Performance Assessement of ANJ Algorithm for the identification of excessive gambling : Background, objectives and hypotheses

Since 2019, the detection of excessive gambling has constituted a legally mandated duty-of-care requirement for licensed gambling operators in France. Within its 2024–2026 strategic plan, the ANJ established the objective of reducing both the proportion and total number of excessive gamblers contributing to operators’ GGR. Accordingly, the ANJ developed a behavioural algorithm designed to identify excessive gamblers based on player account data. This algorithm was designed as a regulatory tool to support operator compliance. Its primary purpose is to assert the number of players with excessive gambling behaviour that operators should identify. More concretely, the tool assigns each player, within each operator’s database, a risk score reflecting the likelihood of excessive gambling. It is intended for operators offering account-based gambling services, both online and in retail settings where players use an account. The aim of this study is to scientifically validate the heuristic model for screening online excessive gamblers using independent survey and account-level data. All this work is carried out under the supervision of, and in close collaboration with an international, multidisciplinary group of recognised researchers who have themselves developed algorithms to detect problem gambling and published their work in peer-reviewed international journals. Following consultations with operators in April 2025, we designed an online, self-administered survey for individuals who gamble online on platforms operated by licensed operators or monopoly operators in France. It included the 9-items PGSI self-assessment questionnaire (Ferris & Wynne 2001), complemented by five additional questions on overall gambling activity and socio-demographic characteristics. Between July and September 2025, a sample of 1251 players was recruited across all licensed online gambling platforms in France. These data were securely linked to participants’ previous year account records stored in the ANJ tracking data “vault.” Model performance is assessed through performance metrics widely used in scientific research (including sensitivity, specificity, the Area Under the ROC Curve, precision, accuracy and the F1 score), using the Problem Gambling Severity Index (PGSI) as a recognized public health–based criterion standard. To this end, the ANJ’s behavioural algorithmic model is evaluated by comparing participants’ PGSI responses with the risk scores generated by the algorithm. The analysis is based on recorded gambling activity between July 1st, 2024, and September 30th, 2025, for the 1 251 individuals included in the final analytical sample, using data stored in the Vault.

Authors

Publication Details

Journal
Open Science Framework
Published
2026-11-14
DOI
https://doi.org/10.17605/osf.io/24qru
Primary Topic
Gambling Behavior and Treatments
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Performance Assessement of ANJ Algorithm for the identification of excessive gambling : Background, objectives and hypotheses

Kàtia Lurbe i Puerto, Jean‐Michel Costes, Sylvia Kairouz, Gaëlle Challet‐Bouju et al.
Open Science Framework
Gambling Behavior and Treatments
article

Performance Assessement of ANJ Algorithm for the identification of excessive gambling : Background, objectives and hypotheses

Kàtia Lurbe i Puerto, Jean‐Michel Costes, Sylvia Kairouz, Gaëlle Challet‐Bouju, Pauline Hot, Amandine Luquiens, Thomas Delafosse, Morgane Austruy
article en

Abstract

Since 2019, the detection of excessive gambling has constituted a legally mandated duty-of-care requirement for licensed gambling operators in France. Within its 2024–2026 strategic plan, the ANJ established the objective of reducing both the proportion and total number of excessive gamblers contributing to operators’ GGR. Accordingly, the ANJ developed a behavioural algorithm designed to identify excessive gamblers based on player account data. This algorithm was designed as a regulatory tool to support operator compliance. Its primary purpose is to assert the number of players with excessive gambling behaviour that operators should identify. More concretely, the tool assigns each player, within each operator’s database, a risk score reflecting the likelihood of excessive gambling. It is intended for operators offering account-based gambling services, both online and in retail settings where players use an account. The aim of this study is to scientifically validate the heuristic model for screening online excessive gamblers using independent survey and account-level data. All this work is carried out under the supervision of, and in close collaboration with an international, multidisciplinary group of recognised researchers who have themselves developed algorithms to detect problem gambling and published their work in peer-reviewed international journals. Following consultations with operators in April 2025, we designed an online, self-administered survey for individuals who gamble online on platforms operated by licensed operators or monopoly operators in France. It included the 9-items PGSI self-assessment questionnaire (Ferris & Wynne 2001), complemented by five additional questions on overall gambling activity and socio-demographic characteristics. Between July and September 2025, a sample of 1251 players was recruited across all licensed online gambling platforms in France. These data were securely linked to participants’ previous year account records stored in the ANJ tracking data “vault.” Model performance is assessed through performance metrics widely used in scientific research (including sensitivity, specificity, the Area Under the ROC Curve, precision, accuracy and the F1 score), using the Problem Gambling Severity Index (PGSI) as a recognized public health–based criterion standard. To this end, the ANJ’s behavioural algorithmic model is evaluated by comparing participants’ PGSI responses with the risk scores generated by the algorithm. The analysis is based on recorded gambling activity between July 1st, 2024, and September 30th, 2025, for the 1 251 individuals included in the final analytical sample, using data stored in the Vault.

Open Science Framework
Gambling Behavior and Treatments
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon