Predicting self-exclusion among online gamblers: An empirical real-world study

Area of Science:

• Behavioral science
• Gambling research
• Machine learning applications

Background:

• Problematic gambling is a significant concern for stakeholders.
• Responsible gambling tools, like voluntary self-exclusion, aim to mitigate harm.
• Understanding predictors of self-exclusion is crucial for intervention.

Purpose of the Study:

• To empirically investigate behavioral and monetary factors predicting voluntary self-exclusion in online gamblers.
• To assess the generalizability of predictive models across different countries and operators using machine learning.

Main Methods:

• Analysis of player tracking data from 25,720 online gamblers across three platforms and six countries.
• Statistical modeling to identify factors associated with future self-exclusion.
• Machine learning algorithms to predict self-exclusion and test cross-country generalizability.

Main Results:

• Higher odds of self-exclusion were linked to increased voluntary limit changes, more payment methods, higher deposit frequency, and greater game variety.
• Monetary gambling intensity (stakes, losses, deposits) did not significantly improve prediction when behavioral factors were included.
• Machine learning models demonstrated effective generalization for predicting future self-exclusions across different countries.

Conclusions:

• Behavioral patterns, rather than solely monetary intensity, are key predictors of voluntary self-exclusion.
• Machine learning offers a viable tool for predicting and potentially preventing problematic gambling behavior on a large scale.
• Findings support the development of targeted interventions based on behavioral indicators of gambling risk.