Identifying High-Risk Players in Online Casinos: A Machine Learning Approach
Session Title
AI in Gambling: Machine Learning & Risk Detection
Presentation Type
Paper Presentation
Start Date
26-5-2026 12:00 AM
Abstract
In many states, players may voluntarily opt-in to limit-setting options, but, typically, operators are not required to use evidence-based indicators to identify or intervene with players who are gambling at the highest levels. In addition, the absence of financial player information further limits our understanding of how or if betting patterns correspond with actual risky play. The goal of this study is to use machine learning (ML) techniques with six years of online casino betting data to identify players gambling at increasingly higher risk levels not tied to actual spend. Unsupervised ML techniques were used to identify behavioral risk profiles from 8 million daily wager records from 128,547 individuals gambling in online casinos from 2016 through 2021. Feature engineering produced indicators of temporal dynamics, volatility, escalation, and anomalies. Highly correlated features (|r| > 0.9) were excluded, and dimensionality was reduced via PCA. K-means clustering was applied on the standardized feature set; the optimal number of clusters was selected using the elbow method and validated with silhouette score and Dunn index. The model used 35 indicators related to betting volume, wager changes, betting outcome, and outcome-contingent wagering to detect high-risk gamblers. Players with inconsistent betting volume and wager amounts, frequent increases in betting after losses or wins, and sudden changes in net gain/loss on rolling and exponential metrics were at highest risk.
Identifying High-Risk Players in Online Casinos: A Machine Learning Approach
In many states, players may voluntarily opt-in to limit-setting options, but, typically, operators are not required to use evidence-based indicators to identify or intervene with players who are gambling at the highest levels. In addition, the absence of financial player information further limits our understanding of how or if betting patterns correspond with actual risky play. The goal of this study is to use machine learning (ML) techniques with six years of online casino betting data to identify players gambling at increasingly higher risk levels not tied to actual spend. Unsupervised ML techniques were used to identify behavioral risk profiles from 8 million daily wager records from 128,547 individuals gambling in online casinos from 2016 through 2021. Feature engineering produced indicators of temporal dynamics, volatility, escalation, and anomalies. Highly correlated features (|r| > 0.9) were excluded, and dimensionality was reduced via PCA. K-means clustering was applied on the standardized feature set; the optimal number of clusters was selected using the elbow method and validated with silhouette score and Dunn index. The model used 35 indicators related to betting volume, wager changes, betting outcome, and outcome-contingent wagering to detect high-risk gamblers. Players with inconsistent betting volume and wager amounts, frequent increases in betting after losses or wins, and sudden changes in net gain/loss on rolling and exponential metrics were at highest risk.