THE BOOK--Playing The Percentages In Baseball

This was discussed in a Lookout Landing thread couple of weeks ago.  I get different results than Matthew.  Using the same data source, Gameday 2007-2009, with all pitchers with at least 500 BIP, I get the following:

Count adjusted Linear Weights on fly balls + line drives by ground ball rate:

lw = gb*.27 + .06

R-Squared = .236, P-Value < 2.2e -16

Ground ball percentage was defined as ground balls divided balls in play.  Count adjusted Linear Weights are using John Walsh’s method.

So for every 10% increase in ground ball rate, you’d expect .027 increase in the Linear Weights per ball in air.

In 200 innings, the average pitcher will allow around 590 balls in play.  The league average pitcher allows around 45% ground balls, so that means he’d allow around 310 line drives + fly balls (if you eliminate pop ups).  Going by the formula above, you’d expect him to give up .18 runs per line drive + fly ball (and that matches up perfectly with my league linear weights).  So that’s 56 runs.  If you give that same pitcher a 55% ground ball rate, you’d expect 64 runs on balls in the air.  So that’s a difference of 8 runs. 

However, that 55% ground ball pitcher wouldn’t allow 310 balls in the air, he’d allow more like 265.  In that case, the difference in absolute runs is exactly 0, however, the difference in runs per ball in air is pretty big (and significant with a relatively high R-Squared).

BTW, if you have the threshold at >= 1000 BIP for the original sample, you get a slope of .24 and an R-Squared of .27.  The only way I get numbers as weak as Matthews is if I include all pitchers with no minimum BIP threshold.  Doing that, I get a slope of .1 and an R-Squared of .02.

Some if it got cut off by the pre tag.  The equation on top should be:

Count adjusted linear weights on fly balls + line drives by ground ball percentage.

Oh this is interesting.  When I do the same regression with pitchers having at least 500 BIP, except using HR/(FB+LD) as the y-axis, I get a negative slope and an R-Squared of .036.  If I limit to pitchers having at least 1000 BIP, I get a slightly higher R-Squared, but roughly the same slope. 

So it appears that while the groundballiness of the pitcher has basically zero effect on his HR/(LD + FB), it has a pretty significant effect on his Linear Weights for LD + FB.  I think this probably warrants further investigation.

I am glad you added #3.  In your first post, you said that you “Got different results from Matthew.”

As I even explained in the opening post. Matthew only looked at HR per fly line drive, fly ball, and both combined.  He did not look at anything else. How are your results “different” when you looked at different things?

Did you even read his article before you made your first post?  He looked at something completely different that you did, and he made no reference to whether the fly balls had lesser or greater run value.  He only looked at HR per fly ball rates.  Nothing else.

I am frankly getting tired of starting threads and then having people blatantly and irresponsibly misread (or not read at all) and misinterpret other people’s writings.  In another thread, someone tells me what I was referring to, as if they could read my mind!

Sometimes I think people just like to argue things even when they have to make things up to argue about.  I’m serious about this.  I am going to ask people to seriously read and try and understand what people are saying before they shoot their mouths (or keyboards) off!  I know I sound like a prick but this blog is turning into a dick measuring contest and I don’t like it.  Sorry to take it out on you, Nick, so don’t take it too personally, although yours is a blatant offense (with respect to what I am talking about)…

This was also covered by David four years ago:

http://www.hardballtimes.com/main/article/the-truth-about-the-grounder/

studes - That was a very good article by David.  Thanks for linking it.

No dick measuring intended.  To be honest, I didn’t think too much about my choice of words and was pretty much musing about the idea that ground ball pitchers get hammered when they allow balls in the air.  Honestly, I assumed that was the goal of his study, not just to look at HR/FB, so that’s what I was referring to when I said “different results”.  My mistake.

No biggie.  Just another one of my middle of night rants! At least you corrected yourself.

Matthew looks at non-home run slugging percentage now:

http://www.fangraphs.com/blogs/index.php/groundballs-and-slugging-rates

He finds a small, but significant effect for fly balls.  I really wish he would have just started with Linear Weights on LD + FB.  Trying to infer just how much of an effect groundballiness has on performance on LD + FB based on a combination of HR/FB + HR/LD and non-home run slugging on FB and non-home run slugging on LD is kind of confusing.  I can’t really tell if his results agree with mine in post 1 or not now