THE BOOK--Playing The Percentages In Baseball

Everyone assumes that the choices are:

A) Joba is a starter
B) Joba is a 3-6 out bridge to Rivera

I wonder if the optimum usage is as a long-relief closer? Lots of high scoring games this year, especially in Yankee Stadium. That means higher pitch counts and more relief innings. Why not use Joba every 3rd, 4th, or 5th day (as the situation warrants) to come into games in the mid innings of close, high-scoring games to pitch the last 4 or 5 innings?

For that matter, why not turn Wang or Hughes into that pitcher?

Mark Eichorn, 1986:
http://www.fangraphs.com/statss.aspx?playerid=1003712&position=P

69 relief
0 starts
157 IP
1.4 LI

157 x 1.4 = 220 effective IP in relief (or, if you prefer to use 1.2 instead of 1.4, then 188)

However, from 1988 onwards, he was a regular reliever.

I also quite enjoyed how Bob Stanley was used for most of his career:
http://www.fangraphs.com/statss.aspx?playerid=1012409&position=P

You’re NEVER going to get anywhere as an outsider, even as the greatest sabermetrician in the world, trying to figure out which pitchers should be starters and which should be relievers.  The reasons are, one, it is always going to be close, statistically, two, we really don’t know nearly the exact conversion formula for starter to reliever or vice versa, three, looking at an individual pitcher’s actual starter and reliever performance, if he has both, tempting as it might be, is going to be fraught with sample size issues, and just plain old noise in general, and fourth, and most importantly, it probably varies A LOT from pitcher to pitcher what their own, actual, and unique difference is going to be as a starter and reliever, based on their repertoire of pitches, their stamina, and other things.  That last, most important thing is best left up to the team’s manager, coaches, scouts, and front office - by far.  Whether they are capable of doing that, I don’t really know.  But we certainly are not.