Hardball Times - Outfield Arm Ratings (3/15)

[JinAZ]

New article today on the Hardball Times with new data on outfield throwing arms. Interesting stuff:

Leftfielders and Centerfielders (15 March):
http://www.hardballtimes.com/main/ar...nd-left-field/

Rightfielders and explanation of method (from 21 Feb):
http://www.hardballtimes.com/main/ar...outfield-arms/

Dunn, Griffey, and Kearns all rated as essentially average in terms of the runs saved via "kills" (assists from outfield) and holds (preventing an extra base). Kearns showed up in the leaders for kills, though was slightly below average in holding runners, resulting in the overall rating of average.
-JinAZ
...gotta love any statistics given in terms of kill rates.

[RedsManRick]

Judging outfield arms statistically is difficult b/c you can't easily account for the decision of the runner to advance. There is some pretty substantial selection bias. The guys with the best kill rates are likely those with the best "good arm to bad repuatation" ratio. Guys just don't run on Ichiro, Vlad, Guillen, etc. and those that do run are probably more likely to be the speed demons.

[elfmanvt07]

I don't think anyone who has seen Dunn throw can deny his amazing arm strength. There are some things that stats cannot account for.

[Johnny Footstool]

Guys just don't run on Ichiro, Vlad, Guillen, etc. and those that do run are probably more likely to be the speed demons.

If you look at the charts in the article, Ichiro and Guillen had amongh the highest kill rates, indicating that guys do run on them and get nailed in the process.

Guerrero has a high hold rate, indicating that runners don't like to chance it when he has the ball.

[JinAZ]

If you look at the charts in the article, Ichiro and Guillen had amongh the highest kill rates, indicating that guys do run on them and get nailed in the process.

Similar findings were made for Andruw Jones in Centerfield. The metrics that Walsh uses in his articles are adjusted relative to the opportunities that the outfielders had, and are apparently based on play by play data. Using league averages, he is able to generate an "expected" number of holds and kills for each player given their opportunities, and then see how players stack up to those expected values.

So it's better than just looking at straight assists and such, though I'll agree with Rick that there is still the potential for changing on-field strategies to influence these measurements (i.e. you'd be more likely to run on a guy with a great arm *only* when you're very likely to make it). I just don't see much evidence for this being a problem in his data.
-JinAZ