The Science of Risk Prediction
- Identify players with the right combination of on-field talent and off-field character.
- Minimize the risk of player selection errors with Achievement Metrics.
- Provide teams, organizations, insurance underwriters, financial service institutions, and sponsors the tools to make more fully informed decisions.
Achievement Metrics can help NFL teams improve their player selection decisions. We believe it is vitally important that teams make every effort to gain an understanding of the character of new players joining the team. Such knowledge can help NFL front offices select players that complement the existing members of their teams.
A team’s ability to anticipate and proactively address potential character issues will likely be reflected in improved win-loss record and increased stability. Players will also benefit, if matched with the right team, a player with character concerns could see his career extended and his earnings increased.
Achievement Metrics is in a unique position to assist teams with these concerns during the player evaluation process. Using software and statistical techniques developed and refined over the past decade by Social Science Automation, Inc., our business partner, Achievement Metrics is able to predict a player’s likelihood of arrest or suspension while in the NFL based on an analysis of his speech while in college.
We have identified statistically significant similarities in the pre-NFL Draft speech of players who have gone on to be arrested or suspended for specific types of activity while in the NFL. These similarities in speech are very subtle and do not rely on racial, cultural or regional differences in players’ speech patterns or expressions. We do not search for readily identifiable verbal cues (that is, the use of what some would consider obvious “problem” words or certain examples of slang or regional dialects).
Examples of the types of variables used in our models include the relative frequencies of verb tenses, adjectives and adverbs spoken by the players. These similarities in speech, combined with the results generated by automated coding schemes developed by Social Science Automation, are used to construct our player risk prediction models.
The speech data on which the statistical models rely come from transcripts of players’ football-related speech (for example, mid-week and post-game press conferences). Because we work with transcripts, the models do not depend on non-verbal cues such as body language or eye movements.
Please CONTACT us if you are interested in learning more about our statistical models, the sampling of current players in our database, or to receive more information on a particular player.