Matthew-Donald Sangster, RPI Graduate Student

 

Matthew-Donald Sangster, RPI Graduate Student

Sage 4101

September 27, 2017 12:00 PM - 1:30 PM

Who won? Who is the best? These two questions may have different answers. Although those individuals who cross the finish line first and those teams that beat the others during playoffs are the winners, were the skills they displayed and the decisions they made really better than those they beat? Can one or more individuals on a losing team be said to have played better than their counterparts on a winning team? Can a diagnosis in clinical psychology be correct even if treatment fails? Can a recommendation be considered sound even when unanticipated events or the performance of other team members cause the actions to fail? Such discordant outcomes are not captured by simple measures. We tackle them here in the context of individual performance in a team setting by building a ``Snapshot'' performance metric of ``how well that team member performed in this match''. With this metric, it becomes possible to find the "I" in "Team" by determining whether an individual played well even though her team lost. Our paradigm is the widely popular competitive team game League of Legends. Our dataset consists of 1.9 million records from 539 thousand matches. We use these data to develop a behavior-based performance metric that, for this task, is effective at assessing individual contribution to task outcome in a team environment. This metric lays the groundwork for applying theories of learning to individuals within a team setting.

Add to calendar
Share|