Advanced Statistical Evaluation

Virtually all traditional statistics used to evaluate player and team performance owe their ubiquity to their computational ease more than their descriptive powers.  Unfortunately, while these measures are easy to use, their practical value is questionable.  To illustrate, consider using a multiple choice exam to evaluate mastery of a topic.  Such exams are very easy for the instructor to grade, but often fail to measure students' understanding of the material (particularly, the more subtle aspects).  Such is the case with most sports statistics -- they're very easy to "measure" but they don't measure what you really want (namely contribution to winning).

At SportsQuant we've developed more sophisticated measures of performance, which measure contribution to team success more directly than those found in a typical box score.  Application of this work to in-game decision-making can be found at our strategy page.  On this page, we present development and interpretation of new metrics applied to a general setting (without regard to particular strategic context).  As always, if you want help building a better mousetrap, drop us a line and tell us about it.

Football:

  • Our strategy discussions are based on maximizing the probability of winning the game.  Another reasonable criterion is maximizing expected point differential.  When do these objectives result in the same optimal strategy and when do they differ?  We answer this question by assessing the marginal value of an additional point scored.
  • Are all turnovers created equal?  Are interceptions more damaging than fumbles, or vice versa?  Does it matter?

    Baseball:

  • The Pythagorean Theorem of Baseball relates the win-loss record of a team to the runs it scores and allows.  Presumably, the runs scored and allowed indicate the inherent ability of the team and are a better evaluation of talent (and therefore future performance) than their win-loss record.  We propose an improved method which accounts for the conditional relationship between runs and wins explicitly.  Those interested can read the complete manuscript.
    •  
    Send mail to webmaster@sportsquant.com with questions or comments about this web site.
    Copyright © 2005-2009 David H. Annis, Ph.D.
    Last Modified October 25, 2009