Breakthrough Machine Learning Model Gives Regulators Like the KSA an Independent Tool to Monitor Operator Duty of Care and Protect Players
Researchers from the University of Amsterdam (UvA) have unveiled a pioneering open-source algorithm designed to estimate the risk behavior of online casino players based on actual gaming activity. Unlike commercial analytical tools developed by or in close collaboration with gaming operators, this transparent model is available for free, equipping gambling regulators, such as the Dutch Gaming Authority (KSA), with an independent instrument to detect risky behavior early and better evaluate whether online gambling companies are fulfilling their legal duty of care.
Since the legalization of online gambling in the Netherlands, the domestic market has expanded rapidly, making digital casinos constantly accessible via mobile apps, games, and social media. Citing recent figures from the French regulator, experts note that approximately 60 percent of online casino revenue originates from excessive gamblers, individuals who gamble frequently and for extended periods, often suffering severe financial, psychological, and social consequences.
The Need for an Independent Model
While Dutch online casinos carry a strict legal duty of care toward their customers, they also possess the most detailed consumer data, which has historically been utilized for commercial retention and marketing purposes. Virtually all existing risk-detection tools have been proprietary systems created by commercial operators.
By introducing a public and independent model, regulators worldwide now have access to a transparent frame of reference free from commercial bias. This enables supervisory bodies to independently calculate risk scores using machine learning and compare them against operator metrics, establishing a new standard for regulatory oversight.
Unprecedented Data Access and Research Background
The initiative originated from Charles de Leau, now a PhD candidate at UvA, whose personal experiences with the impact of gambling addiction among family and friends inspired the research. After pitching the concept, funding was secured from ZonMw through the KSA’s Addiction Prevention Fund. De Leau developed the model alongside UvA professors Reinout Wiers (Psychology) and Johan Bollen (Computer Science), in close collaboration with the Spanish regulator DGOJ.
The algorithm was trained using comprehensive historical betting data covering all players across 13 Dutch online casinos over a two-year period, spanning from July 30, 2023, to July 30, 2025. This dataset was accessed through a unique statutory provision in Dutch law requiring casinos to provide user data for independent research, marking the first time an independent researcher has utilized this provision.
“The fact that we were able to analyze all bets from 13 different casinos over two years has never been done before by independent researchers. With this massive amount of data, which is normally used by casinos themselves for marketing purposes, we can see for the first time on such a scale which patterns in gambling behavior often precede serious problems.”
How the Machine Learning Model Works
The algorithm evaluates actual player behavior across multiple dimensions to compute an objective risk score encompassing all forms of online gambling. Key behavioral indicators include:
- Betting Patterns: Volume and frequency of individual bets.
- Frequency and Timing: Unusual playing habits, such as continuous multi-day or late-night sessions.
- Streak Dynamics: Player reactions and behavioral shifts following winning and losing streaks.
Because the software is entirely open source, its underlying code and methodology are fully public. This allows international researchers to verify its functionality, build upon its framework, and enable regulatory authorities worldwide to utilize it as a benchmark for supervising online gambling operators.
The model went live on the KSA website, making the resource accessible to global governments and regulatory agencies seeking to strengthen player protection in rapidly growing digital markets.
