from Amazon
Saturday, May 31, 2008
Do speedy outfielders (and those with good arms) save extra-base hits?
I’m talking about by cutting off hits in the gap (or wherever) and by the threat of a strong and/or accurate arm. As far as I know, none of the defensive systems takes this into consideration. UZR and my “arm lwts” do not.
I looked at 5 years of data. I grouped all outfielders into 3 groups, according to speed scores, slow, average and fast, where the slow and fast groups are each about 20% of all players. In each park and in each field, LF, CF, and RF (in order to control for park and field), I summed the difference between the lwts value of all hits for each group and the average lwts value of a hit to that field in that park.
IOW, let’s say that in Safeco, all hits to LF had an average lwts value per hit of .60 (where s=.47, d=.78, and t=1.03). If hits with speedy outfielders in LF at Safeco had an average lwts value per hit of .59, then the speedy outfielder group would get credit of .01 runs per hit, weighted by the number of hits to left field with speedy outfielders in the field. So on and so forth for all fields, all parks, and all three groups of outfielders.
The final result and answer is that yes, speed in the OF results in fewer extra base hits, but not by a whole lot.
If I break it down by field:
LF
Difference between fast and slow players (the 20% fastest and 20% slowest) is around 2 runs.
CF
3.9 runs
RF
2.4 runs
If I isolate the fastest and slowest 5%, the difference is around 4.8 runs in CF, 2.7 in LF, and 3.6 in RF.
So I think it is fair to say that the difference between the fastest and slowest players in terms of cutting off extra base hits in the gaps is plus or minus 3 runs per 150 in CF, and plus or minus 1.5 to 2 runs at the corners, which isn’t wood I guess.
UZR does not include this, and I don’t think any of the other defensive metrics do either. Of course, another way to do this for individual players is to look at the average hit value in each section of the field and compare that to the average hit value for each section when that player is on the field. I think you are going to get too much noise using that method, and I am not sure that all of the databases track where the ball lands as opposed to where it is fielded. I like the idea of just using an outfielder’s speed rating and adding a constant to his UZR or other defensive metric.
What about the same thing using “arms?” Players with better arms should be able to hold players to fewer extra base hits. Again, my “outfield arm linear weights” does not keep track of this, only of “holds” and “kills.”
For the best and worst 8% of the arms in CF, the difference is 1 run (per 150 games). In LF, it is also 1 run. In RF, it is 2.4 runs. Again those are differences between the best and worst 8% in “arm”. So, in CF and RF, it is plus or minus .5 run per 150, and in RF, it is plus or minus 1.2 runs per 150.
As I said, I think this is the first time anyone has quantified this, but I am not sure.