THT: How valuable is UZR?

This is an interesting look at how reliable UZR is as a metric of defensive ability, complete with graph. Definitely worth a read, given our frequent forays into The Skip Schumaker Experiment via UZR.

4 months ago bgh 4 comments 1 recs |

Story-email Email Printer Print

Comments

Interesting read

I wish they had provided a little more meat. Pretty much the conclusions I would have expected.

I wonder if you only used players who played the same positions in the same park for a number of years if you would find the correlation to be a little stronger? I think OFs are at the mercy of the shape/size of the park they play in somewhat.

Those Pilgrims ain't lookin' so proud now...

by giveml on Jul 2, 2009 1:04 PM EDT reply actions 0 recs

What meat do you want?

I has a blog. | Ask a sabermagician!

by cwyers on Jul 2, 2009 2:12 PM EDT up reply actions 0 recs

This definitely enforces the belief that UZR data isn't very predictive after 1 year

It would be nice to see a two year correlation study, but I suspect that would kill your sample size.

Derosa.

by vivaelpujols on Jul 4, 2009 9:03 AM EDT reply actions 0 recs

The sample size issue wouldn't be too large, I don't think.

The selective sampling issue is probably larger – at that point you’re looking at players who played in three consecutive players, who are (typically) better players. That’s going to shrink the spread between players and push the correlation up.

But you can use the basic regression equation for any number of chances. After two seasons, I think you’d use roughly 75% a player’s actual UZR rate and 25% of the league mean.

I has a blog. | Ask a sabermagician!

by cwyers on Jul 4, 2009 9:26 PM EDT up reply actions 0 recs