Industry predictions: Using big data to tackle fraud and the rise of computer vision

Sergey Portnov, CEO, Parimatch

Esports to be in top five categories by revenue

Over the past 12 months, esports has not only entered the mainstream, but become one of the fastest growing sports in the world. With live sporting events cancelled due to the Covid-19 pandemic, esports tournaments have given online betting customers a new and exciting alternative. With major tournaments like Dota 2 attracting hundreds of players and generating millions of views on platforms like Twitch, the success of esports will undoubtedly outlive the difficulties of the pandemic.

The integration of bookmakers’ presence into esports, and more generally sport events, has become more evident over the past year. Online betting companies which support esports and sports teams through sponsorship, brand ambassador deals and joint marketing campaigns have gained great exposure. This year, we will see more bookmakers investing in the creation of their own events and tournaments, which will appeal to fans regardless of their team affiliation.

The rise of big data in the battle against fraud

2021 marks the start of a new era for tackling betting fraud. With fraudsters becoming smarter and better equipped with state-of-the-art software and robots, and the adoption of biometric authentication tools being held back due to privacy concerns, fraudsters can easily blend into a crowd of legitimate online users. They also take advantage of the fact that anti-detect browsers like Linken Sphere and mobile/residential proxies have never been so cheap and user-friendly.

Going forward, betting companies can no longer rely on industry-agnostic B2B anti-fraud solutions like ThreatMetrix. Instead, large online operators will try to tackle fraud issues through the adoption of sophisticated data analysis tools and application of domain-specific knowledge to algorithms, as well as ML and big data approaches, so that customer information is ring-fenced from malicious attack.

Greater investment in payment systems over gaming products

World-class gaming products are nothing without user-friendly and efficient payment systems. Customers now expect high levels of data privacy and seamless transactions which do not interfere with their online betting experience. Operators will increasingly be looking towards encryption and tokenisation as methods of securing payments, moving away from the traditional methods of credit card payments which are often targeted by fraudsters. As customer bases grow, more needs to be done to protect personally identifiable information – therefore investment in secure online payment systems is paramount.

Ben Burdsall, CTO, Sportradar

Increased use of computer vision

Computer vision (CV), a form of artificial intelligence (AI) that teaches models to interpret and understand the visual world, will significantly evolve the sports data and sports betting experience in 2021 and beyond. While we, as individuals, are very good at understanding what is happening in front of us at sporting events, we have limited capacity in the amount of data we are able to manually interpret, record digitally and upload. CV provides a 100-fold increase in the level of statistics and data-points that we can offer our clients, significantly scaling the commercial opportunities available to them. Deeper analysis: by calculating velocities and trajectories of players we could predict when players will, for example, become offside or predict when an attack is mounting.

Simply put, CV will provide coverage of more matches, increase the volume of data we are able to provide and reduce the time of data acquisition. For bookmakers, this will be of huge significance as they will be able to offer a greater number of new betting markets – in-point betting in tennis, for example, could be something we see. The richness of insights that CV creates will make sporting judgements less opinionated and more informed based on the data.

Additionally, we have identified how CV can improve the integrity of sports and alert to doping among athletes. Every athlete has a unique, signature style of play and certain biometric characteristics, which can be tracked over time. That signature will naturally evolve over their career, but unnatural changes can also be identified. If you suddenly see a marked change in performance on these mechanical characteristics, it gives you valuable insight into suspicious activity. A sudden change in human performance is consistent with use of steroids or other drugs.

Growing importance of supervised machine learning

From an integrity perspective, machine learning will enable us to scrutinise further how games are played out. Over the course of 2020, we trained algorithms to replicate the work of human operators through a process of supervised learning. This is achieved by processing all the historic data of games and presenting the outcome of what the human experts reported. Based on this line of data, we can conclude whether the game was fair or not, and the computer, through the use of Deep Neural Networks, has the power to find the intrinsic patterns in the data. Pattern identification emulates what experts call “years of experience” – where they cannot articulate why they believe something looks odds or normal, they just know.

While machine learning can cover more games in a fraction of the time it takes a person, the output has to be compared with human intelligence to provide greater accuracy. It is important to remember that we still need human operators as they generate that expert supervised set that trains the new machine learning networks.

Migration to the cloud

We will see more organisations move to public cloud-based infrastructure as it is the most effective way to achieve true data quality and quality of service. The timeliness of data has never been more crucial and being cloud-based will improve latency regardless of customer geographical location if operators take full advantage of decentralised system architectures. We anticipate as much as a six-fold decrease in the latency of live data delivered to our Asian and US customers early in the new year. Providing customers with access to the same data, at the same time creates a level playing field for all. Betting markets prosper when there is no unfair advantage.

The added advantage of cloud migration is the increased security it offers, given that a public cloud will have thousands of cyber engineers more than a company who may have only 20. The belief that there is any value in hosting your own systems and data for security, quality or availability will evaporate. Certain regulators need to re-examine their policies about stipulating hardware being on-premise.