Interview With Mitchel Lichtman, Creator of UZR

If you utilize the sabermetric stat WAR (Wins Above Replacement), you probably know that the defensive component of it uses a metric called Ultimate Zone Rating (UZR). If you don’t know about WAR, then buckle up and get ready for a bumpy ride to Nerd Town. UZR is a method of determining a player’s defensive value to a team by how many plays he makes in relation to other players at his position.

Through the magic of Twitter, The Point of Pittsburgh was able to contact UZR’s creator, Mitchel Lichtman, to gauge his interest in doing an interview with us. The 55-year old Lichtman has been doing sabermetric analysis for over 25 years, long before Michael Lewis wrote Moneyball that would later star Brad Pitt. Lichtman, a graduate of Cornell and owner of a Juris Doctor in law from University of Nevada Law School, snowbirds in the Las Vegas suburb of Henderson, so unfortunately this won’t be one of TPOP’s odd interviews where we hang out with our subject and do oddball things with them. Most of that reasoning is that it’s not a good idea for me to go to Las Vegas, because:

a) My wife would kill me

b) There’s a non-zero chance I’m wanted on outstanding charges after my last visit in fall of 2008

So Lichtman was kind enough to answer a few questions for us over email instead.

 

How did you first come up with the concept for UZR?

Mitchel Lichtman— I got the idea from STATS Inc. They published their results and explained the methodology in one of their STATS Scoreboard publications in the 1990 ?s. It was an offshoot of their regular Zone Rating. I ran with it, and continued to refine it over the years. There are several similar ?batted ball ? metrics, notably DRS (Baseball Info Solutions), DRA (Humpheys), PMR (Pinto), and SAFE (Jensen), among others.

 

From your 2003 articles on Baseball Think Factory, UZR is based on dividing the field into 78 zones. How do you actually analyze all of the data from all of the games — is it just you or do you have a team?

Mitchel Lichtman ? Just me. Everything is automated in one computer program (actually several preliminary programs and then one final one). The program uses BIS (Baseball Info Solutions) batted ball data. That data provides the type, location and speed (among other things, like whether there was a shift or the ball hit the outfield wall) of every batted ball. It provides the exact ?x,y ? location, as estimated by the “scouts”, of each ball in play, but the program consolidates those into around 78 discreet IF and OF locations or zones.

 

Is there a park that is more difficult than other ones for you to standardize?

Mitchel Litchman ? Yes, although I use ?park factors ? which attempt to standardize the data, some parks are more difficult than others in that regard. Coors Field is difficult as it is for just about any analysis, because of the expansiveness of the OF and the speed and trajectory of the batted balls, due to the altitude (low air density) of course. Parks with OF quirks are also difficult in those locations, like left and left center in Fenway, left field at Minute Maid, right field at Yankee Stadium, etc. Basically any park with a non-standard part of the OF, especially short and high walls, can be problematic.

 

Is there a certain type of player that “breaks” your model?

Mitchel Lichtman ? The model works the same way for any player. It basically determines how often players make or don ?t make plays more or less often than the average player at that position given the type, speed, and location of the batted ball, as well as the inferred approximate initial position of the fielder, based on outs, base runners, speed and power of the batter, and the park. However, the results have noise in them for various reasons, such that some percentage of those results will not truly represent how ?good ? or ?bad ? a player actually played, defensively.

In addition, like all ?samples of performance, ? empirical results do not necessarily reflect a player ?s true talent, just like your score on one or two (or three) aptitude tests do not necessarily reflect your real aptitude. Offensively, a bad offensive player can have a fluky good year or vice versa. Same with defense. A good defensive player might have ?bad luck ? on defense in any given time period, for whatever reasons. Obviously the longer the time period (the larger the sample), the less likely that will be the case. So larger samples of UZR will increase the signal-to-noise ratio such that the results, in runs saved or cost, will be closer to not only their actual defensive performance, but their defensive true talent during that time period as well.

 

How do you respond to the criticism that the weakness of fWAR is the use of UZR ? Or defensive metrics, in general ? It seems as if there is a general distrust of them, even setting into the sabermetric community.

Mitchel Lichtman ? It is a matter of degree. It is not a weakness per se. It is a matter of reliability. Some metrics, by their nature, are more reliable than others, given a certain sample size. Those last 5 words are critical. UZR or DRS, or a similar defensive metric, is about as reliable as maybe a 1/2 to 3/4 of a season of wOBA (a similar offensive metric, and essentially the basis for offensive WAR, which can easily be converted into runs), for example. So, it is a little problematic adding together two numbers (say, UZR runs and offensive runs) when one is a little more reliable than the other, but not hugely so. People who express a mistrust or disdain for defensive metrics or for WAR, because of that, don ?t really understand WAR and/or defensive metrics, or baseball metrics in general.

 

When you worked for an MLB team (and did consulting work), were you privy to some in-house alternatives to UZR ? In terms of fundamentals, what were some differences?

Mitchel Lichtman ? Nope. When I worked for the Cardinals (2004-2005), we used UZR. Consulting for several other teams, I was not privy to any in-house metrics they might have used. Of course, with the Cardinals and other teams, objective metrics are generally ?combined ? in some manner with scouting reports, as should be the case. That helps to reduce the ?noise ? in the metrics.

 

With Field F/X and MLBAdvanced Media on the horizon, are you concerned that UZR will become obsolete?

Mitchel Litchman ? I am not concerned. I don ?t get any royalties from UZR (I might get something from Fangraphs – I ?m not sure). It is basically open source. I think it will become obsolete, slowly, and if it does, that means that there is something better, which is a good thing. So I ?ll be happy about that! The only thing I care about is the advancement of knowledge. Recognition does not interest me in the least.

 

What concept are you working on next?

Mitchel Lichtman ? I am actually updating UZR a little. Trying to add some kind of regression so that in small samples the results more accurately reflect performance. I am also working on some pitch F/X stuff related to game theory and pitch sequencing. Some other proprietary stuff as well. Most of the stuff I do these days is actually for private consumption, although I sometimes write an article or two based on research I have done or am doing. I have my own blog where I occasionally publish some of these research pieces and other commentary, mostly during the season.

 

When you co-wrote the seminal The Book: Playing the Percentages in Baseball, did you actually meet Tom Tango face-to-face ? His desire for secrecy is legendary. Do you see a time when he reveals himself and gives his true name?

Mitchel Litchman ? (e-laughs). I have never met him face-to-face, although that probably has little to do with his desire for privacy. It was never necessary and he doesn ?t attend any of the saber conferences and events. We know each other ?s real names, addresses, phone numbers, etc. We have spoken on the phone, and email one another often.

I doubt that he will ever reveal his name, but I don ?t think it ?s a huge deal to him, although he would have to speak to that. He seems to value his privacy, which I think is a good thing, and it really isn ?t necessary for anyone to know his real name or any other personal information. It ?s not like the media is going to camp out on his doorstep. By the way, he is one of the nicest, smartest and most honest guys around. And of course one of the legends in Sabermetrics.

 

For more from Mitchel Lichtman, visit his blog MGL On Baseball. As always, please feel free to follow The Point of Pittsburgh on Twitter @thepointofpgh or like us on Facebook.

Kevin Creagh is the author of the sci-fi novel Creating Christ, available now on Amazon