A Q&A with BetBuddy’s Simon Dragicevic

Go back

BetBuddy announced itself to the world back in late 2014 when it announced it had received funding from Innovate UK to further develop its AI-based socially-responsible gambling tools and it would be working alongside the Research Centre for Machine Learning at London’s City University. Since then it has announced partnerships with Ladbrokes and the Ontario Lottery Group (OLG). Here founder Simo Dragicevic talks through the company’s tech difference and its work at the cutting edge of machine learning and AI.

GamCrowd: Could you describe how your early-warning system works and the technology that lays behind it?
Simo Dragicevic: Our regtech platform is based on our deep research understanding of the pathways towards gambling harm. This is undertaken using a three-tier approach: the exhibited behaviour using data collected directly from player account management system, the second tier looks at declared behaviour, or the results of self-assessment test, and the third tier revolves around inferred behaviour, which is deduced by combining results from the first two to create predictive models based on actual data gleaned from problem gamblers.

GamCrowd: Could you go into detail about what Power Crunch is?
Simo Dragicevic: The core of what we offer is a single platform that embeds our domain expertise in understanding risk across any form of gambling, whether land or online, and across all gaming verticals.  This enables any operator to quickly and accurately classify consumer risk. The platform is built around three key algorithmic services: behavioural risk calculation, prediction and classification, and model interpretation. Wrapped around this core algorithmic logic are an operator front-end to view player risk profiles, a data science back-end that enables model configuration and tuning, and an API to enable an operator to access personalized player messages that can be integrated to any customer interface or channel.

GamCrowd: Do you think operators are aware of how much they might gain from fully understanding the data they see in relation to their customers?
Simo Dragicevic: Operators understand there is incredible value in understanding data, however most still see this from customer acquisition and marketing perspective. The UK Gambling Commission’s consumer-first approach is now placing greater emphasis on using data to improve consumer protection and compliance processes. Using data to inform consumer protection in this context has three key elements: first, risk assessment, second, customer interaction and third, evaluation. Whilst most operators are still grappling with how to assess risk, we think their main priority should be in implementing robust and repeatable experiments to test which responsible gambling interventions work best, just like they would with A/B testing new customer experience and product features. The industry is currently going through a stage of trying to understand how to best build player risk algorithmic solutions internally, a trend we have seen in the financial services industry many times over in the past two decades following continually increasing compliance demands. Most internal systems eventually become slow, difficult to adapt to changes - whether business strategy, consumer consumption habits or regulatory, and costly to maintain, and are ultimately replaced. As new regulations become part of business as usual, companies typically look to find a class-leading solution which continues to be improved by specialists to ensure it remains relevant. Their focus moves away from internal building and more to generating value-add to gain competitive advantage, whether that be improved customer experience and retention through better use of insights or though operational efficiencies that technology and AI can bring to solving these problems. Large and mature companies often struggle to keep pace with the change and opportunities that technology and AI is bringing, and most internal corporate technology and innovation centres rarely replicate the technical creativity that start-ups bring. Companies like BetBuddy are essential to industry if they are to trial, test, and ultimately adopt emerging technologies. Successfully commercialising R&D in high risk technologies brings tremendous benefits to larger operators and the industry as a whole and plays a critical role in de-risking innovation.

GamCrowd: What help have you received form the Machine Learning Group at City University?
Simo Dragicevic: Our collaboration with City University’s Professor Artur Garcez, who heads its Research Centre for Machine Learning (http://www.city.ac.uk/machine-learning), is incredibly important in helping to continually improve our product. For example, whilst we had developed a robust application that adopted mature machine learning technology, feedback from industry was that uptake of machine learning solutions in risk would be inhibited unless the system could explain to the end user the rationale for a classification. This is an age-old problem in AI and machine learning - the technology has tremendous benefits in solving complex, non-linear problems using very large data sets, but the structure of models, whether they be random forests, neural nets, or Bayesian nets, are incredibly difficult for humans to understand. Garcez and his research team have decades of unique experience in knowledge extraction, a research field of AI aimed at making black-box algorithms more human like or explainable. We've applied these techniques to our product, which now gives operators greater transparency and greater ability to test and use our algorithmic outputs. Garcez and his team are also helping us to apply new AI techniques, specifically around deep learning, to develop new applications. For example, improving anti-money laundering (AML) processes is a major challenge for the online industry to solve. Whilst traditional rules-based systems and supervised learning models may be effective at profiling known money-laundering schemes, they are less so at detecting new ones. Here, combining our expertise in traditional knowledge-based approaches together with City's expertise in emerging online learning techniques, such as deep recurrent neural networks, will help to tackle the dynamic nature of AML more effectively.

GamCrowd: What are the questions you most often face when speaking to gambling companies?
Simo Dragicevic: The process is very consultative as it's normal for operators to want to gain trust in the supplier brand, solution, and team's expertise prior to trialling or using the solution. We find being a specialist in responsible gambling and data is important in building trust in this domain. Operators are naturally interested in how the solution works. Common questions are centred around the behaviours that are analysed and the variables we generate from our algorithms and feed into various models. The benchmarks we use to model harm is a common question, given there is a sparsity of good 'ground truth' data. Medium and larger operators are very keen to understand how the platform can be tuned and configured to provide accurate classifications for different parts of their businesses e.g., retail, online, casino, sportsbook, etc. Integration of data is a common question, but this rarely causes issues given that all industry player-account management systems capture the core data we use. Operators also increasingly keen to understand how our machine learning models can be interpreted effectively and used to interact with the customer. They are also interested to understand our experiences with our current customers.