THE BOOK--Playing The Percentages In Baseball

I haven’t read the article, but I’m assuming this data includes pop ups? More so than bunts, I would think pop ups could distort the data as some pitchers never field them while others (presumably Maddux would be one of them) always field them. Since 99% of pop ups collected by pitchers would become outs anyway, it would be nice to remove them from the study.

It’s also dependent on the 1b. James talked about Bill Buckner (good arm, bad legs), who had lots of assists to pitchers because he made them make the play.

I don’t think pitchers catch too many popups.

It depends what you count as being involved in the out.  If you do it the PBP way, it’s usually done by “first touch”.

Right, I’m doing first touch to determine play credit. An assisted putout is always counted as a play made for the player with the assist, not the putout.

The tricky part comes with unassisted putouts of the batter. Some of those are popups (and very occasionally liners - Kerry Wood catching a comebacker without looking is probably one of my favorite fielding plays of all time).

Since I’m doing this without using any batted ball data, I’m sure that I’m not doing this perfectly, but there are clues about the play (batter handedness is a big one) that can give us a good idea. Pitchers probably are having the most popups sneak in as plays made, but I don’t think it’s a big impact on the final scoring, at least on a career level.

It occurs to me now to simply just count how many plays the other infielders are making. So, yeah. I’m gonna do that.

Nothing to add. Just subscribing to this thread, cause it looks like it could get really interesting.

Couldn’t you use regression to estimate the marginal value of each extra play made by pitchers?  Make infield outs the dependent variable, and pitcher’s own plays made the predictor variable.  Obviously, the marginal value won’t really be the same for every pitcher, but I think that should give you a good average estimate.

Can you give any extra credit to pitchers that start double plays? A softly hit ball that might’ve been fielded by the 2B or SS instead might be a double play if the pitcher gets it and only one out if a 2B or SS fields it. Groundball pitchers might even get an extra affect from this.

Guy, that’s interesting enough. 

Colin, can you present the results of this:
1. Take all the pitchers in your top 10, and show us the outs made by infielders per presumed GB BIP
2. Do the same for the pitchers in your bottom 10.
3. Compare.

Ideally, the two rates will match, and therefore, there is no ball hogging happening.  Realistically, we will see *some* drop in outs per GB for the good fielding pitchers (which shows ball hogging).

Well, I can present the results. They don’t match. For the top ten, looking at each pitcher’s plays made above average, and then the other infielder’s plays above average while playing behind those pitchers:

Then, the bottom 10:

So yes, there seems to be a ballhogging effect. I also took plays made divided by average plays made, for pitchers and their other infielders. The correlation was -0.52.

Would it be useful to compare pitchers outs versus fielder outs for pitchers that pitched on the same team, and see if the fielder out ratio changed based on how good of a fielder the pitcher was?

For example, Glavine, Maddux and Smoltz were starting pitchers on the Braves from 1993 to 1999. Assume that they basically pitched in front of the same batch of fielders (ATL) over that time frame. Basically, ATL is the “control” group.

Does the ratio of Glavine outs to ATL outs in games that Glavine pitch in differ from Smoltz and the ATL outs in games Smoltz pitched in, and similary from Maddux? If there is a difference, does it provide insight into good fielding or ballhogging?

Wow, that’s a big effect.  Each marginal play by a pitcher is really only .5 outs at the team level.  I suppose the pitcher sometimes makes a “better” out, getting a lead runner.  On the other hand, pitcher outs probably prevent a GDP some of the time.  So .5 is probably close to the real relationship, and Tango’s .35 rule of thumb works extremely well.

I agree, huge effect.  I love learning stuff like this. 

So, only about half the outs are actually saved, and the other half are simply transferred.  (Colin, in addition to the r= -.52, can you should the actual correlation equation so we can see the magnitude of the slope?)

Look especially at Glavine v Maddux.  ALL the infielders (including themselves) are +290 for Glavine and +211 for Maddux.  But, if you exclude the pitchers, it’s +82 for Glavine and -178 for Maddux.

Colin, I think you have a pretty good test case here of two great fielding P behind identical infields (if you limit the study to their overlapping Braves years.. might as well include Smoltz too).  We might learn some fascinating styff.

This could also mean an adjustment for IFs. Braves IF playing behind Maddux and Glavine were a bit better than Colin’s metric thinks.

Braves starting pitchers, ‘93 through ‘02:

If there’s something there, I’m not seeing it.

Good stuff.  Maddux looks interesting, as the bulk of his bad fielders must have happened with the Cubs then.

That would be another interesting thing: Maddux with different teams.

Really, we can go through this and dissect tons of interesting viewpoints, much like WOWY gave us a new perspective.

The main takeaway is the negative correlation between plays made by pitchers and plays made by infielders, showing an extreme ball-hogging effect.

Colin:  are you using the pitchers’ GB/FB distribution (as opposed to the team’s) when you rate the infielders behind them?

For the particular purposes of this study I am, Guy.  I am not generally, as it’s rather more computationally intensive than I would like, for seemingly little return in turns of results. (The MOE drops just barely per play when I use a pitcher’s own batted ball distribution instead of the team’s, but it has an exponential effect on query run time.)

Here’s the scatterplot (with formula) for plays:

http://flic.kr/p/8pLEqC

I agree with Colin’s basic point.  Unless you have infielders being paired with pitchers (which I suppose IS possible if you pair a good infielder/bad hitter with Derek Lowe and a bad infielder/good hitter with a FB pitcher), then it won’t matter what you do.

Totally agree.  Just thought it was important for this particular study.

Are the OIF-PM affected by the pitcher’s BABIP abilities or only the infields defensive ability?  In other words, does it capture all plays made above or below regardless of who gets the responsibility or is it just showing the fielders responsibility?

Question - if Glavine’s infields were + 81 total for his career, and only +23 from 93-02 when they were at their alleged pinnacle, than that means from 1987-1992 and 2003-2008 (with some presumed mediocre to poor 1987-1990 and 2003-2005 seasons included) his infields were +58.  Am I missing something?

Also, what can we derive from the fact that Glavine is the only pitcher on wither list to not have a negative correlation from PIT_PM to OIF_PM?  Anything?

#14 Is CH just infield chances or all chances?

I wonder if the difference in infield performance might be a right handed/lefthanded thing somehow… maybe Glavine and Maddux didn’t get the same performance from their infielders because one pitcher got their pitches pulled to one side of the infield and the other pitcher got theirs pulled to the other side. Or maybe hitters off one pitcher were more likely to hit it down the line while the other pitcher was more likely to hit it up the middle.

An additional thought… though I wouldn’t know how to go about it (as if I ever do). But you could use these kinds of pitcher vs home infield defense comparisons to determine how well a team fields routine groundballs, or the flipside, how well a pitcher generates routine groundballs (as opposed to hardly hit, unfieldable ground balls). If a pitcher good at generating infield outs switches teams and no longer generates those outs, it would imply a deficiency in team infield defense.

I’m going to try to try to tackle this idea tomorrow, but what what if one were to check groundball hit percentage—Batting Average on Ground Balls in Play—for Pitcher A, then compare it to the GB BABIP for Pitcher B, another pitcher on the same team, during the same years?  (If GB BABIP is too cumbersome, it might not even be a big deal; the effect, if there is one, should show up in regular BABIP as well.)

There are a few other things that could need to be controlled and/or examined—for example, each pitcher’s career BABIP—but it should present a pretty clear picture of the value of a pitcher’s defensive contribution. 

Also, there is going to be diminishing marginal value to Pitcher plays made as the infield defense behind him improves, right? It seems logical that the more capable the defense, the more likely it is that any play made by the pitcher could be classified as boll hogging.  If this relationship does exist, what position has the biggest effect on ball hogging?

Finally, might there be some effect—independent of the quality of defense mentioned above—when a team tends to put the shift on more often?  What about bad pitchers who are pitching with runners on base more often? 

Anyway, this is all great stuff.

post #22

We know for a fact that Glavine had a largely disproportionate number of BIP go to the right side of the field compared to other pitchers.  Not sure what impact that has.  Looking at Colin’s scatterplot, we see Glavine as an extreme outlier. which is nothing new to Glavine, who seems to be an outlier in many regards.

Going through the list:

20 - I’m just counting plays made, compared to the baseline. If the pitcher is somehow generating “easier” chances in ways that I can’t measure, fielders are getting credit for that.

21 - I don’t know if it means anything; maybe Glavine just happens to be the guy who fielded well but played on nothing but great-fielding teams.

22- The average baseline controls for batter handedness.

Most fielding metrics have shown Maddux to have been saved 30-60 runs by his defenses and Glavine 50-80 runs.  How does this info. on their fielding abilities change those numbers, if they do?

Something I just thought of, but maybe ballhogging is a good thing with a good defensive pitcher. Wouldn’t Tom Glavine on the Mets make the Mets team defense look better if he’s getting to balls that Luis Castillo and Carlos Delgado wouldn’t have? Maybe both Maddux and Glavine got to balls that Chipper wouldn’t have.

I’d say it’s a good thing.  You don’t know if the infielder behind would make the play, if the pitcher can then getting that out is a very good thing.

Might have helped getting to balls that Chipper wouldn’t have gotten to.  Probably got to a good number that Chipper would have had as well.  That would explain why Chipper made many fewer plays than other 3B, but the play by play data does not show an increased amount of hits going through the 3B area.

That would explain why Chipper made many fewer plays than other 3B, but the play by play data does not show an increased amount of hits going through the 3B area.

And this is the exact reason why my BZM fielding metric doesn’t consider any GB that the pitcher touches in the 3B’s zone a “chance” for the 3B.  Likewise for pitcher touches in SS, 2B, and 1B zones, and 3B touches in the SS zone and 1B touches in the 2B xone.

It’s hard to tell on a lot of this, and I really don’t know the intricacies of the various metrics. I don’t know if a pitcher is ballhogging chances that other fielders would have made easily, in effect, giving his fellow infielders more difficult plays to handle (and likely miss).. or if a pitcher is ballhogging difficult chances. I would imagine this kind of effect would be more obvious in statistics not related to zones like range factor or fielding percentage, but I figure those are really only used in mainstream articles.

Rally -

do you see anything in this study that may make you adjust WAR at all, or is all of this already captured in WAR some way or another?

Here is one thing I, in my limited sabermetric understanding, cannot figure out.

Take Glavine for example: Glavine + the fielders behind him are about +140 hits saved on BIP compared to league average.  Weyer’s has Glavine/infielders +290 hits saved.  That does not even account for hits saved by the OF, which presumably would put the total to well over +300.  What is causing the massive gap between the two?  Does it have something to do with adjustments for GB rate?

Thank you for the clarification.