Predicting Problem Gambling: An Analysis of Ladbrokes and Coral Self Excluders
14 March 2023

The MRS Census and GeoDems group champions new thinking and new talent; one area they have been particularly impressed with is the CDRC Masters Dissertation Scheme (MDS)

This programme offers an exciting opportunity to link students on Masters courses with leading retail companies on projects which are important to the retail industry. The scheme provides the opportunity to work directly with an industrial partner and to link students’ research to important retail and ‘open data’ sources. The project titles are devised by retailers and are open to students from a wide range of disciplines.

MRS CGG are proud to have been granted permission to publish abstracts from the dissertations and we are sure the students have a great future ahead of them.

This abstract is by Alexander Szoke

Title: Predicting Problem Gambling: An Analysis of Ladbrokes and Coral Self
Excluders

Academic Institution: University College London

Industry Sponsor: Entain

Introduction

Background
As the popularity of online gambling has grown, so too has the public health concern for gambling addiction As a result, gambling operators have been required to implement Responsible Gambling (RG) tools to provide better player protection and increase safer gambling awareness.

One of the most extreme RG tools is self exclusion, a programme where gamblers who recognise that they have a gambling problem can ask the operator to deactivate their account for a pre set period, typically 6 months.

The ability to predict self exclusion is beneficial for gambling operators because
1. High risk players can be pre emptively managed and protected
2. Players will interact with Entain's products in a more financially sustainable way, stabilising long term revenues.

Objective
Using the gambling activity of a sample of Ladbrokes and Coral Sports bettors, the objectives of this study were to
1. Explore the difference in the gambling behaviour between Self Excluders and control gamblers (i.e. those who did not self exclude)
2. Develop binary classification models, capable of predicting self exclusion
3. Explore how the performance of these models varied based on the length of the observational period on which the models were trained

Data Overview

Data Origin
Almost all data used was sourced from Entain's database. The final data set summarised the 2021 sports betting activity of 60,000 UK-based Ladbrokes and Coral players who played sportsbook for the first time in 2021 with real money at least once.

Variable Overview
The variables created could be grouped into 5 categories:
1. Player characteristics: Included age, gender, brand and geospatial features
2. Bet Behaviour: Included summary statistics for all bets placed e.g., number of bets, number of distinct days played, Std. of bet value (£)
3. Bet Type: Summarised the risk level of bets placed e.g., average odds, percentage of accumulator bets
4. Transaction Behaviour: Included summary statistics for all deposits and withdrawals made
5. Trajectory variables: Linear regression coefficients capturing the trajectory of bet behaviour over time.

Exploratory Data Analysis

Both classes of players tended to be losers over the observation period. However, Self-Excluders tended to be the bigger losers.

Self-Excluders deposited more often but also failed their deposits more often (due to lack of funds etc.)

Modelling and Results

Modeling
A set of 12 different supervised learning algorithms were developed to determine which was most effective in predicting self-exclusion. After hyperparameter tuning, the optimal model was deemed to be an ensemble Voting Classifier consisting of a KNN classifier, a Gradient Boost classifier, a Random Forest classifier and a Neural Network.

The final model was able to correctly classify 73% of players using only a two-week observation period of data, surpassing any test accuracy seen in prior literature.