Dutch Researchers Release Open-Source Gambling Risk Model

The University of Amsterdam research was led by PhD candidate Charles de Leau in collaboration with psychology professor Reinout Wiers and computer science professor Johan Bollen. ZonMw financed the project through the KSA’s Addiction Prevention Fund.

The project reflects the KSA’s wider investment in evidence-based gambling harm prevention, including initiatives focused on earlier risk detection, intervention and support.

The researchers obtained betting data covering every player at 13 Dutch online casinos from 30 July 2023 to 30 July 2025. They accessed the information through a provision in Dutch law requiring casinos to make user data available for independent research.

De Leau was the first—and, at the time of publication, the only—researcher to use this statutory provision. According to the University of Amsterdam, an independent analysis of every bet made across 13 casinos over two years had not previously been conducted on this scale.

The model assesses several aspects of actual gambling behaviour:

• How much and how frequently a player wagers;
• When gambling sessions take place, including repeated night-time activity;
• Whether a player continues gambling across consecutive days;
• How behaviour changes during winning and losing streaks.

The model uses these patterns to generate a risk score. All forms of online gambling were included in its development.

The resulting score highlights behaviour that may require further investigation. It does not provide a clinical diagnosis or, by itself, confirm that a player is experiencing gambling-related harm.

Many existing player-monitoring systems are developed by gambling operators or commercial technology providers. Their methodologies are often inaccessible to regulators, making it difficult to determine how risk scores are calculated or compare results across platforms.

Publishing the new model’s code and methodology gives supervisors an independent reference point. Researchers can examine its assumptions, test its performance and build revised versions, while the KSA can compare its output with risk classifications produced by licensed operators.

The KSA describes the algorithm as a transparent tool for the early identification of risky gambling behaviour. Its open structure may help supervisors investigate why similar patterns receive different risk classifications across operators or why individual systems trigger interventions at different times.

The model was developed in close collaboration with Spain’s Directorate-General for the Regulation of Gambling (DGOJ), which is developing its own player-risk model. Regulators in other jurisdictions may use the Dutch methodology as a reference, although differences in gambling products, reporting standards and player behaviour will require market-specific testing and validation.

The model’s release does not create a new regulatory obligation, and the KSA has not required Dutch operators to adopt it. Licensed providers may use the tool to enhance their player-protection controls, but the regulator emphasises that its use does not guarantee compliance.

The distinction between detection and effective intervention is evident in recent KSA enforcement against ComeOn Casino. The regulator fined operator Tulipa Ent €750,000 after reviewing 10 young-adult player files and finding that signs of risky gambling had been identified too late or followed by inadequate interventions.

The KSA has also penalised LeoVegas over duty-of-care failures, demonstrating that identifying risky behaviour must be followed by timely and appropriate action.

Risk scoring is only the first stage of player protection. Operators remain responsible for determining when identified behaviour requires:

• Manual review of a player’s account;
• Direct contact with the customer;
• Affordability or source-of-funds checks;
• Deposit limits, temporary restrictions or account suspension;
• Referral to gambling support services.

The model also supports the KSA’s 2026 supervisory priorities, which place the protection of vulnerable players and supervision of operators’ duty of care among the regulator’s central enforcement themes.

The open-source model may make risk-detection systems easier to test and evaluate, but operators must still demonstrate that identified warning signs result in timely and proportionate action. Its principal regulatory value lies in providing a transparent, independent basis for assessing whether operators are effectively identifying and addressing risky gambling behaviour.