Predicting Home Runs

            So after a very long hiatus due to graduate school (which is almost over!), I have come back to this blog.  The fact that Opening Day has just passed has really renewed my interest in statistical analysis of baseball, so I hope not to have another 7 month layoff between posts.

Also, my NCAA brackets aren’t doing well, so there’s less interest in that…

            Anyway, I recently read a series of posts on FanGraphs (by Chad Young and Mike Podhorzer) about predicting home runs.  More specifically, they were trying to find metrics that would predict HR/FB%.  HR/FB% is the percentage of fly balls that go for home runs.  The metrics the two guys used were the angle of the ball (Left field? Center? Right field?) and the fly ball distance measured in feet.  Predicting how many home runs a player will hit is obviously useful for predicting value in terms of runs.  Since runs are the most important part of winning baseball games, projecting runs is something every front office should be doing.  I am going to use some of their work to talk about Royals players here.

            First, Billy Butler.  Last year, Butler hit a career high 29 home runs.  He has never showed that much power before; his previous career high was 21 home runs.  Butler’s 2012 HR/FB% was 19.9%, which means that about 1 of every 5 fly balls went for a home run.  Butler’s career HR/FB%, including 2012, is 11.4%.  Using the equation Podhorzer and Young created, Butler’s expected HR/FB% (xHR/FB%) in 2012 was 12%.  From 2009-2011, the equation predicted Butler’s xHR/FB% within 1%, so the equation is moderately accurate (for the inquisitive, the R^2 of the equation was about 0.6).  Butler’s fly ball distance in 2012 increased from 294.6ft to 295.8ft, which is not nearly enough to explain the jump in home runs.  Given these data, I do not expect Butler to improve upon or even repeat last year’s performance.  It seems 2012 was the exception, not something that will stick long-term.  I expect something similar to 2011, when he hit 19 home runs.

            Next, Alex Gordon.  In 2011, Gordon hit 23 home runs and garnered a bit of MVP consideration.  In 2012, Gordon’s home runs dropped to 14.  Gordon’s HR/FB% go like this: 2010-11.3%, 2011-12.6%, 2012-8.5%.  The equation predicted 2012 to within 0.2%, but the equation was a bit off in 2010 and 2011 (1.4% and 3%, respectively).  Gordon’s average fly ball distance actually decreased from 2011 to 2012.  Given his age and his distance decline last year, I expect Gordon’s HR/FB% to decrease further.

           

            Third, Mike Moustakas.  In 2011, Moose’s HR/FB% was 4.4%, which is exceedingly low.  He was a rookie, and he generally struggled in the minors whenever he was promoted a level, so that was expected.  In 2012, his distance increased by about 10 feet, and correspondingly his HR/FB% rose to 9.0%.  The equation expected an increase from 2011, but not that large of an increase.  Young and Podhorzer showed that the younger a player is, the less of a decline is expected in distance.  Given Moose’s young age and increase from last year, I expect Moose’s distance to increase further with a corresponding increase in HR/FB%.  I expect Moose to surpass his home run total from last year.

            Last, Eric Hosmer.  In 2011, Hosmer’s HR/FB% was 13.5%, but the equation expected 9.2%.  In 2012, Hosmer’s HR/FB% was 11.3%, but his xHR/FB% was 7.9%.  Hosmer has confused the equation, I think.  There was a 3ft increase in distance from 2011 to 2012, but Hosmer ‘s angle decreased significantly, suggesting that the fly balls he did hit were in larger parts of the outfield, away from the foul lines.  It is difficult to predict what Hosmer will do this year.

            So what did we learn from this analysis?  Well, average fly ball distance and the angle of the batted ball are important factors for projecting home runs.  As the equation did not predict everyone correctly, there is room for improvement.  However, this equation can help shed light on whether an increase in home runs is going to stick the next year or not.

            Projection systems, of which this equation is a small part, come with error bars.  It is extremely difficult to account for injuries, slumps, mechanical changes, and luck when trying to project talent.  The equation won’t be able to predict everyone correctly, but the equation will predict many others fairly accurately.  Hitting 75% of projections and missing on 25% of projections would be pretty awesome and in the long run would benefit teams.