IN-DEPTH 5 January 2018
Sports trading and AI: Taking the human out of sports betting
With Artificial Intelligence now playing an ever-increasing role in the determination of odds all across the sports trading world, what future is there for the human sports trader?
By Gambling Insider
Greater technological integration into betting has revolutionised the sports betting market, giving individuals access to millions of new betting markets and presenting bookmakers with an opportunity to grow revenue and custom in an economically efficient way.
Online platforms enable gamblers to bet whenever and wherever they want, reducing the need for the traditional bookmaking shop. However, it has proven to be a double-edged sword, with the increasing number of available betting markets growing beyond the capacity of human odds makers to efficiently manage them.
Tom Daniel, Head of Risk and Player Analytics for sportsbook provider Kambi discusses the changing face of the industry: “The scale of available sports betting markets is so vast in comparison to 13 years ago when I joined the industry. Everything was priced manually, and a single trader working on one event might price up to ten markets, but with everything being done manually they spent more time changing their odds than actually watching the events.
“It is very different now of course; you have a single trader managing the output for scores of events offering hundreds of markets, with systems that require no direct input from them whatsoever so we’ve definitely come a long way.”
Daniel is talking about the second technical revolution in sports betting which is currently in its infancy; the increasing use of artificial intelligence and machine learning in odds making. Designed to increase the accuracy of odds calculation, efficiency and output, AI sports trading represents the next digital frontier in sports betting.
The basic theory is that greater automation leads a drive towards greater efficiency, however Christopher Langeland, Managing Director of BettingCloud explains: “In reality it will take some time and considerable skill to set up AI/automated systems to do a lot of the jobs that have traditionally been done by humans, so the most valuable people in a bookmaking organisation will become the analysts/programmers who set up this automation.”
Moving into unknown territory carries with it dangers to both the bookmaker utilising the technology and the consumer taking advantage of it. In this gold rush to get to the next stage of sports trading is the industry ignoring the inherent perils in favour of the profits?
Just as more scientific analysis of sport is changing how coaches, trainers and clubs play their respective games, greater analysis of sporting events is helping odds making database operators evaluate the potential permutations of each sporting event, increasing the accuracy of that respective odd and thereby making the subsequent odds determination easier.
Human traders simply cannot utilise the huge amount of sporting analysis data all of the time and if they did they would simply not be able to recall the data for use in odds determination in the same time as an AI could, especially given the increased number of betting markets being offered by bookmakers.
There is also a greater degree of accuracy to the odds calculations of AI based sport trader than a human sports trader, who will always be liable to human error regardless of their skill. An AI odds maker can make hundreds of thousands of calculations per second, leaving their human counterparts in the dust.
Connall McSorley, Commercial Director of Metric Gaming believes that artificial intelligence algorithmic based trading “is vital in order to run the quantity of markets required to be seen in a competitive benchmark perspective because quite frankly human sports traders cannot manage that type of output.”
AI in sports betting is still in its relative infancy, with programmers still developing the language that will ensure its security and accuracy. While they may be able to develop programming language and algorithms to replicate and exceed the knowledge of a human odds maker, AI still cannot currently replicate the experience of a human.
Kambi’s Head of Business Intelligence Daniel Tidström commented: “Coming from a data and algorithmic side, AI is becoming more and more of an accepted practice in the sports betting industry, however the Holy Grail of a fully autonomous general artificial intelligence sports betting system is still some years away.”
However, with the gap between the operational effectiveness of AI and human sports traders becoming smaller the role of human odds makers will inevitably have to change to a one of risk management ensuring that the system remains free of potential corruption and that problem gamblers are identified/stopped, all things that AI systems cannot currently do without
McSorley believes that the role of human sports traders will “morph more into one of risk management and price biasing, using AI algorithms as a tool in this role.”
Aside from setting odds, AI can make significant contributions to player analysis and risk management. Kambi has been involved in using AI in this way for some time, as Daniel explains: “We’ve developed fully automated player management systems that actively manage the profile of every player on our network, perhaps hundreds of thousands of players. This gives us the means to manage players in a way that a human operator could not in terms of speed and coverage.
“We’ve been doing a lot of work aimed at removing the human element from player profiling in particular, however we do still retain an element of input from human risk assessors. The sheer number of players currently participating in sports betting necessitates the need for practical ways to assess every player on the network, which is only possible with a high degree of automation.”
The world of odds compiling has increasingly moved away from an exercise where an odds-maker starts with a blank sheet of paper and compiles prices/lines based on his own subjective skill and experience, and moved towards a role where the skill is in interpreting the market and positioning your firm within the boundaries that the market provides you.
Langeland believes that there is a future for human traders “in designing and pricing unique markets, and also pricing events and markets early before a strong market has formed. Sports traders should also be able to have an active and valuable role in designing and programming the AI trading systems, so at least in the short-term sports traders should see AI as an opportunity not a threat – albeit one that will eventually replace the traditional compiling/trading jobs they once did.” Placing AI odds makers in total charge of an operator’s books potentially exposes its financial revenue to attack or manipulation by hackers. Given the significant levels of revenue involved and the number of high profile corporate entities which have fallen victim to hackers over the last five years, betting operators may be reluctant to place the entirety of their odds making under AI control.
“Traders are constantly monitoring the output of what AI is generating and if there is seen to be a denial of service or a system failure the markets are just pulled and voided,” explains McSorley. “There is an absolute requirement for the sort of monitoring and management of what content is being produced, the accuracy and sanity checking against other operators to ensure that you are in line and not creating arbitrage opportunities.”
“There has always been a potential for hacking out there, with denial of service attacks being a feature of the industry for
quite some time,” warns Daniel. “B2B providers must have robust cyber security measures in place and we take security
“It’s important that traders are involved in the development of all our systems, spending as much time with our development guys, quant specialists and the people who build the algorithms as they do on trading. Traders need to be able to understand our systems and we are working to ensure that our systems do not become some great black box that no one understands and can therefore not detect when things go awry. We have always had redundancies and expert traders who can
step in should the need arise.
“It’s counterproductive to have a system with great security that no one understands and is vastly complicated; it is better to have experienced individuals operating in partnership with systems so that the provider can be prepared for any issues that arise.”
One area where people have AI beat hands-down is the ability to read subtle human behaviours of those involved in a sporting event. For example, no computer could understand the shock and resignation on David Luiz’ face when Brazil were five-nil down at half time in the world cup semi-final, or see the pained body language of Rafael Nadal when he injured his back in the Australian Open final but carried on playing.
“We have human traders with decades of experience who are experts at reading body language,” explains Daniel. “The statistical modelling and algorithms simply do not exist for these sorts of situations, so there will always be a requirement for an expert human trader who can read the body language of the participant and interpret the emotional aspect of the event.”
Connall McSorley agrees: “Human sports traders will never be replaced by AI, the overview component of their offering is absolutely huge, for example is a player crocked but has he stayed on the pitch because he is the team captain is just one example of all sorts of nuances and sentiment in sport that an algorithm will never get. Human sports traders are also able to offer bettors a company’s position from regionality or marketing perspective or a competitive one, all an algorithm can do is offer a price.”
“In betting there is a big incentive for people to look to manipulate the market with disinformation,” Langeland explains, “so an AI social media monitor would need to learn which sources to trust, as well as how to weigh what is said. When police detectives can be replaced by AI we will know that AI bookmaker traders are ready to take over from human counterparts!” Daniel feels there is “a definite need for a human component, you simply cannot remove it, however what you need to do is to free up human sports traders to watch and interpret the sporting events rather than be constantly changing the prices and doing the manual tedious work that machines can be doing, instead allowing the human trader to concentrate on the things that machines cannot do. A modern trader requires a more diverse skillset than ever, he must speak a common language with developers, understand and help refine the mathematical models and have a fundamental understanding of the mechanics of both sporting events and betting markets. You need to create a pipeline to develop talent with the right experience. Maths alone will only take you so far, you need experts who know when correlation is not causation”
If the technology behind AI sports trading becomes more robust and the scope for its usage in betting grows then this may lead to a symbiotic relationship between the human sports trader and the AI trader. Betting operators could look to retain the services of human sports traders for higher profile sporting events and ultimately as a checking mechanism on the work of their AI counterparts.
One thing remains certain: the sports betting industry has accepted artificial intelligence algorithmic sports trading into the fold, but the nature of the relationship between human sports trader and AI algorithm will be an ever evolving one, with both seemingly being required to be of benefit to the industry.