In the context of free-to-play (FTP) mobile games, this paper seeks to examine one question: Which factors affect a player's play duration or in-app purchases (of virtual items)? This research question has not been examined in the research literature. But it is an important one because the revenue of an FTP game developer is based on in-app ads watched by players (which depend on play duration) and in-app purchases. Using different regression models to analyze weekly data associated with 100,000 players' activities and expenditures of an FTP mobile game over a 3-year period, we provide three key results and their implications. First, game performance has an "inverted-U" effect. Players with exceptionally good/bad performance in 1 week tend to play for a shorter duration and make fewer purchases in the following week. This result implies that a game developer should monitor a player's performance and offer rewards to prevent players from dropping out. Second, virtual item novelty has a positive effect. Players who acquired new virtual items in 1 week tend to play for a longer duration and make more purchases in the following week. This result suggests that, to entice players to extend their play and increase their expenditure, the game developer should design customized virtual items with personalized pricing. Third, social interaction has a positive effect. Players who are "clan members" or played more with friends in 1 week tend to play for a longer duration and make more purchases in the following week. Hence, the game developer can benefit from this result by developing incentive for players to invite their friends to play together.