After the thread about our pitchers' GIDPs, I got to wondering what "caused" GIDPs. With all the talk of Bradford and his crazy ground balling ways, I decided to look at how certain factors effect a team's GIDPs. So, using some of Basevall Prospectus's fantastic tools I set up a little analysis.

I wanted to look at every team from the past 5 years and run regression analyses to see how much three factors (groundball rate, pitcher strike-out rate, and defensive efficiency) effected a team's likelihood of turning a double play. I collected DP% (the rate at which teams convert double plays in all oppurtunities), K%. and Def_eff for all the teams. Then, I ran the regression. Surprisingly, to me at least, none of the factors held a strong enough relationship to account for differences in teams' DP% (which ranged from the 2006 Nats at 9.7 % to the 2005 Cards at 17.5%). So, I took a look at the data.

I noticed something that seemed way off. The GB% year by year averages went 53.73, 45.66, 45.28, 45.12, and 45.33 (for 2004-2008 respectively). I don't know what the hell was going on in 2004, but this number was throwing everything off. After taking out 2004, the analysis was much more in line with what I expected.

GB% and K% had a statistically significant (at 95%) relationship with DP%. Def_eff was close to significant, but not quite. The coefficients for GB%, K%, and def_eff were .27, -.25, and .18. This finding, though, was inconclusive, to me, because it included a variable that did not, in this sample size of 124 teams, seem to be statistically significant. Here is how the relationship between the variables plays out, though, with def_eff in the mix:

DP%= -.072 + .269*GB% + (-.248*K%) + .175*def_eff

This tells us, that in a vacuum as a pitchers GB% goes up, K% goes down, or def_eff goes up we expect DP% to go up.

As I said I was unhappy with these results, I ran another regression with only GB% and K% as the variables. This time the relationship with the two variables was even greater. The t-score (for stat-geeks) increased for both. Our new model for predicting DP% looked like this:

DP%= .038 + .276*GB% + (-.203*K%)

So, what does all this mean? Well, the expected outcome that groundball pitchers are much more likely than any other pitchers (especially more than strike-out artists) comes out to be true. As it pertains to the Rays, this year is our best year for DP% despite the fact that our projected DP% is lower this year than any of the last 4. The difference between our expected DP% (11.55%) and actual  DP% (13.5%) is sixth out of the 124 teams. This does not surprise me having seen the effectiveness of our turning the double play.

One other thing. Looking at our relievers and theirexpected GIDP% (w/o accounting for our defense), it's fairly obvious just how effective Bradford is in situations like this.

Per skyking, all the pitchers and how they fared in each aspect.

Data for 2008 teams using RZR

This shows the GB%, SO%, Fielding rating for infielders, the predicted (by the three measures) and actual DP%, and the skill implied (s1, s2, and s3). The new coeffecients are .24, -.2, and .0000392. The effect of RZR looks small, but that's because this is the first timeit isn't a percentage.