Football teams spend billions of dollars yearly signing new players to improve their squad and to fill identified areas of need. Despite having available complete statistics from players all around the globe, teams often do not use this data at its total capacity to identify potential signings. Looking to fill this void, this work presents models intended to suggest to teams which players should they buy and sell to fulfill their specific needs, adequately assemble the roster and gain a competitive edge. A stochastic two-stage Mixed Integer Linear Programming model is presented to optimize a team’s roster choices considering their budget, desired attributes and absences throughout the season due to injury or suspensions. A data-science framework is also proposed for data collection and treatment to input it into a data-driven model. The framework is applied to real-world data from top leagues and some case studies are presented in order to showcase its results and roster suggestions.