Graph of the Day: Range Factor

by Jack Moore on Aug 4, 2009 10:00 AM EDT in Visuals Above Replacement

Yesterday, I posted a graph that showed quite conclusively that Fielding Percentage is a meaningless statistic when talking about defensive impact on runs allowed. As commenter CoachOfEarl brought up in the corresponding thread, a slightly more advanced statistic called Range Factor (RF) also has disagreements with UZR. We see this specifically in the case of Adam Jones, who has the best RF in the majors and yet has a -13.1 RngR (the range component of UZR).

Range Factor can be calculated as either RF/9 or RF/G. In this case, I will use RF/9, which is defined as follows:

Range Factor rewards chances. Outfielders who play for fly ball heavy pitching staffs will have high range factors. Infielders who play for ground ball heavy pitching staffs will have high range factors, and vice-versa. This begs a few questions. Does RF measure defense better than Fld.%? Does RF even properly measure range?

With that in mind, I took the quick step of comparing UZR and RF, as well as comparing RngR and RF. The graphs would get squished and distorted before the jump, so please follow the jump and take a look at what kind of correlations we see (or don't see).

First, let's compare UZR and RF.

Interestingly enough, we see a slightly (but not significantly) negative correlation. This isn't terribly surprising, though, given that RF completely ignores the Error, Arm ,and Double Play components of UZR. We understand that, and that's part of Range Factor - it's not designed to answer those questions. Still, we should expect RF to correlate positively with RngR. Let's see if the expectation holds.

Interestingly, it does not. In fact, the correlation is even more strongly negative. I'm not sure I can exactly answer why that is without looking into the backbone of UZR. However, I will say that given the simplicity of RF, I will certainly trust UZR and it's measurement of range, which credits plays out and on the edge of a players zone much more than RF does.

Discussion is welcome and encouraged. What's the next step?

Tweet Comment 13 comments | 0 recs |

Do you like this story?

Comments

I have actually created a defensive metric using RF and E that has a range from -20 to 20

It generally picks out the good and fielders, but there are some definite inconsistencies with UZR. JJ Hardy for one that I have much worse using RF vice UZR. I created mine to look at older players.

Jeff Zimmerman - Protecting the world from RBI's and Wins from my mom's guest house.

by Jeff Zimmerman on Aug 4, 2009 10:35 AM EDT reply actions

I will say that Hardy definitely passes the eye test.

---
http://www.beyondtheboxscore.com
http://www.rightfieldbleachers.com

by Jack Moore on Aug 4, 2009 10:42 AM EDT up reply actions

I agree, that is why I am not confident with it. It is on another computer and I will post it later.

Jeff Zimmerman - Protecting the world from RBI's and Wins from my mom's guest house.

by Jeff Zimmerman on Aug 4, 2009 11:23 AM EDT up reply actions

The problem with Fld% and RF...

…is that it doesn’t account for opportunities in any meaningful sense. You have to look at team BIP at least in order to get it to come out to something meaningful.

I has a blog. | Ask a sabermagician!

by cwyers on Aug 4, 2009 1:05 PM EDT up reply actions

So what is your opinion on Adam Jones...

Is UZR mis evauluating him?

by Markakis and Wieters 4 Life on Aug 4, 2009 12:56 PM EDT reply actions

If I had to bet on either stat mis-evaluating him, I'd go with RF.

A quick glance at his career stats (250 career games, roughly, still not a whole lot to go on) has him at +11 with the arm and -8 with range. I don’t know, really, I haven’t watched him enough.

---
http://www.beyondtheboxscore.com
http://www.rightfieldbleachers.com

by Jack Moore on Aug 4, 2009 2:01 PM EDT up reply actions

To reiterate from the last thread...

…range factor doesn’t have the sense that God gave a dog. That doesn’t mean that UZR is right, necessarily, but range factor shouldn’t enter the conversation at all.

I has a blog. | Ask a sabermagician!

by cwyers on Aug 4, 2009 2:12 PM EDT up reply actions

Some quick thoughts

(I’ve been swamped doing analysis for the people who pay me to do it…)

Orioles pitching FB% is 3rd highest (40.0%). League spread is 32-41%. Ahh, but their IFFB% is also 3rd highest (11.5%). They give up the 3rd most HRs. (So many correlations!) So what we’re really looking for is OFFB%, which I don’t have time to put together right now.

One would have to assume that the distribution around the outfield would be the same for each team, or close. This may not be correct.

UZR RngR might not be really measuring just range, but positioning effectiveness. Having a negative might nto mean you don’t have range, but you’re playing far from where the ball is being hit.

I’m not sure why assists are included in RF for OF…they measure arm, not range.

You really can’t compare different OF positions with RF. A CF has many more chances. The only thing I find RF useful for is as a position relative metric, and it should be taken with a couple shakers of salt.

Where are you getting your RF numbers from? Fangraphs has AJ as the highest at 3.1, but the range here is 4-4.3.

"I don't like to be here and just thinking about in October I'm going to go on vacation " -- Melmo moaning about being benched when he's hitting .256/.321/.330

by CoachOfEarl on Aug 4, 2009 6:02 PM EDT reply actions

This is at a team level.

And yes, fangraphs.

---
http://www.beyondtheboxscore.com
http://www.rightfieldbleachers.com

by Jack Moore on Aug 4, 2009 7:42 PM EDT up reply actions

Your third point makes sense to me.

I’m not sure how to describe it without using diagrams though, and I’m kinda tired right now.

@bs_uf15bosox9be The Original Gameday; Learn to use SB Nation

by bdalebs on Aug 4, 2009 11:27 PM EDT up reply actions

What is wanted is a batted ball hit/out chart

And/or access to the data that would generate it ;)

"I don't like to be here and just thinking about in October I'm going to go on vacation " -- Melmo moaning about being benched when he's hitting .256/.321/.330

by CoachOfEarl on Aug 5, 2009 10:09 AM EDT up reply actions

That would work.

I was talking more about the general concept of positioning. If we imagine a player’s range (for a certain time period that a ball is in the air) as a circle, if he’s not playing in the correct spot (where balls tend to fall most offen), then he’s missing a lot of balls that RngR assumes they should have been able to get to if they were positioned correctly.

@bs_uf15bosox9be The Original Gameday; Learn to use SB Nation

by bdalebs on Aug 5, 2009 7:16 PM EDT up reply actions

A problem with UZR is that it is meant to be used to find how much a player is above or below league average for a single season, not season to season.

Say you have a shortstop that makes the exact number of good and bad defensive plays in two different seasons. He is the exact same defender. In the second year, the top 4 shortstops all get injured. These 4 are all replaced with below average defensive shortstops. The original shortstop’s UZR value will then go up in the second year because the average league ability went down even though he fielded the same.

Jeff Zimmerman - Protecting the world from RBI's and Wins from my mom's guest house.

by Jeff Zimmerman on Aug 5, 2009 12:58 PM EDT reply actions