Pulse
Nothing’s new under the sun in match-fixing
By Lewis - 17 January 2025
As we head into 2025, the integrity space in sports betting is poised to benefit from continued advances in AI and data analytics, making detection and prevention of match-fixing more efficient than ever. As the data ecosystem continues to grow and technology evolves, Sportradar’s approach will adapt explains Jack Kennedy, VP Anti-Match-Fixing, with a focus on not only improving betting market monitoring but also expanding its role in supporting intelligence gathering and enhancing regulatory compliance across the industry.
How is the integrity space changing as we head into 2025?
The world of match-fixing is constantly evolving. The challenge is to stay on top of the latest trends to ensure that, as a technology company, we are continually moving forward and have the correct data infrastructure and resources to keep up with the pace of advancement. Nothing’s new under the sun in match-fixing. We often see the same characters popping up throughout the years, but adapting to changes in the landscape is always important. As we’ll likely touch on later, there is an increased reliance on betting data from operators, so we need to continue building out those relationships and the information flow is key. The biggest challenge is that, because of the confidentiality of ongoing investigations, sometimes it’s difficult for operators to get feedback from sporting federations and that creates difficulties. These cases take time, they’re complex and you’ve got to ensure operators are understanding of the fact that their information being used to benefit these cases is behind closed doors for positive change.
What’s the ratio of suspicious activity detection that’s now analysed by AI as opposed to human oversight?
It’s not so much that these matches are detected by AI and these matches are detected by humans. The real reason that the technology is in place is for the AI component to support human analysis. We monitor hundreds of thousands of matches per year and the AI model is there to separate the wheat from the chaff as it were and identifies the small subset of potentially suspicious matches that require further human analysis. Our Universal Fraud Detection System (UFDS), a system we built at Sportradar, processes data taken from a wide range of operators, from major to local operators. The system ingests all the betting data, and the AI model is there to identify suspicious betting activities – be it pre-match markets with a large odds change, increased turnover on a certain betting market, or in live betting markets, increased turnover or odds deviation from expected levels. The AI and alerting system are there to essentially flag out matches that require further human analysis.
The AI model has been a terrific addition it’s constantly learning, improving and ingesting data from multiple sources. It does things that humans can do at a much greater speed, but there is always going to be a need for human analysis. Strong betting doesn’t always point to suspicious betting. There are various reasons for strong betting. For example, as we came out of the pandemic and live sports resumed there were situations where large swathes of players would be ruled out. There are other examples when a team has been spotted travelling to a match at the airport and several star players aren’t travelling. That have even been situations where a team has started 10 players and been forced to play a second goalkeeper outfield. These are all instances that require human analysis to differentiate between what is legitimate betting and what is suspicious betting.
Does every global sports activity receive the same level of scrutiny, or is the computational power of the UFDS more targeted than that?
One of the greatest things about the machine learning model and the increased use of AI in our monitoring is the ability to monitor in a horses for courses manner. Each sport has its own betting dynamics. Different markets are popular in football and basketball, volleyball and tennis. The AI model is there to help ensure each sport is monitored in a way that’s related to how it’s previously been manipulated because the machine learning model is continually learning from our human assessment. That’s the dataset it works with so it’s able to make sure that each sport is being judged based on the individual risk factors that have been seen in the past. That is one of the biggest advantages in the model – that it’s able to differentiate between different sports – and one of the great benefits of Sportradar acting as a data provider and ecosystem means we have access to large data sources on a broad range of sports. This gives us the ability to look at each sport in an individual manner.
We have teams of analysts dedicated to certain sports. To address your question, yes, each sport gets its own kind of attention but because it learns from human behaviour and our previous classifications of matches. For sports like football and basketball where there has been a higher number of suspicious matches recorded previously, naturally the system is better equipped because it’s got a greater pool of data to work with.
You’ve spoken on Sportradar’s role in determining and then validating legitimate betting from suspicious betting. What about the enforcement action?
There are three different areas Sportradar supports. The first is prevention and we have well-equipped integrity education and prevention teams. These are people who’ve come from an education background, and they’ve got tailored material to help players, referees or any stakeholders in the sport recognise match-fixing approaches and how to report it which is often not as easy as it sounds. Then there is the detection component, and this is where the UFDS comes in. When the system flags up strong betting, an analyst will conduct an initial review. If they can’t see any immediate factors to determine why the strong getting activity is in place, we’ll do further research, and the data gets validated by a team within Sportradar. This is to ensure that the goal times are accurate, the system has correctly recorded any red cards, and so on. Then a request is sent out to a local expert with a list of questions that gives our team of analysts all available information. For instance, were there any local rumours? Was there any extra team news before the match? Often there is stuff talked about on local radio that is difficult for analysts to find from the desk. This next set of questions is run through and once the data has been validated and checked, they’ll collectively come to a decision about whether the betting activity is explainable by reasonable factors.
We have a high threshold, so we look at every betting activity in multiple ways to ensure that there is no legitimate reason for the betting activity. Once that threshold is hit a report is sent to the relevant sporting federation and outlining the betting markets targeted, why we determine the betting activity to be irregular and various other different factors. It is also listed in a data appendix – things like the league table, a glossary of betting terms and any previous suspicions of the teams or players involved. This report contains information for federations to undertake their own action and the approach here has been validated by the Court of Arbitration for Sport. Then I come onto the third pillar – action. Within our anti-match-fixing team, we have a dedicated team of intelligence analysts and investigators who come from a variety of career backgrounds. For example, the head of Sportradar’s integrity operations worked in Navy intelligence for 10 years. This team is set-up so that if the sporting federation doesn’t have an infrastructure to investigate the match-fixing activity themselves and isn’t able to add more context onto the yes/no of the bet monitoring report, our team is there to support. We’ve got dedicated intelligence holdings, relationships with law enforcement partners, data systems we can use to help build a picture for the federation and ultimately help them build towards sanctions because that is the biggest deterrent.
I like to think we’ve got it all covered so that even partners with the smallest level of resourcing can call on Sportradar’s expertise and we can cover as much for them as they’d like us to do. Alternatively, if they want to keep things in-house, we can just provide complementary assistance.
How often do these cases lead to sanctions?
The work that we’ve done in Integrity services has resulted in 900 sporting sanctions and 73 criminal convictions. We can’t jeopardise any ongoing investigations, but some previous and public examples include the huge support we provided for UEFA in the match-fixing case involving Albanian club KF Skënderbeu. We provided a significant amount of intelligence support and, if you read the case, you’ll see that our reports are used. The club were banned from European competition for 10 years and fined one million euros. We were also heavily involved in a FIFA case with Ghanaian referee, Joseph Lamptey. The first identification came from our team, and we were involved in the intelligence sides on that too. I could reel off many more.
How do you balance the objectives of the operators with those of the sporting organisations, state authorities, national platforms and law enforcement agencies?
Thankfully, they’re all aligned insofar as match-fixing is a negative for all these different bodies – betting operators, sporting federations, law enforcement. That’s a positive – everybody wants to pull in the same direction. Because of Sportradar’s position we’re very close to betting operators and sporting federations and we recognise that makes us unique in these circumstances. It’s why Sportradar is the best in integrity matters because of the company’s position in the data ecosystem. As a recognition of that, a couple of years ago we set up the Sportradar Integrity Exchange, a free initiative for betting operators to join that now has over 100 in that network. Essentially, it’s a network of information sharing on potential suspicious betting. It gives operators an outlet to Sportradar and other operators, as well as an inlet to sporting federations. Data from betting operators is fundamental in sporting and criminal investigations. Sportradar is at the centre of that and we’re facilitating this connection because many sporting federations don’t know where to begin getting information from betting operators, so this integrity exchange is a huge part of that. Of course, we always respect information sharing confidentialities.
How will Sportradar improve its integrity services offering in 2025?
We are going to look at product development and how we use AI beyond the bet monitoring space. How can we continue to use the latest technologies to help improve results to our partners, particularly in the intelligence space? That is not to say that AI will be used to perform any of our intelligence and analysis, but it will help us to digest different sources in a quicker time frame. We have accomplished this on the bet monitoring side, and we’ll turn our attention a little more to the intelligence aspect. Across Sportradar, we will continue strengthening the support we provide to regulators, including in the responsible gaming space, which is just one element of the regulatory landscape where we are using technology to solve challenges. Within anti-match-fixing and beyond there are various targets, but for me it’s all about continually adapting and engaging with betting operators. We want to bring operators deeper into the ecosystem and create more direct conversations between betting operators and sporting federations. The support that we can provide in the anti-match-fixing space is largely given to sporting federations, but we’ll examine what we can further provide for betting operators as we look to build on those relationships.