Real-Time AI for Predictive Responsible Gambling Using Live Wagering Transaction Data
Session Title
AI in Gambling: Machine Learning & Risk Detection
Presentation Type
Paper Presentation
Start Date
26-5-2026 12:00 AM
Abstract
This paper presents a real-time artificial intelligence framework for identifying elevated gambling risk using live wagering transaction and behavioral telemetry. Unlike traditional responsible gambling approaches that rely on self-reports, periodic reviews, or retrospective analysis, this framework evaluates player behavior continuously during active play. The system organizes telemetry into three variable groups reflecting increasing behavioral sensitivity. A baseline layer captures cumulative exposure and engagement, including total session amount, total session time, and visit frequency. An intermediate layer measures behavioral change, including game switching, bet size variation, and denomination shifts, which are associated with escalation and loss-chasing behavior. A high-resolution layer evaluates short-term dynamics such as cash-in and cash-out cycles, betting speed changes, and win–loss streaks, which reflect impulsivity and emotional volatility. These variables are processed using time-series modeling, behavioral clustering, and anomaly detection to generate continuously updated player risk scores. Rather than treating gambling risk as a static classification, the framework models it as a dynamic process that evolves within and across sessions. All detected events are recorded to support structured analysis and regulatory review. This approach enables objective measurement of risk trajectories and intervention timing in regulated gambling environments.
Real-Time AI for Predictive Responsible Gambling Using Live Wagering Transaction Data
This paper presents a real-time artificial intelligence framework for identifying elevated gambling risk using live wagering transaction and behavioral telemetry. Unlike traditional responsible gambling approaches that rely on self-reports, periodic reviews, or retrospective analysis, this framework evaluates player behavior continuously during active play. The system organizes telemetry into three variable groups reflecting increasing behavioral sensitivity. A baseline layer captures cumulative exposure and engagement, including total session amount, total session time, and visit frequency. An intermediate layer measures behavioral change, including game switching, bet size variation, and denomination shifts, which are associated with escalation and loss-chasing behavior. A high-resolution layer evaluates short-term dynamics such as cash-in and cash-out cycles, betting speed changes, and win–loss streaks, which reflect impulsivity and emotional volatility. These variables are processed using time-series modeling, behavioral clustering, and anomaly detection to generate continuously updated player risk scores. Rather than treating gambling risk as a static classification, the framework models it as a dynamic process that evolves within and across sessions. All detected events are recorded to support structured analysis and regulatory review. This approach enables objective measurement of risk trajectories and intervention timing in regulated gambling environments.