Sportradar has identified more than 11,000 suspicious matches worldwide over the past two decades through its combination of technology and human expertise. The global sports technology company, which provides data, content, and integrity services to sports leagues, media companies, and regulated gaming operators, continues to enhance its monitoring tools with artificial intelligence (AI).
Since 2018, Sportradar has integrated machine learning into its Universal Fraud Detection System (UFDS), which now relies heavily on AI to analyze and detect irregularities in sports competitions.
“We have been assisting sporting stakeholders to safeguard their events for over two decades by monitoring regulated operators’ data to uncover suspicious patterns in the global market,” said Andy Cunningham, SVP, Global Partnerships, Integrity Services, Sportradar, in an email interview with Back End News.
Through this monitoring system, Sportradar can analyze large datasets, identify regional trends, and compare matches to detect the methods used by match-fixing groups.
While fixers have traditionally targeted major markets such as match results or total points, Sportradar has observed a shift toward smaller, more specific wagers. These include bets on which player will receive a yellow card, how many corners will occur before halftime, or whether a break point will happen in a particular set.
“These bets are small and often spread across platforms to avoid detection,” Cunningham said. “But systems like UFDS AI can now spot them effectively thanks to advances in technology.”
UFDS AI processes hundreds of data points for each match and can review thousands of transactions per second. Its models monitor shifts in market behavior across multiple games at the same time, allowing the system to flag questionable activity for review by Sportradar’s analysts and integrity experts.
The platform also visualizes when irregular activity occurs and highlights data points that triggered alerts.
Cunningham said “probabilistic models” then translate these findings into a “suspiciousness” score.
“For instance, a score of 99 indicates the match looks more suspicious than 99% of previously reviewed events,” Cunningham said. “Beyond the score, the system provides clear explanations, flagging factors such as abnormal volumes, sharp price movements, activity from historically suspect accounts, or intelligence shared through the Sportradar Integrity Exchange, a collaborative information-sharing network for operators.”
Since its launch in 2005, Sportradar has built what it calls the world’s largest database of match-fixing intelligence and related data. This foundation has allowed the company to refine its machine learning algorithms to detect new forms of manipulation.
“Over the past eight years, Sportradar has leveraged this data to advance its AI technology, significantly strengthening the quality of monitoring and detection,” Cunningham said.
The UFDS AI system integrates multiple datasets into a single algorithm that evolves continuously. The growing number of suspicious matches detected since 2005 shows the tool’s increasing accuracy, the company claimed. It also allows Sportradar to report potential fixing methods to sports federations and authorities.
Detection speed has become another focus area. The company upgraded its system last year to provide faster, real-time insights. The improvement gives analysts immediate access to AI-generated data when matches are flagged as suspicious, helping streamline operations and improve accuracy in assessment alerts.
“With comprehensive visibility across global markets, including a wide range of derivative markets through Sportradar’s extended operator network, we remain agile in uncovering new manipulation methods as fixers adapt and diversify their strategies,” Cunningham said.
Sportradar’s UFDS AI combines technology and human analysis in what the company describes as a two-tier approach. Algorithms first process vast amounts of data and generate alerts when unusual activity exceeds set thresholds, which are tailored to each competition. These thresholds are based on factors such as market liquidity, historical price movements, and other data insights.
“Despite these technological advancements, UFDS AI maintains the critical ‘human in the loop’ approach,” Cunningham said. “We’ve seen cases where legitimate patterns, perhaps driven by team injuries or tactical changes, can appear suspicious to algorithms. That’s why humans make the final decision, and we do not apply completely autonomous monitoring in our process.”
Analysts use statistical, sports, and investigative expertise to interpret findings and confirm whether a flagged case truly warrants attention.
“Explainability” is also important to this process, according to the company. Sportradar said it ensures that each AI-generated finding can be traced, verified, and clearly presented to regulators or sports bodies.
“Stakeholders must clearly see why a match was flagged: what data points drove the alert, when patterns emerged, and how they compare to historical cases,” Cunningham said. “Without this, results risk being dismissed as arbitrary.”
To maintain transparency, every system action and analyst decision is logged and reviewable. Independent audits are also conducted to ensure the system’s credibility, particularly when findings are used in legal or regulatory cases.
“AI must remain a decision-support tool: it flags anomalies, but human experts review and escalate,” Cunningham said. “Black box models, even if potentially more accurate, are not favorable in integrity investigations where due process is critical.”
Sportradar said it uses AI insights to help sports organizations prevent manipulation before it happens. The company provides partners with detailed analyses of match-fixing trends, high-risk regions, and the specific tactics used by fixers.
“This intelligence enables federations and regulators to take proactive measures rather than reacting only after incidents occur,” Cunningham said. “For example, they can allocate education resources to high-risk areas.”
You must be logged in to post a comment.