Being versatile -having the ability to perform different functions- can be a desirable trait for firms that must compete in different environments. However, attaining versatility may come at the high cost of sacrificing the kind of excellence that can only be achieved through specialization. Furthermore, when strategic interaction involves any opportunities for cooperation, then being more versatile might lead to lower equilibrium payoffs, even if such greater versatility were costless. It follows that whether versatility tends to pay off or not is an empirical question, which needs to be asked in each specific setting. We use detailed match-level data on players' interaction from the Spanish La Liga 2018-2019 soccer season and the Colombian 2018-1, 2018-2 and 2019-1 seasons to define a measure of observable versatility. Namely, for each team, we count the number of distinct sub-teams of all possible sizes (2-11) that are ever observed to interact during the season and thus build a team's 10-dimensional observable versatility profile. We observe that teams displaying a greater number of distinct sub-teams of large size (4 and above in the case of the Spanish league) tend to perform better. We compute 1- dimensional observable versatility indices by applying PCA to the 10-dimensional observable versatility profiles and show that these simple indices are strong predictors of teams' performances.