This paper focuses on predicting player behaviour in two-player games with microtransactions. Typically the games are for free and companies generate their revenue by selling in-game goods. We show creation of a users behaviour model, which are then used in a recommendation system increasing in-game goods purchases. We focus on learning techniques in a novel way, predicting the time of purchases rather than the most likely product to be purchased. The player model is based on in-game signals, such as players success, curiosity, social interactions etc. We had access to a Pool Live Tour game dataset made by Geewa. We report promising results in predicting the purchase events.