EDIT: Thanks for BobbyMac in the comments for pointing out a flaw in my data. It's not quite as much fun as the original results, but it's still pretty conclusive. Expect a couple more graphs
later today tomorrow, too.
Listening to broadcasts crews and the mainstream media talk about defense can be painful. I think most of the broadcasters understand the basic concepts behind what makes a good defense. A good defender has good instincts, positions himself well, gets a good first step, throws hard, can make plays on both sides as well as right at him, has good range, and makes plays on balls that he gets to.
Unfortunately, the one stat that many people refer to plainly as "defense" is fielding percentage. Fielding percentage is an attractive statistic, much in the way batting average is. The formula is simple: Fld% = Errors/Total Chances. As such, the statistic is too simple to measure the complexities of defense. It only measures one of the criterion mentioned above - making plays on balls a defender reaches.
UZR is, so far, our best attempt at quantifying all these criteria. Right now, UZR is the sabermetric communities best attempt at quantifying run prevention by fielders. As I showed way back before I was an author here, there is a very significant correlation between UZR and the run differential we see between FIP and actual ERA. As UZR as well as our advanced pitching statistics are fine tuned, I would expect
Still, one would naturally expect that fielding percentage would strongly correlate to run prevention. Batting average, despite all its flaws, still has a very significant corrleation to runs scored. It's just not as good as OBP, SLG, OPS, or wOBA. Does fielding percentage fit this mold? Is it just the limited first step in a line that leads up to UZR? The following graph will reveal the truth.
Using the data from this season, which is more than 8000 innings of defensive data per team, I've plotted our two statistics in question. UZR, we already know, is directly related to run prevention. So if fielding percentage is also direclty related to run prevention, we should expect at least some sort of positive correlation between the two.
In fact, we no correlation whatsoever. That means that we cannot say anything about how good a team is defensively, even if we know its fielding percentage. I suspect that this relationship is as such because of the inherent flaws of fielding percentage, the biggest of which being its complete and total inability to measure range. Many of the balls that the teams with the high UZRs are reaching (and those with low UZRs aren't) are very tough plays even after getting to them. Scorers may see these players with great range get to these balls and expect them to make plays that other fielders wouldn't have a chance at. This is not to say that a high fielding percentage is bad, but merely that the subjectivity of errors leads to many plays being incorrectly labeled as poor.
This is what I call the Nate McLouth factor. This is why Nate McLouth can win a gold glove despite being one of the worst defensive center fielders in the game. He is merely the most recent example. This concept is not new. But it is an important starting point when one tries to improve his or her knowledge of what to look for in a defender.