, in just one day.
As well as assessing the facts on offer, machine learning is able to play out millions of simulations, collating the results and even amending its own source code in order to improve its future prediction efficiency.
It’s this incredible technology which in 2016 saw a Google-led project, AlphaGo, take on one of the most successful Go players of all time, Lee Sedol. For those unfamiliar with the game, Go is a highly complex board game which has been prominent mainly in Asian culture for thousands of years.
Whilst we’ve had computer technology that can take on the world’s best chess players since IBM’s Deep Blue machine defeated Grandmaster, Garry Kasparov, in 1997, Go represents a completely different challenge altogether. There are 2 x 10170 possible outcomes in any Go game – to put that into perspective, that’s significantly more permutations than there are atoms in the known universe.
[wbcr_php_snippet id=”16172″ title=”Three Top Tipsters”]
The team uploaded a huge number of top amateur moves into the AlphaGo system which then played itself millions of times over, correcting human mistakes as it went. It would then use this learned information to play moves based on the percentage chance of it leading to success.
In one of the most highly anticipated matches of all time, the AlphaGo team played Lee Sedol in a five-match series which was watched by almost 100 million people worldwide. Sedol was incredibly confident prior to the first match, predicting a comfortable 5-0 victory.
He lost 4-1.
Many claimed it impossible for a computer to defeat a human player in this field due to the perceived need for creativity – something which has until now baffled experts in both its definition and execution. The 37th move in game two completely redefined the battlefield.
Sedol took a moment to enjoy a cigarette in the midst of battle and, upon returning to the table just a few minutes later, was faced with a move that no human player has ever, or indeed would ever, have chosen to make. It was perfection, baffling the champion and resulting in a collective gasp from those across a continent which lives and breathes the game.
What Does This Mean For Sports Betting?
Predicting the future is something that should ironically be reserved for the machines we’re trying to create, but it’s clear to see that AI/GI and machine learning, in general, is developing at an exponential rate.
Applying this tech to the sports betting industry opens a plethora of opportunities for the future market of tipping, betting and bookmaking alike. Human tipsters will take into account a huge range of factors when assessing races. From wider factors such as the form of the horse, jockey and trainer through to the going, course and distance, there’s plenty that gets taken into account by all of Betting Gods tipsters.
However, imagine a piece of kit which can base these tips on every single factor of every single race since data collation began – and then some. Stride patterns, deep lineage information, data breakdowns of distance to the yard, jockey racing habits and even fence success can all be combined before millions of race simulations are carried out in minutes to produce a profitable percentage return.
It’s potentially a dream world for punters, but it’s wishful at best to assume that bookmakers won’t use this technology to offer more accurate odds that favour themselves. Like humans, horses aren’t machines and so it’s safe to say that they’ll never be a system which can predict the outcome every time, but it could certainly lead to favoured horses appearing at significantly reduced odds.
The free market will inevitably dictate the release of a number of competitor products when it comes to providing AI tech to the public. All will undoubtedly claim to offer the greatest returns, but the reality is that once we reach a stage of accurate predictions, you’ll see software updates on a daily basis that result in greater win percentages.
Horse racing markets seem the most likely entry point for AI tipping systems given the fact that horses are perhaps more consistent in their form and performances than human beings. But what of football and other human-centric sports where the variables are considerably more abstract? Could a machine predict a player having a rush of blood to the head and getting sent off? Could it envisage certain tactical and player selection changes that a manager might make in the run-up to a certain fixture? What of refereeing decisions which are often hotly debated after several replays without a concrete agreement amongst pundits?
It’s difficult to see an ultra-consistent machine being able to take into account human error and subjectivity, meaning that some sports will invariably produce a higher rate of AI predicted success than others.
What Technology Do We Already Have In Place?
Data systems in sports have been around for years – after all the traders within each bookmaker use the myriad of information in the public domain to formulate their odds in the first place. This same data is, of course, available to our own tipsters in this cat and mouse industry.
For example, OPTA is the dominant force in the football industry, a company which has stretched its reach into the hands of football managers around the globe. The data provided offers a foundation on which to base tactical decisions and even player transfers. Take a look at the film Moneyball for a perfect example of how this has been most prevalent in baseball.
Companies such as FlatStats and BetRadar are other sources of data operating across wider sports which also funnel data to punters, tipsters and bookmakers alike.
The Sports Analytics Machine (SAM) is perhaps the closest we have to an existing football results predictor, although it still falls somewhere short of profitability compared to the top professional tipsters. Created at Salford University, it takes into account a plethora of data, including player form, team form and home advantage whilst removing the human emotions of ‘gut instinct’ and personal bias. Using decades of data, for example, it has calculated that home teams score an average of 1.2 times the number of goals as away teams in top flight English football.
Championed by the BBC, it takes on former professional footballer and pundit, Mark Lawrenson, each week in a prediction competition and is by far the most successful of its type developed to date. Don’t get me wrong, you’ll still be at a loss when it comes to following its tips each week, but it’s effectively a mass data cruncher rather than the market-leading AI technology which can self-correct and learn over time. It ultimately relies on human data input and, whenever we as a species are involved, the door is always open to missed criteria and human inaccuracies.
But the big wigs at Salford University aren’t the only ones on the trail of the perfect football prediction technology. They too are currently in the process of raising £25m to create a hedge fund that will offer ‘investors’ a viable alternative to volatile stock markets. According to the founder, Andreas Koukorinis, he sees a repeatable event with a fixed set of rules in place that he feels can be exploited to find shortfalls in bookmaker odds.
They’re the first company to begin talking about this predictor-style technology in terms of ROI as our own tipsters do at Betting Gods. However, whilst our network of tipsters work exclusively to assist punters with a proven investment strategy, Stratagem is also open to selling their data to bookmakers as well as professional gamblers.
Whilst this approach is perfectly in line with the free market of trade, it does raise concerns over a potential conflict of interest which begs the question – should these AI sports betting firms declare which side of the battlefield they’re going to work with? Will we see some businesses siding with the bookmakers and others with punters, creating a veritable arms race of AI in a quest for a perfect system?
When Can We Expect To See Mainstream AI Sports Betting?
How close this vision is to becoming reality is difficult to say. The AlphaGo team hit their target of defeating a world champion Go player an entire decade faster than analysts originally predicted. What’s more is that the stronger that AI technology becomes, the quicker and more efficiently it will be able to amend and improve itself, accelerating the rate of development even further.
At Betting Gods, we talk regularly about responsibly managing a betting bank, setting wagers that limit losses and provide a consistent stream of long-term profit. Gambling will very, very rarely be a ‘get rich quick scheme’ for anyone and should be treated as a long-term, hedge-fund style investment rather than a way to smash and grab some extra cash in an afternoon.
It’s for this reason that even these early days of AI advancement will worry bookmakers and should excite punters. Being able to increase your percentage odds of a win by even 5% can result in a big chance to your overall return on investment. We’ll, of course, be monitoring this over the coming months and years to ensure that our members continue to receive market-leading tips that offer reliable returns.
Either way, one thing is for sure – we will eventually reach a point whereby a system is created that drags the AI sports betting sector to a new level of accuracy and, even if for a brief moment, one individual, business or group will be in possession of a tool that gives them a markedly improved chance of getting one over on the bookies.
Similarly, if the bookies manage to get their hands on the technology before the punters do, it will be interesting to see with what transparency they operate. Will odds be superstitiously slashed for seemingly no reason based on AI predictions or will it be released as an open source product to maintain a level playing field?
Whichever side is first to make the all-important breakthrough, it remains to be seen whether it will be shared enthusiastically with those on the other side of the fence!