I’m embarking on a fairly sizable project to see if pitch F/X can influence the world of player projections (in fact if my adviser will allow it I’m going to try and use it as my masters thesis).  The first step I’m taking is coming up with predictive models for batted ball type based on “stuff” (horizontal and vertical movement and velocity) and location.  My initial fooray into that has been GB%.  Here’s what I’ve got so far….  I’ll give you direct R output and then give a quick explanation.  NOTE this is for fastballs (all kinds) only, other pitch types still to come.


Estimate Std. Error t value Pr(>|t|)

pfx_x -0.227381 0.268034 -0.848 0.3968

pfx_z -2.503483 0.235354 -10.637 < 2e-16 ***

start_speed 0.550177 0.046822 11.750 < 2e-16 ***

Way.Down 0.391974 0.063755 6.148 2.01e-09 ***

Edge.In 0.163802 0.077946 2.101 0.0363 *

Edge.Away -0.006597 0.071885 -0.092 0.9269

Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ‘ 1

Residual standard error: 10.01 on 375 degrees of freedom

Multiple R-squared: 0.9506, Adjusted R-squared: 0.9498

F-statistic: 1202 on 6 and 375 DF,p-value: < 2.2e-16

A quick explanation of variables:  pfx_x,pfx_z, and start speed represent the same things they traditionally do.  Way.Down is % of fastballs knees or below.  Edge.In and Edge.Out are % of pitches on the edges of the strike zone and beyond.  Those terms with *s behind them (the more the better) were deemed significant, so vertical movement (duh!!), velocity, % of pitches down in the zone (duh!!! again), and to a lesser extent % of pitches inside were all significant. Note that the adjusted R squared is high which is promising (it basically means 95% of the variation of GB% around its average is explained by this set of variables).

Now’s as good of a time as any to mention that this technique is not meant to provide insight into cause and effect, it is merely meant to predict GB%, so be careful what you read into it.  That being said, an additional inch of sink (all else being the same) will predict a 2.5% increase in GB% and throwing and extra 5% of your fastballs at the knees will increase the GB% by ~2%.  Neither of these results are particularly suprising.

Next up, I’ll apply the model to some Cardinal data and see how they shake out.

Steve Sommer

Simulation analyst by day, father and baseball nerd by night

More Posts - Twitter

Leave a Reply



You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>

© 2011 Gas House Graphs Suffusion theme by Sayontan Sinha