Prof. Roi Reichert and his doctoral students Nadav Oved and Amir Feder has developed a model based on deep learning that can predict the performance of a player in a certain game according to his pre-game interviews.
In order to test their model, Nadav Amir and Roi ran it using pre-game interviews given by 36 MBA players and found that in some of the parameters checked, the model’s success predicting the players performance was as high as 60.33%, compared to 53.06% success seen with systems which are based on data from previous games.

More about the research in an article published in Haaretz newspaper.

The research paper, “Predicting In-Game Actions from Interviews of NBA Players”, was published in the journal Computational Linguistics.