Arm ratings in defensive metrics look at three things: holding runners, assists, and kills. It stands to reason that the runs saved attributed to an outfielder would be greater for killing the runner as opposed to being an accomplice. Being an accomplice to the out would then be more valuable than holding a runner. A concern on how well runs saved represent the talent in a player's ability to throw a ball is whether or not he has the opportunity to show off that talent. In other words, if base runners fear an outfielder's arm then the outfielder will be given fewer opportunities to wipe the runners off the base paths as a function of the runner's hesitancy to test the arm. However, based on my conversations with baseball folk and through my own work on assessing how a pitcher's fastball velocity changes as he ages, arm strength should be relatively consistent through the majority of a player's starting career. If the metrics trying to represent the value of a player's arm actually coincide with how talented that play is in throwing the ball, then you would expect a flat line. Arm value should remain constant as a player ages if his arm quality remains the same.
In this post, I wanted to look at something simple. A measure of talent is typically a good measure of talent if the measure is consistent. That if you have an 80 arm in year 1 resulting in 20 runs saved, then, if the metric is strongly related to talent, in year two the 80 arm should result in another 20 runs saved. I decided to take an elite group of arms and observe how their runs saved attributed to their throwing changes from year to year. Now, this is a simple study with a population of only ten, but it may serve as a decent launching pad for further discussion.
The list below are the top ten cumulative arms using rARM (DRS methodology) from 2005-2012 using their first five seasons and by defining their first season as the first year they achieve 900 innings in the outfield. Both Adam Jones and Nick Markakis are in this grouping.
Here is the same list, but with the ARM metric from the UZR methodology.
rARM - DRS Year 1 2 3 4 5 Jeff Francouer 3 10 2 4 8 Shane Victorino 4 5 3 -1 3 Adam Jones 5 10 5 8 -1 Alex Rios 7 4 8 7 8 Alfonso Soriano 7 14 4 1 -4 Hunter Pence 0 8 6 -1 5 Jayson Werth 3 3 5 3 -1 Melky Cabrera 3 9 1 0 4 Nick Markakis 2 2 10 2 0 Matt Kemp 8 7 -4 4 4
I took the sum of both the rARM and ARM metrics by year as well as those metrics normalized to what would be expected over 1400 innings. What I mean normalized or adjusted values would be the following: if player A saved 10 runs over 1000 innings then he would be projected to save 14 runs over 1400 innings. Whether right or wrong, I did this to cut down the variation in the numbers above by changing the counts into rates. If there is an issue with doing this, then I am sure someone will be kind enough to inform me.
ARM - UZR Year 1 2 3 4 5 Jeff Francouer 3.5 16.6 2.5 5.3 9.7 Shane Victorino 5.3 3.9 3.0 -1.0 2.2 Adam Jones 3.3 6.6 2.5 5.4 1.7 Alex Rios 11.7 6.5 5.6 6.0 5.6 Alfonso Soriano 4.9 14.3 4.8 -3.5 0.5 Hunter Pence -0.1 8.2 6.6 -1.0 3.5 Jayson Werth 4.2 4.4 3.3 5.4 -1.4 Melky Cabrera 3.4 4.4 1.1 -3.9 3.0 Nick Markakis 1.1 2.7 6.7 0.9 1.1 Matt Kemp 8.0 4.3 -6.1 4.0 3.4
The difference between the cumulative runs saved looks significant. I decided to run an ANOVA on the adjusted Runs Saved for the population (not cumulative), which resulted in significant p values for both DRS (p=0.04) and UZR (p=0.01). Further analysis of both metrics indicated that year 2 performance was significantly greater than years 3-5. This is illustrated below.
The alternative hypothesis would be playing into the early defensive peak. It may be that around what normally would be the second year of a player having a starting job. The narrative here would be that a player's ability peaks early in his career and then tails off. That certainly is the case with range and it may well be that the opportunity to kill a runner has as much to do with a player's arm as it does with a player's range. It may be that by not being able to put himself in a position to get to a ball quickly that the fielder is simply reducing the number of chances he has to impact base runners. Additionally, it may be that ability in range or route running may decrease, resulting in the player putting himself in worse position for a throw. That would be a situation that likely would have more affect on accuracy than arm strength.
Simply put, I have questions. Perhaps, as my post on UZR in the Camden Yards outfield led to a great deal of words being hashed out, including by John Dewan in his Fielding Bible, maybe this post will also launch a thousand blogs just the same.