View Full Version : Defensive Metrics

07-27-2006, 08:54 AM
Several people reciently have posted comments that defensive metrics weren't very good. FCB humerously posted that it would be best to forget them all together and save yourself the trouble.

What are the short commings of these metrics? Is there something inherrent to defesense that makes it hard to measure in statistical terms?

This is a serrious question, not a swipe at stats. I'm currious about the methodoligies that go into defensive metrics and why they aren't as usefull as offensive ones.

07-27-2006, 09:04 AM
Is there something inherrent to defesense that makes it hard to measure in statistical terms?How do you measure plays that don't get made due to a lack of range? That seems to be one of the main issues.

the 3 defensive metrics used most often

FLD% - plays made/chances
RF - # of plays made per 9 innings
ZR - % of plays made within your defensive zones

They each have serious shortcomings. I try to use the metrics in combination with subjective observation to get a more accurate picture.

I would suggest reading Dewan's Fielding Bible.

07-27-2006, 09:38 AM
Fielding percentage is easily the goofiest of all stats. Far as I'm concerned it has no use.

Range factor suffers from an apples to apples problem. It's impossible to compare the RF of guys who played in different parks behind different pitchers. It has some use in comparing guys at the same position on the same team, but even then you need to be careful about outliers and bogeys. It's best used for large sets of data and sweeping generalizations. For instance, for six season running, every other player the Reds have stuck in CF have made a ton more plays than Jr. After seeing it happen consistently year-in, year-out, you can safely say that Jr. isn't a very good fielder.

Zone rating has its adherents, but I knew a sportswriter who used to tabulate it and it had a a major cognitive glitch to it. Yes it dices up the field into different sectors and then assigns responsibilty for those sectors to the fielders, but it was left up to human perception as to whether a given ball was within a player's sector. That may have been fixed in recent years, but before that you could actually get punished for showing range and almost getting to a laser while another guy with less range would watch it from afar.

Positioning also could screw up ZR. Say you traveled 20 feet to make a play at SS while another guy made a play in the exact same spot, but only traveled 12 feet. It's not an equivalent play, but it's treated as such. The hope is that such things normalize over larger numbers and that the guy who goes more distance eventually makes more plays, earning a higher ZR. That's always been too much of a leap of faith for me. My take has always been that ZR is probably missing some important functional range differences.

Some of the newer defensive stats are much more painstakingly derived, weeding out human perception as best they can and attempting to get at the hard-measure-but-ultimately-key component of range (which is by nature a distance over time equation). Bill James probably took traditional fielding stats as far as they can go when he developed his defensive rating system for Win Shares. Dewan's, which Ricardo mentioned, has produced some illuminating numbers. Baseball Think Factory's recent 2006 number follow in the Dewan's vein. Ultimate Zone Rating had its adherents, but the guy who did it got hired by an MLB organization and those numbers are no longer made public.

07-27-2006, 12:08 PM
...but the guy who did it got hired by an MLB organization and those numbers are no longer made public.

There are at least 7-8 clubs (off the top of my head the A's and Bosox) that have stated at one time or another they have their own method of creating defensive statistics. We will likely never see those.