Online pay-per-bid auction is a new entertainment shopping format with a mechanism that integrates the characteristics of traditional e-commerce, online auctions, lotteries and games. Previous researches have shown this auction format possesses high variances and uncertainties. However, research on the players’ participation patterns and performance is rare. In this study, we attempt to study how players’ participation patterns can affect both their own and websites’ performance. An empirical study with data collected from an online pay-per-bid auction website with 5,650 players’ participation behaviors is performed. Particularly, the role of players’ preference is also explored through a cluster analysis of their historical participations. Multiple linear regression model and regression model with beta distributions are used for hypotheses testing. Findings confirm that players’ participation performance and websites’ revenue are strongly associated with the players’ participation patterns. The results show that (1) players with longer lifetime perform better and contribute more revenue to the website; (2) players participating in more auctions show worse performance and websites obtain more revenue from them; and (3) players with a strong preference behave more irrationally and website benefits more from them. Both theoretical and managerial implications are discussed.