Long-term behavior of casino games
Session Title
Gambling Mathematics: Casino Game Theory
Presentation Type
Paper Presentation
Start Date
26-5-2026 12:00 AM
Abstract
One of the central questions in gambling mathematics is how total return (or total profit) compares, in the long term, with the total amount bet. This ratio not only measures how favorable or unfavorable a game is, but also quantifies the extent to which it favors one side or the other (typically, the casino). Although this problem has been widely studied through fundamental parameters such as return to player (RTP) and house advantage (HA), its analysis usually relies on strict and unrealistic assumptions. In particular, wagers are typically assumed to be independent and identically distributed, effectively modeling repetitive and nonadaptive play. In this presentation, we introduce a general framework for studying long-term return and profit under very mild assumptions on the sequence of wagers. We show that players may adapt their betting behavior arbitrarily over time, yet no betting strategy can escape the constraints imposed by the game’s intrinsic parameters in the long term. The analysis relies on martingale methods, in particular strong laws of large numbers for martingales. The set-up applies to a broad class of games, including compound games and those with delayed resolution. Representative examples include roulette, slot machines, and craps.
Long-term behavior of casino games
One of the central questions in gambling mathematics is how total return (or total profit) compares, in the long term, with the total amount bet. This ratio not only measures how favorable or unfavorable a game is, but also quantifies the extent to which it favors one side or the other (typically, the casino). Although this problem has been widely studied through fundamental parameters such as return to player (RTP) and house advantage (HA), its analysis usually relies on strict and unrealistic assumptions. In particular, wagers are typically assumed to be independent and identically distributed, effectively modeling repetitive and nonadaptive play. In this presentation, we introduce a general framework for studying long-term return and profit under very mild assumptions on the sequence of wagers. We show that players may adapt their betting behavior arbitrarily over time, yet no betting strategy can escape the constraints imposed by the game’s intrinsic parameters in the long term. The analysis relies on martingale methods, in particular strong laws of large numbers for martingales. The set-up applies to a broad class of games, including compound games and those with delayed resolution. Representative examples include roulette, slot machines, and craps.