I like numbers and I am generally pretty comfortable with them, but I have to acknowledge there are limits to that comfort.  After all, I am by education a liberal arts/social science kind of guy and never really had any advanced statistical training or any cause to use advanced statistics in my work.  Nevertheless, I have always loved the numerical side of baseball and was thrilled to find a community of similarly minded people, especially since it included a number of people I could learn from. 

I have recently been trying to get my arms around the various methodologies and measures of defensive ability and I am writing this post because I am hopeful there are many similarly confused people who are hoping to find some firm footing in the area of defensive metrics.  I feel like I am wading around the continental shelf in cool, refreshing water, but the 500 foot drop could just be a couple of steps away.  So, I am going to present some information that may well demonstrate that I am in over my head.  I am counting on the more statistically sophisticated to throw me a lifeline when I flounder and hopefully not ridicule me too terribly much for being out in water over my head when I don’t know how to swim very well. 

I make no claim that I have figured out any thing of earth-shaking significance.  Mine is a simple quest to try to understand baseball a little better and to actually be able to participate in seemingly esoteric discussions about the worth and desirability of various players with some degree of coherence. 

I have seen a number of measures purporting to objectify defensive prowess and many of you have no doubt seen them as well.  I will not endeavor to explain all the various methods here as others more capable than I have done that very well in other places.  I have found a couple of articles useful in gaining what I hope is a basic understanding of the various approaches.  Justin Inaz has what is about as comprehensive an overview as I can wade through on defensive metrics at his site, jinaz-reds.blogspot.com, as part of what he calls his Player Value Series.  You can check it out by scrolling down the left hand side of the home page.  Many, but certainly not all, VEBers seem to like Justin’s approach.  Another useful spot, at least for me, is the fine article by David Gassko at The Hardball Times called “Evaluating the Evaluators” (http://www.hardballtimes.com/main/article/evaluating-the-evaluators/) with additional information available at his blog that is linked in the article.

What I am searching for is a publicly available, up-to-date statistic that can be used with some confidence when talking about the relative value of a particular defensive player.  I like rate stats in general because I think they give you a more accurate picture of a talent level and seem to allow discussions of players who may play multiple positions, or are just part-time players, or have been injured and missed some time.  For example, if you want to get into a discussion of the defensive abilities of various Cardinal OFs or MIFs the various amounts of playing time they have been afforded would skew their rating by counting stats. 

Pay attention, because here comes the part where I am going to get out over my skis a little and will need some guidance from the more sophisticated.  The various rating systems I see quoted most often are John Dewan’s +/-, David Pinto’s Probability Model of Range (PMR), the fielding component of Justin Inaz’ Total Value System, Zone Rating (ZR, readily available on espn.com), Revised Zone Rating (RZR, readily available on hardballtimes.com), and MGL’s Ultimate Zone Rating (UZR).   I believe they all draw their data from either STATS, Inc. or Baseball Information Solutions (BIS). Justin’s Player Value Series has a nice overview of the features of each and both scatterplot and correlation matrices if you care to see them. 

For me as a relative novice, each of these systems has at least one significant drawback. They are either not freely available, are not compiled until the end of the season, don’t correlate well with each other, are difficult to manipulate, or are no longer being published.  As far as I know, the only way to get comprehensive +/- info from Dewan is to subscribe to it.  He provides the top ten in MLB at each position and the bottom six at www.fieldingbible.com or in The Bill James Handbook.  I like the method they employ and I think the results jibe with what I see with my own eyes, but it has limited value to me as a chaotic blogger who wants to do ad-hoc analyses.

I am somewhat intrigued by David Pinto’s PMR data, but I have to admit to some skepticism.  It doesn’t really seem to correlate well with anything else, so either David is just way smarter than anybody else (very possible) or there are some problems with either his data or his methodology.  David uses a predicted defensive efficiency ratio (DER) and then compares it with an actual DER to see who outperforms, or underperforms, the expectation.  One thing I can’t seem to understand is that his top five rated SS all have low predicted DER, while he has guys like Adam Everett, Yunel Escobar, and Jimmy Rollins as barely above average.  David, if you happen to read this the prevailing assumption here is that I am not qualified to pass judgment on anyone’s method, but I am just groping for a useful tool.  So please don’t be too harsh to me, even if I deserve it.  PMR is also only available after the end of the season and is in a pretty static format, so it doesn’t serve my purposes well. 

Justin Inaz’ Total Value system has the benefit of being expressed in runs so it can easily be used to calculate player value in dollars, plus it shows Albert Pujols as the best player in the known universe by a ridiculously wide margin, so I like that.  While Justin’s discussion on his site indicates he thinks the ideal metric would be a blend of UZR, PMR and Tom Tango’s Fans Scouting Report (FSR), I think the data he actually uses is based on RZR.  FSR is a compilation of subjective impressions from serious fans.

ZR has the advantage of being readily available on espn.com, while RZR is accessible at hardballtimes.com.  I am not certain what other differences their may be in the data, but the most obvious one is that RZR includes out of zone (OOZ) plays made while ZR does not.  My understanding is that Mitchel Litchman, known as MGL, no longer makes UZR data available, at least partly because he is currently an employee of the St. Louis Cardinals. 

So what is a poor neophyte to do?  I got tired of attempting to draw conclusions from the readily available RZR only to get stomped by someone with a better tool, so I decided to see if I could cook up something that might be a better indicator.  I think I may have found a reasonable dataset that is freely available, has data for the last five years, is sortable on a number of parameters and may have some statistical validity as well.  I call it POOZBIZ. 

POOZBIZ is a simple acronym for the components for RZR and it seems like you can use those components to come up with a reasonably accurate indicator of defensive ability.  All you have to do is do the simple calculation of: (Plays + OOZ) / BIZ.  What that tells you is how many plays in the zone (Plays) and out of the zone (OOZ) were made as a percentage of the total balls in the zone.  I started looking at this because I had read in multiple locations that one of the problems with the STATS and BIS data was that they used different methods of assigning a batted ball to a zone and were somewhat misleading.

It seemed sort of common sensical to me that most good players field most of the balls within their field of play, but I kept seeing that many players missed a relatively high number of plays yet managed to get to quite a few balls out of their zone.  That just didn’t make sense with what I was seeing when I watched games.  What I would call a true OOZ play seems to be a relatively rare event.  For example, RZR shows Cesar Izturis missed a whopping 41 plays in his zone, or almost 15%, while he made 58 plays, or over 20% of his total plays, outside his zone.  I think PIZBOOZ corrects that, at least in a practical sense, and shows that Cesar made 105.43% of the plays in his zone.  Since he made only made six fielding error all season at SS, that would mean he made about 7-8% of his total plays outside his zone.  That sure sounds reasonable to me.

I thought I better do a sanity check by correlating my results against the systems whose data was readily obtainable.  So I put together some data for SS only to see whether I might be on to something.  I should note that my baseline was all the shortstops in both leagues that The Hardball Times refers to as “qualified”.  In this instance that meant 20 SS with a minimum of 880 innings at the position.  The correlations are below: 

NOTE:  I got my +/- data from Dewan’s website, and he only allows freeloaders like me to see the top ten and bottom six at each position, so all the correlations from to +/- are based only on eight hits to my baseline.  The remaining data is for all twenty players.  I have very little concept as to whether this is truly statistically significant, but I was encouraged at the extremely high correlation between Dewan’s and Inaz’ numbers.  I would probably be willing to run the data across the remaining six positions (no P, no C) if the community thinks it would actually be a worthwhile exercise. 

I thought it would also make sense to take a quick look at the data for other positions to see who would rise to the top as a sanity check.  So, here it is, the first ever all-POOZBIZ team:

That is all I have.  I hope it is interesting and I hope it was not a waste of my time.