PIN-UP GLOBAL Holding creates its own Machine Learning and Artificial Intelligence models to spot fraudulent behaviour and address it at the right time. Volodymyr Todurov, the Chief Analytics Officer at PIN-UP.TECH, shared some insights into how the team handles AI and creates game-changing safety solutions in his interview – https://next.io/news/promoted/pin-up-global-ml-and-ai-combating-fraud/. We, in turn, are going to focus on the key points of his conversations about improvement and innovations in the anti-fraud domain.
Key Principles of Risk Management according to Volodymyr Todurov
First, Volodymyr Todurov notes that in sports betting, risk management is mostly about setting the right individual limits. They depend on multiple factors, such as the type of sport, the tournament, specific events, and betting markets. That’s a hierarchical system with its own rules.
Generally speaking, there are two main types of betting limits: the maximum bet for a customer and the time delay for accepting their bet. This includes managing the sportsbook’s margins (profitability).
There’s also the customer hierarchy. Segments like beginners, regulars, VIP customers, and high-risk players also have personalized limits, which can be changed during different marketing campaigns. The system is flexible and dynamic, which allows the operator to balance risks and profitability while meeting the demands of all customer groups.
Classic Gambling Antifraud — Volodymyr Todurov’s Opinion on the Pre-AI Solutions
As noted by Volodymyr Todurov, any antifraud system, even a classic one, is based on a simple principle. You first determine the norm and then spot anomalies — anything significantly different from the norm you set.
There are many examples of anomalies in iGaming and sports betting specifically. For instance, you may spot a sudden change in betting volume for a certain tournament, a serious and non-typical imbalance between pre-match and live bets in one or some of the above-listed customer segments, a spike in average bet size, or abnormal growth of maximum bets placed by players. If you notice any of these anomalies, there’s a need for a deeper analysis of the segment where something went wrong or not as initially expected.
For PIN-UP GLOBAL, Volodymyr Todurov says that analytics of user behavior and preferences are another top priority. You should consider how much the customer usually bets, at what time they place their bets, and which markets they usually choose. Such analysis allows identifying clients who exploit errors in odds or use the bookmaker for other types of fraudulent schemes, such as financial or casino fraud, bonus abuse, and so on.
Lastly, arbitrageurs may use software solutions to practice arbitrage betting, which is actually one of the most common types of fraudulent betting. It’s simple and can be automated. However, there’s something that arbitrageurs usually don’t consider. A bookmaker is also using the software to detect such arbitrage opportunities, identify specific odds that can be abused, and take action against customers who focus on these odds exclusively within a certain time.
Basically, all these three aspects are the key elements of classic betting antifraud, Volodymyr Todurov says.
Benefits of AI: Taking iGaming Safety to a New Level
When asked about how Artificial Intelligence may boost the classic anti-fraud approach, Volodymyr Todurov emphasizes that ML and AI models are more than just helpful when dealing with huge volumes of data. You can generalize information more easily and spot specific customer behavioural patterns. You can’t do it with older methods, such as linear rules or manual transaction analysis.
All in all, AI is an additional, ultra-effective tool that helps reduce errors, allows the team to make decisions much faster in every single case, and gives you a lot of new information about the customers.
The differences between human and AI decision-making patterns aren’t a bad thing. By contrast, they deepen our expertise and help us see wider contexts, Todurov notes. AI is already capable of completing certain tasks as an autonomous decision-maker.
Simply put, with AI, you can increase business volumes without scaling the anti-fraud and risk-management team. It also reduces routine and provides a unique opportunity to focus on more creative tasks and a deeper analysis of other important processes instead of doing more or less standardized, boring tasks.
Antifraud Tools by Volodymyr Todurov & the Team: How Do They Work
When asked about the specific AI model used by the team, Volodymyr Todurov says their anti-fraud AI-based system processes a lot of data focusing on multiple parameters for every single transaction. The most important criteria are the sizes of bets made, their frequency, preferred events, and the overall history of the actions taken. The AI model analyzes it all, identifies the parameters that have the biggest predictive potential and uses all that data to make ultimately accurate forecasts.
Volodymyr Todurov also highlights the importance of using the extreme gradient boosting model, which is currently the best AI model for structured data. This method allows for creating a set of models that progressively upgrade each other, boosting the accuracy of predictions.
Basically, this model takes a bet history as input and outputs the probability, ranging from 0 to 1, as an indicator of the likelihood that a particular user is engaged in a fraudulent activity.
Workflow for Creating Top Anti-Fraud Solutions
As noted by Volodymyr Todurov, analytics, or rather its accuracy, will largely depend on the quality of the training data. The better the data, the better the results of the AI model’s work.
As a chief analytics officer at PIN-UP GLOBAL, he made a decision to hire an employee who has a lot of experience and expertise in sportsbook risk management. He was responsible for validating the chosen segments and evaluating their predictability potential for identifying fraud behaviour.
For over three months, his main responsibility was carefully reviewing a large amount of de-identified data from hundreds of users’ accounts and choosing the information that would be suitable for training the AI model. This was a significant boost for us at the initial stage of our work. Due to such in-depth analysis, even the first proof-of-concept models demonstrated a reliable balance of type 1 and type 2 errors, which is crucial for operational use.
Then, they focused entirely on onboarding Data Science experts who would use these models. It was essential that everyone in the team fully understood all the main principles of its work and used the right terminology to increase the overall effectiveness of data processing.
Todurov notes that their team trained the custom AI models through multiple iterations, testing them on data from various periods of time. This ongoing process greatly improved user segmentation and overall data quality for the next versions of the AI model.
Future of PIN UP Anti-Fraud Platform According to Volodymyr Todurov
According to Volodymyr Todurov, PIN-UP and his team, in particular, plan to create a decision-making model based on generative AI and neural networks. This will help save a lot of time spent on decision-making, boost accuracy, and allow us to identify the customer’s fraudulent patterns sooner and more accurately.
The best thing about such a system is that it doesn’t stand still. It’s learning and adapting. Moreover, it’s capable of studying the history of every customer in detail, applying multiple different scenarios to identify groups with specific behavior, including fraudulent behavior.
All this allows for achieving a new level of accuracy and reducing the manual work of the entire anti-fraud team.
Large businesses may see this technology as an opportunity to do retrospective cross-checks, see whether you labeled a customer as good or bad correctly, and reduce the burden on managers who won’t have to worry about whether they missed a dangerous user or not.
Small businesses may find it a great additional tool, too. It offers many opportunities for optimization, so you don’t have to hire a large team of analysts to manage all the processes effectively. Artificial Intelligence will do manual work so that the department may spend more time on more important and unique tasks.
***
Volodymyr Todurov is now the chief analytics officer at PIN-UP GLOBAL. He has over 8 years of experience in the field, holding the positions of bookmaker and risk manager in gambling and contributing greatly to the safety of different software solutions, particularly sports betting solutions.
He also specializes in risk management in such fields as casinos, affiliates, marketing, and payment systems. Volodymyr Todurov’s unique approach is based on decision-making, economic modeling, and giving more and more accurate predictions regarding potential risks.
