What’s the “hold” up? A simulation study of information flow in a connected Bayesian network of slot players
Session Title
Gambling Mathematics: Slots & Player Experience
Presentation Type
Paper Presentation
Start Date
26-5-2026 12:00 AM
Abstract
It is very difficult for individual players to detect changes to the theoretical hold of slot machines without playing an unrealistically large number of games. This inability is due to the slow convergence of realized losses to theoretical hold for typical slot pay tables, which are volatile. Nevertheless, many operators believe that players can detect slot hold increases, particularly in locals casino markets, and therefore resist increasing hold. This viewpoint is supported by our Bayesian Monte Carlo simulations, which demonstrate that information about casino-wide hold differences (or changes) can rapidly propagate through a socially connected network of friends. Paradoxically, differences in hold that are imperceptible to any individual player are easily detected by the locals community – as a whole – without requiring any intentional collaboration between players. This talk reviews our simulation methodology and explores the sensitivity of this key result to the magnitude of hold changes; the fraction of games changed; the connectedness of the players; the strength of their communication; and how strongly they weight new information. We also explore the factors affecting how quickly the community notices a change. Finally, we examine the correlation between player beliefs about slot hold and their time on device, allowing us to provide practical advice for operators wishing to increase their slot hold and monitor for signs of players reacting negatively.
What’s the “hold” up? A simulation study of information flow in a connected Bayesian network of slot players
It is very difficult for individual players to detect changes to the theoretical hold of slot machines without playing an unrealistically large number of games. This inability is due to the slow convergence of realized losses to theoretical hold for typical slot pay tables, which are volatile. Nevertheless, many operators believe that players can detect slot hold increases, particularly in locals casino markets, and therefore resist increasing hold. This viewpoint is supported by our Bayesian Monte Carlo simulations, which demonstrate that information about casino-wide hold differences (or changes) can rapidly propagate through a socially connected network of friends. Paradoxically, differences in hold that are imperceptible to any individual player are easily detected by the locals community – as a whole – without requiring any intentional collaboration between players. This talk reviews our simulation methodology and explores the sensitivity of this key result to the magnitude of hold changes; the fraction of games changed; the connectedness of the players; the strength of their communication; and how strongly they weight new information. We also explore the factors affecting how quickly the community notices a change. Finally, we examine the correlation between player beliefs about slot hold and their time on device, allowing us to provide practical advice for operators wishing to increase their slot hold and monitor for signs of players reacting negatively.