Friday, November 06, 2009

What's Wrong with UZR?

The Twins just made a deal trading Carlos Gomez to Milwaukee for JJ Hardy. UZR is being sited by all sides of the discussion of whether this was a good deal for the Twins. The proponents of the trade point to Hardy's UZR as evidence that scouting reports that his defense is declining are wrong. People concerned about the trade point to the Twins outfielders low UZR's as evidence that, without Gomez, the Twins pitchers are going to struggle because of poor outfield defense.

So what is UZR and how is it calculated. At its most basic UZR measures how many batted balls a player turns into outs compared to other players at his position. This is done by dividing the field into zones and then recording each ball hit to each zone.  Players are then rated based on more calculations of how many runs the result represents based on how many runs are scored on average after that outcome. I am not going to go into the accuracy of this last calculation because whatever innaccuracies it introduces probably have little or no impact on the relative scores of players at the same position.

Lets look at the general idea. That you can determine a players range by where balls are hit and how often he turns them into outs. In essence, a bouncing groundball up the middle is the same as a line shot. A lazy fly ball caught at the wall is the same as Willie May's catch. The fact is, UZR ignores the most important factor in whether a fielder catches the ball which is how hard it was hit. And, as anyone who has followed the Twins ought to know, the nature of the field surface is also important. Balls hit on long grass are going to move much slower, than balls hit on the old Metrodome surface. And UZR gives no credit for cutting off a hard hit ball and holding a player to a single.

In short, UZR is all but useless in evaluating player's defense. It ignores the most important factor in whether a ball is a hit, how hard the batter hit it.

 

10 comments:

John said...

Would you feel more comfortable with a defensive metric like UZR if it included the time from hit to fielding play (or non-fielding play) as a parameter? I think I would.

TT said...

Sure. And if you can measure that you can also measure how far the player was positioned from where he needed to be. You can also measure the number of balls that he prevents from going by him for extra bases. You can also measure the strength of the throw. So you can measure the components of a fielder's defense.

But we have found that even on something as simple as pitch speed, it is hard to get consistent measurements. So when someone says they are doing it, I would want some way of verifying their data but my guess is those numbers will become a lot like pitch speed an location. Not always reliable, but close enough.

Jesse said...

What stats do you look at when you evaluate a players defense? Fielding %, Put Outs, Errors ect.... or do you think all stats for defensive play are flawed and choose instead to rely on "scout" reports and observed play? Unless I have a stat to look I probablly can't evaluate any players in this league except for the Twins and potentially our central division rivals.

Specific example, You have written a number of comments in response to other blogs about Hardy's defense:

The scouting reports on him last year when he was sent down were that his defense had fallen to average at best. You don't send a "great defensive shortstop" to the minor leagues to give playing time to a prospect or to manipulate his major league service time.

https://www.blogger.com/comment.g?blogID=11215059&postID=512998951188683738

Where are these scouting reports? How can they define him as average at best? Have they also scouted the other shortstops in the majors? At least stats like UZR are equally biased against all players. I can go to fangraphs and compare all SS together and see that Hardy is 6th in UZR amongst all SS (http://www.fangraphs.com/leaders.aspx?pos=ss&stats=fld&lg=all&qual=y&type=0&season=2009&month=0) which tells me he is still pretty darn good. Where can I go to get a list of all the scouting reports. How do they rank all the scouting reports together to get a ranking? In short, UZR is flawed but what do you suggest that is better?

TT said...

Jesse -

"which tells me he is still pretty darn good"

It shouldn't tell you anything about Hardy. What it probably should tell you is that Milwaukee pitchers got a lot of slow ground balls and soft liners to the left side of the infield. Or that the grounds keepers kept the grass watered and let it grow a little long.

Nick N. said...

It is because of the flaws that you mention that UZR shouldn't be taken as gospel in small samples, which nearly everyone who uses the metric acknowledges. Your conclusion that it is "all but useless in evaluating a player's defense" is quite a ridiculous stretch, though. Over time, UZR is generally quite accurate.

Still waiting to see these scouting reports on Hardy's defense that you keep vaguely referring to, by the way.

TT said...

"Sample size" is irrelevant.

"Over time, UZR is generally quite accurate."

Based on what?

It claims to measure the number of runs that scored, compared to the runs that would have been scored for the average team as a result of balls hit to a particular zone. It claims to do that by simply measuring where a ball was hit and whether it turned into an out. It attributes the outcomes entirely to the individual player's fielding ability. Regardless of how hard the ball was hit, how fast the runner was, the speed of the turf or any of the other factors that go into those outcomes.

I would love to see some, even semi-objective, evaluation that claim is accurate. It seems to be a matter of faith.

"nearly everyone who uses the metric acknowledges."

That isn't true. If they understood the "flaws", they wouldn't give the number any credence at all.

"Still waiting to see these scouting reports on Hardy's defense"

Try google, they aren't hard to find.

Jesse said...

I googled "JJ Hardy Scouting Report" as you suggested (http://www.google.com/search?sourceid=navclient&ie=UTF-8&rlz=1T4GZAZ_enUS325US325&q=JJ+Hardy+Scouting+Report) and here are the first 3 I saw:

1. When you think of a defense-first shortstop, you usually think of a guy who plays great defense and doesn't hit. This doesn't hold true with James Jerry Hardy. J.J. has excellent range at shortstop, more due to pure instinct than quick feet. He puts away virtually everything he gets to, and has a strong and accurate arm. His defense would have played in the majors the day he was drafted. (http://www.brewerfan.net/ViewPlayerProfile.do?playerId=160)

2. Assets - An intense player, he makes consistent contact and has growing power. In the field, he boasts a very strong arm and sound instincts for a shortstop. Flaws - Is prone to nasty hitting slumps that can go on for prolonged stretches. Isn't much of a base-stealer. Still makes too many errors at the shortstop position. (http://www2.sportsnet.ca/baseball/mlb/players/J.J._Hardy)

3. A Piggly Wiggly-sponsored scouting report shown during an at-bat by Brewers shortstop J.J. Hardy Sunday indicated that his major weakness is a hankering for Piggly Wiggly-brand certified angus beef boneless rib-eye steaks, now just $4.99 a pound. "J.J. is very strong when it comes to purchasing three Ole El Paso Mexican Dinner Kits for $5, but he's struggled lately with Golden Flake-brand potato chips, as he hasn't bought a single bag since last August (http://www.theonion.com/content/news_briefs/piggly_wiggly_scouting)

Those "Scouting Reports" while soewhat descriptive and a little funny, don't provide me with a way to rank JJ Hardy versus other players and seem to simply be opinions.

TT said...

"don't provide me with a way to rank JJ Hardy versus other players"

Why don't you just use the player's weight? There is a known correlation between how much a player weighs and the number of home runs they hit.

Or you could use runs scored. Afterall, isn't scoring runs the purpose of offense? So the player who scores the most runs is the best offensive player, right? And very, very objective.

"seem to simply be opinions."

Opinions based on statistics are simply opinions as well. They only have whatever meaning you give them.

Mike I said...

Actually, UZR does in fact use batted ball data that has been labeled as "soft, medium, or hard" hit. Both UZR and +/- use BIS data break down each batted ball into a vector on the field, the type of ball hit (fly, liner, fliner, grounder, bunt, etc.) and the approximate speed of the ball.

It's explained here: http://www.fieldingbible.com/ - check the "Overview of the Plus/Minus System," around the 6th paragraph or so.

So, basically the only basis of your criticism of UZR is entirely wrong. If you did your research about UZR and +/- before spouting off such fallacies, you'd find that there's a lot more accuracy and precision in the data than you originally thought.

TT said...

Actually there are several different versions of UZR and you are correct, there is a "hard, medium, soft" for those that use Baseball Info Solutions. But, unless things have changed, the versions that used the more common Stats Inc. data do not include that information. It appears that Fangraphs has now changed to using BIS data, which I did not know.

Adding "soft,hard,medium" to the equation certainly improves things, but that was not, and is not my only criticism of UZR. It is only the most glaring problem.

For people who want more information here is a link to an article by one of UZR's creator/advocates:

"http://www.baseballthinkfactory.org/files/primate_studies/discussion/lichtman_2003-03-14_0/"

MLB Twins Updates