Range to left, range to right

Warning No. 1: In this article I will briefly outline a fielding system. I want to make clear that I’m in no way attempting to create something better than UZR, PMR, plus/minus, TotalZone. I went through the work I’m going to show because I needed something that could answer a few questions about minor league fielding.

This leads to Warning No. 2: My last three articles were about evaluating minor league outfield defense (arm in particular); this one is about evaluating minor league fielding for shortstops. Well, I’m not going on with the remaining positions in the upcoming weeks. The article you are reading is, I think, the last on minor league defense for a while.

When I was first exposed to the term “sabermetrics,” back at the beginning of the millennium, most of the articles I stumbled upon dealt with something I subsequently learned to call uberstats. The idea of a single number defining the overall contribution of a player to his team was really fascinating to me, and I still like things like WAR. But what I like most about WAR (as shown on FanGraphs, or in Sean Smith’s database is the split in offensive, defensive, pitching and baserunning contributions.

I want to know that, of two equal players (i.e., same total WAR) one is better defensively, because, maybe I have a pitching staff that induces a lot of balls in play, and my lineup is already equipped with heavy bats.

Similarly, the first thing that I look for when I want information about a player’s defensive value is his UZR number; but then, I’d like more details. The Range Runs and Error Runs that compose UZR are something I find really useful. Plus/minus gives also clues on the ability of an infielder moving on either side to make the plays. While I believe the Fielding Bible proprietary system has ground to cover yet in order to be on par with Mitch Lichtman’s, I really like to be able to know that J.J. Hardy is excellent going to his left, while Jose Reyes is a wizard on the right side (according to plus/minus data for 2008).

Note: Pinto’s PMR charts also give a hint on players’ abilities on either side.

Where is the utility of knowing that a player strength is on the right rather than on the left? Not during contract negotiations, I believe. I can’t imagine Billy Beane looking at some stat sheet and saying “I need this shortstop because our pitchers surrender many more hits in the hole than up the middle.”
We can suppose that this kind of info might be used during the game, to position a fielder according to his strengths, but I’m pretty confident that ballplayers already do that by themselves.

I think that probably, the main use of a stats like “plus/minus on the left” has something to do with training, especially at the minor league level.
Coaches surely know who has a weakness on going to his left and, more importantly, how to work on that, but having numbers back their impressions can’t be a bad thing.

Say a GM is looking at his options at shortstop for the future. The two most promising kids in the farm are very similar in every aspect of the game except that Player A has the following defensive line:
{exp:list_maker}plays on the left: 3 runs above average, to center: +3, to right: +3; total: 9 runs above average. {/exp:list_maker}
Player B line, instead, goes like this:
{exp:list_maker}plays on the left: +6, to center: +6, to right: -3; total: 9 runs above average. {/exp:list_maker}
I assume the GM would make a couple of phone calls to his minor league coaches. Maybe he’ll learn that Player B has an awful first step to his right that they are working on and are confident will improve. Or that he has a sore arm that prevents him from making strong throws from the hole and won’t get any better.

Thinking about things like those above, I tried to figure a way to get plus/minus values for minor league players, using hit location data from MLBAM Gameday.


First, I had to figure the average positioning of each shortstop. Here my reasoning differs form the Fielding Bible‘s: It considers plays to right as plays made on the right relative to the average shortstop’s position; that is, the cumulative average of every shortstop. What I wanted to do, instead, was count the plays made on the right relative to each shortstop’s own average position.

FieldF/X is not on sight yet; we have less sophisticated data on fielders’ positioning. Thus I went through some reasoning to create a method for finding good estimates. I was quite proud of my first idea, which made use of smoothing techniques and, most importantly, seemed to yield good results. It would have made a nice article.

But then I took the median angle at which an infielder successfully collected a grounder as the estimate of his average positioning and got results equal (if not superior) to the ones produced by the more complex method. Simpler is better, thus we’ll stick with this one.

How do I know my estimates are reasonable? Matt Thomas has been collecting fielders’ positioning data at Busch Stadium in St Louis for a few years. Look at what he presented at both PitchF/X Summits (2008; 2009) and you’ll know his data are really accurate.

I compared my estimates with his data and, on aggregate, for some reason, I can’t get an accurate value for second basemen against right-handed batters. For every other infielder I’m satisfied with the small differences.

			average position
		Matt Thomas' data	Estimates
infielder	vs LHB	vs RHB		vs LHB	vs RHB
3B		-32	-37		-28	-34
SS		-11	-17		-13	-15
2B		 17	 10		 17	 15	
1B		 36	 33		 36	 31

Actual vs. estimated position for infielders. Data on games at Busch Stadium, years 2007 and 2008. Numbers are degrees, where 0 is up-the-middle, -45 down the third base line, 45 down the first base line.

Mental Health and the CBA
A particular bit of language in the latest CBA could have negative consequences for some players.

I also checked some individual values, because Matt was very kind to share some of his data some time ago, while I was working on another project (that has yet to see the light). While sample sizes are too small to draw any conclusion, it seems that estimating the positioning from where an infielder collects a grounder gives reasonable results.

Is all of this worth the trouble?

Shortstops do not differ in their average positioning by much. After you regress toward the mean and factor in the low precision of Gameday hit locations data, we can’t blame the creators of the plus/minus system for not taking initial positioning into account.

Left, right, plays made over the average shortstop.

I believe we should lean toward a continuum; thus the best way to present fielders’ skills on each side would be doing charts like David Pinto’s, maybe smoothing the lines, like SAFE does. Nevertheless, given the lack of data on positioning and the imprecise nature of hit locations coordinates, I’ll classify the grounders into three buckets: to the player’s left, to his right and straight to him—just like the Fielding Bible does.

I arbitrarily defined the “straight on” bucket as all the grounders within five degrees from the player’s estimated position and discarded all the instances where I suspected (based on where the ball was collected) an extreme shift was on.

Then I deviated again from the plus/minus method outlined in the Fielding Bible. It counts the number of plays a shortstop made (for example) on his right and compares that count to the number of plays an average shortstop would have made given the same distribution of batted balls. That means that, given two players equally proficient on their right, the one that gets more chances on that side will look better.

I chose to compare the percentage of plays made by the player on his right side with the percentage of plays the average shortstop makes on the right. Both methods have their pros and their cons.

Sanity check

Before getting into minor leaguers, I had to check if my numbers made any sense. Making a comparison with the Fielding Bible 2008 MLB data, I visually found an acceptable agreement. I compared all the shortstops with chart like the ones that follow.

Fielding Bible‘s plus/minus in red.

I think I can’t produce all the charts, since plus/minus is a proprietary system and it would not be fair to somewhat reproduce its numbers. Suffice it to say that I found eight instances in which the two lines roughly matched (like in the Cabrera chart), seven I considered acceptable (like Renteria’s) and four in which they completely disagreed (like Rollins’).

Minor league, a couple of cases.

This time I won’t produce a top 10 list as I did for outfielders; I will just summarize a couple of cases.

Chin-lung Hu is a 26-year-old shortstop from Taiwan. He has been playing in the Dodgers farm system since 2003, when he debuted at Ogden, Rookie League. He climbed the minor league ladder and hit around .300 at every level; at the major league level he has a round .200 batting average in 150 at-bats accumulated in the last three seasons. Currently he is listed right behind Rafael Furcal in the Dodgers depth chart.

Sean Smith‘s Total Zone lists him at 15 Runs above average for 2009, one of the best shortstop performances of the year. He behaved equally well at home and on the road, so his home park at Albuquerque shouldn’t be responsible for his outstanding defense. According to my analysis he is exceptional on grounders to his left (only veteran Mexican Leaguer Ivan Cervantes is better than him), slightly above the regular minor league shortstop on balls hit straight at him, but has very poor numbers going toward the hole.

Tommy Manzella is currently believed to be the starting shortstop for the 2010 Astros. Comparisons with Adam Everett have already arisen both from an offensive (and that’s no good news) and a defensive perspective. Sean Smith values him +3, in the middle third of regular minor league shortstops. You would expect more from the next Adam Everett. My numbers see him as solid going to his left (but not as exceptional as Hu) and on balls hit straight at him but, again, disastrous on balls on the third base side.

Keeping in mind that random variation certainly has played a role in Hu’s and Manzella’s performances, we can suppose that either they have a poor movement toward their right or their arm is not very strong / accurate.

What the fans say.

On Tango’s website fans have been collectively scouting MLB players for seven years. I was curious to run correlations between the numbers I got (or those from the Fielding Bible) on plays to right, to left and straight on and fans’ rating on various aspects. Here they are (2009 data only).

Correlation with Fan Scouting Report values
             plays to center   plays to left 	plays to right
rate        	  0.14		   -0.03	   0.24
reaction      	  0.19  	    0.03  	   0.29
acceleration  	  0.10		   -0.01	   0.18
velocity    	 -0.01		   -0.04	   0.04
hands             0.09		   -0.09	   0.24
footwork          0.13		   -0.07	   0.23
th. strength      0.17	 	   -0.03	   0.26
th. accuracy      0.20		    0.00	   0.32

No particularly high correlation comes out, but that can be expected since only 32 shortstops for just one season are considered. Throwing strength and throwing accuracy have the highest correlation values when compared with plays made to the right, as I hoped. Anyway, fans seem to give out higher ratings in all categories to players who are good on grounders in the hole.

References & Resources
Hit location data for MLB and minor leagues come from MLBAM Gameday.

Fans Scouting Report by Tom Tango.

Lookout Landing has a wonderful compilation of sabermetric resources. Scroll down to the Defense section for links on UZR, PMR, plus/minus and many more.

Thanks to Matt Thomas for sharing some of his data on fielding positioning and to Ben Jedlovec of Baseball Info Solutions for a clarification on the plus/minus system.

Print This Post
Sort by:   newest | oldest | most voted
Sean Smith
Sean Smith

This is good stuff Max.  Sounds like you are well on your way for a top flight fielding system.  I would love to merge the hit location data into totalzone, but I’ve got a lot of work to do.


Pretty sweet stuff Max.

I’m interested in some of the technical details here: how’d you do the translation from pixels to vector for the minor league parks? Do minor league parks use standard pixel maps, unlike MLB? If not, maybe that contributed to some of the discrepancy between the data sources.

Max Marchi
Max Marchi

Yeah, that’s something I should have mentioned.
I used a standard conversion, while pixel maps are not standard.
So your point is correct.

Jeff Sackmann
Jeff Sackmann

Most of the pixel maps are pretty close, but a few parks are really wacky.  If you browse around minorleaguesplits.com and look at the spray charts, there are some that are hilarious.  (I’m also using the standard conversion there.)