Continuing with the theme this week of defensive metrics their interpretations, today's graph examines the ErrR, or Error Runs, component of UZR on a team level. We should expect a team's unearned runs (Runs - Earned Runs, of course) to correlate strongly to Error Runs on a team level. In order to gain a larger sample size, today's graph uses whole team data from the 2006, 2007, and 2008 season. Let's take a look at the data.
Today, common sense actually holds water, which should be pretty unsurprising given that we're using an advance metric like UZR in this case as opposed to the primitive Fld. % and Range Factor. Still, although a .547 correlation coefficient shows strong correlation, I personally expected stronger. Here are my explanations - maybe some of those with more knowledge on the inner workings of the statistic can confirm or deny.
First of all, take a look at this part of the definition of the earned run:
To determine whether a run is earned, the official scorer must reconstruct the inning as it would have occurred without the errors (for purposes of this rule, the "errors" also include passed balls). The benefit of the doubt is always given to the pitcher in determining which bases would have been reached by errorless play.
So this means that continuations of innings after three outs would have been made could potentially result in a lot of earned runs. For example, if there are two outs in the inning, and then an error is made, then every run that scores after the error is unearned. I don't believe that ErrR would account for this, since it is linear weights based (nor do I think you would want it to).
Secondly, and possibly related, it seems like the correlation breaks down at the ends. Teams on the edges don't seem to trend towards the line. Teams that have worse pitchers allow more unearned runs, regardless of defensive talent behind them, and similarly, better pitchers allow fewer unearned runs.
Any other ideas or discussion are welcomed and encouraged in the comments section.