University of Amsterdam researchers have developed a transparent machine-learning model that regulators can use to assess online gambling behaviour and compare operator risk-detection systems.
 

Key Points

The open-source model calculates risk scores using customers’ actual gambling behaviour

Training data covered every wager recorded by 13 Dutch online casinos over two years

Regulators can use the model as an independent reference when assessing operator protections
 

University of Amsterdam (UvA) researchers have released an open-source machine-learning model designed to estimate risky behaviour among online gambling customers and provide regulators with an independent supervisory tool.

The algorithm analyses behavioural indicators including bet size and frequency, play timing and duration, winning and losing streaks, and how customers respond to those outcomes. It then generates a risk score that can be compared with assessments produced by operators’ proprietary systems.

Researchers said the model could help regulators identify harmful gambling earlier and examine whether licensees are meeting their player-protection obligations.

Its public code and methodology also allow researchers to test, verify and develop the system.

The project was led by UvA PhD candidate Charles de Leau alongside psychology professor Reinout Wiers and computer science professor Johan Bollen. It was funded by Dutch health research organisation ZonMw through the Dutch Gambling Authority’s (KSA) Addiction Prevention Fund.

Training data included every wager recorded across 13 Dutch online casinos between 30 July 2023 and 30 July 2025. The researchers obtained the information through a legal provision requiring operators to provide user data for independent research.

This was the first time the provision had been used and enabled independent researchers to analyse gambling patterns across multiple operators at this scale. All forms of online gambling were included.

The model became publicly accessible through the KSA website on 18 August. It was developed in collaboration with Spanish regulator DGOJ, which is working on a separate model.

UvA positioned the open-source approach as an alternative reference point to commercial tools developed by or alongside operators. Regulators internationally could use the algorithm without relying exclusively on closed systems, although its effectiveness in different markets may depend on local data quality, product characteristics and regulatory requirements.

The model arrives as the KSA increases its focus on player protection and illegal gambling. A recent KSA-backed campaign addressed sports betting risks among young adults, while affiliate body KVA identified more than 1.6 million monthly Dutch visits to websites promoting casinos without Cruks self-exclusion.

 

 

Source

 

 

#Netherlands #GamingRegulation #ResponsibleGambling #PlayerProtection #GamblingResearch #AI #MachineLearning #OnlineGambling #Compliance