13 Dutch Casinos Feed New Open-Source Gambling Risk Model

Dutch regulators now have an independent machine-learning tool for assessing risky online gambling. The model was built from two years of player activity across 13 Dutch casinos and is available for public inspection.

Researchers at the University of Amsterdam developed the system using anonymized player activity supplied by licensed Dutch operators. The dataset covers online gambling behaviour recorded between July 30, 2023, and July 30, 2025.

Two Years of Betting Data Built the Model

The dataset includes betting patterns across all forms of online gambling. The model looks at the following:

How much and how often a player wagers;
When sessions take place;
How behavior changes after wins or losses.

Repeated late-night play is one example of a pattern the system can assess. Those signals are combined into a risk score designed to identify behavior associated with gambling harm.

Access to the data came through a provision in Dutch law that requires gambling operators to make customer data available for independent research. According to the UvA, project lead Charles de Leau is the first researcher to use that route.

He developed the model with psychology professor Reinout Wiers and computer science professor Johan Bollen. ZonMw financed the project through the Dutch Gambling Authority’s Addiction Prevention Fund.

KSA Can Compare Scores With Operator Systems

The model was made publicly available through the Kansspelautoriteit, or KSA, on August 18. Its code and methodology are open source, allowing regulators and researchers to examine how the scores are produced.

This gives Dutch supervisors a separate reference point. Online operators already use monitoring systems as part of their duty of care, but many commercial risk tools are closed to outside review.

The KSA can now calculate its own risk scores and compare them with assessments generated by operators. Large differences could give the regulator another reason to examine how a licensee identifies risky behavior and when it decides to intervene.

The project was also developed in close cooperation with Spain’s gambling regulator, the DGOJ, which is working on its own risk model. Authorities outside the Netherlands can test or adapt the Dutch system because its underlying method is public.

Open Model Adds a New Regulatory Benchmark

The next question is how the KSA uses that benchmark in practice. A risk score alone cannot establish whether an operator has failed its duty of care. It can, however, give supervisors something they previously lacked: a model they can inspect and reproduce instead of relying solely on the operator’s own scoring system. That could make future player-protection reviews easier to challenge with data rather than assumptions.