In order for a NFL team to properly evaluate a player's performance, an estimate for the true value of a play must be calculated. Recent work in football analytics has led to the development of the expected points added (EPA; Burke et al 2015) associated with an individual play. We introduce our reproducible method for calculating EPA using the nflscrapR package, as well as different scoring event probability added measures based on our novel multinomial logistic regression expected points model. Next, we examine the repeatability and predictive value of player-level EPA using NFL play-by-play data from 2009-2016. Finally, we propose new metrics for player evaluation based on these measures, such as an EPA weighted completion percentage for quarterbacks, and compare rankings with more traditional football statistics. We emphasize why teams should evaluate players in a similar manner and note that this work has broader applications (e.g. in player contract valuation) that we intend to explore in future work.