Monte Carlo simulations of players on the slot floor: segmentation of slot machines and classification of players by historical behavior
Session Title
Session 1-1-D: Data and Player Behavior
Presentation Type
Event
Location
Caesars Palace, Las Vegas, Nevada
Start Date
28-5-2019 9:15 AM
End Date
28-5-2019 10:40 AM
Disciplines
Artificial Intelligence and Robotics | Gaming and Casino Operations Management | Numerical Analysis and Scientific Computing
Abstract
Player tracking systems collect vast quantities of data that can be used to study how players utilize and move between machines on a slot floor. These data can be used to build a Markov transition matrix for each segment of the player population, which describes the probability that a player occupying a slot machine will move to any other machine on the floor. The size of the transition matrix can be decreased and the statistical noise improved by segmenting slot machines into distinct types, based on characteristics such as denomination, volatility, manufacturer, game type, theme, etc., as well as location on the slot floor. Such a transition matrix can then be used to generate realistic Monte Carlo simulations. We will present simulations based on anonymized slot floor data, in which we investigate, characterize, and classify, and visualize player behavior. We will use artificial intelligence methods to solve a large optimization problem over a space of simulations, in which we minimize a statistical metric that quantifies segmentation error. In other words, we will search for the optimal way to segment both players, based on their observed behavior, and slot machines, based on their most important features determined from their tabulated characteristics.
Keywords
slot floor simulations segmentation optimization artificial intelligence
Funding Sources
This research is privately funded by nQube Data Science Inc., which is partially owned by the author, who is also CEO of the company. The research question addressed is of a curiosity-driven and exploratory nature, and not directly related to our current products or business activities.
Competing Interests
No competing interests beyond the company involvement noted above.
Monte Carlo simulations of players on the slot floor: segmentation of slot machines and classification of players by historical behavior
Caesars Palace, Las Vegas, Nevada
Player tracking systems collect vast quantities of data that can be used to study how players utilize and move between machines on a slot floor. These data can be used to build a Markov transition matrix for each segment of the player population, which describes the probability that a player occupying a slot machine will move to any other machine on the floor. The size of the transition matrix can be decreased and the statistical noise improved by segmenting slot machines into distinct types, based on characteristics such as denomination, volatility, manufacturer, game type, theme, etc., as well as location on the slot floor. Such a transition matrix can then be used to generate realistic Monte Carlo simulations. We will present simulations based on anonymized slot floor data, in which we investigate, characterize, and classify, and visualize player behavior. We will use artificial intelligence methods to solve a large optimization problem over a space of simulations, in which we minimize a statistical metric that quantifies segmentation error. In other words, we will search for the optimal way to segment both players, based on their observed behavior, and slot machines, based on their most important features determined from their tabulated characteristics.
