Abstract

We present a system for predicting gaming related properties from Twitter profiles. Our system predicts various traits of users based on the tweets publicly available on their profiles. Such inferred traits include degrees of tech-savviness, knowledge on computer games, actual gaming performance, preferred platform, degree of originality, humor and influence on others. Our approach is based on machine learning models trained on crowd-sourced data. Our system enables people to select Twitter profiles of their fellow gamers, examine the trait predictions made by our system, and the main drivers of these predictions. We present empirical results on the performance of our system based on its accuracy on our crowd-sourced dataset. Ultimately, we are motivated by the automated discovery of influential gamers in social media, and its potential for streamlining product campaigns.