It does take away some of the simplicity. This was just an experiment for the time being and I don’t see implementing in the regular stuff.

When you change the run environment, you’re messing around with each situation so even the starters/relievers may not be in the exact same order. They should still be pretty close.

The run environments are not currently not park adjusted, but will probably be soon. However the league run environments are set at the correct level.

By changing to run environment, are we saying that instead of comparing K-Rod’s %chance to give up a HR to all pitchers (say HR% of 6%) we’re comparing K-Rod to all relievers (say HR% of 4%)? Ergo, when K-Rod gives up a HR he loses WPA in comparison to what the average RELIEVER would have done vs. the average pitcher.

An adjustment like that makes sense to me, but it does destroy some of the “simple elegance” of WPA.

P.S. I notice that in the 2006 (raw) WPA, Johan has 4.12 WPA and Oswalt has 3.68 WPA, whereas in the revised WPA Oswalt has more WPA than Johan. (See also Halladay vs. Carpenter).

Is this some sort of league/park adjustment?
Follow-up: Is WPA league/park adjusted?

Name                 pWPA
Roy Oswalt           5.10
Johan Santana        4.90
Joe Nathan           4.43
Francisco Rodriguez  4.42
Brandon Webb         4.38
Jonathan Papelbon    4.26
Chris Carpenter      4.06
John Smoltz          4.04
B.J. Ryan            3.89
Roy Halladay         3.86
Bronson Arroyo       3.74
Chris Young          3.68
Nate Robertson       3.56
Anibal Sanchez       3.52
Derek Lowe           3.42
J.J. Putz            3.41
Barry Zito           3.29
Takashi Saito        3.27
Jason Schmidt        3.25
Jered Weaver         3.18

The starting pitchers are more valuable than before while the relievers are somewhat less valuable. 6 of the top 20 are still relievers opposed to 11.

And Mariano Rivera controls his leverage index (LI) as much as Adam Everett controls that he’s getting balls hit to him at SS: the great player will be placed in a position by his manager to maximize his skillset. This means putting Everett at SS, making 4 plays a game, not 2. And this means putting Rivera when the LI is 2.0, not 1.0. In both cases, the players were able to double the impact of their skills by being the great player they are, but letting their managers control how they are used.

There are two things:
1 – The starter/relief problem is that it’s easier to pile up good stats as a reliever than as a starter. The REW for a reliever would be something like +.020 wins per 9 IP (i.e., .520), while that of a starter would be -.010 wins per 9 IP (.490). But, of course, the average starter is a far better pitcher than the average reliever. It’s just that starters have one hand tied behind their back (they have to pace themselves).

2 – You can consider “chaining” in that if Mo is not pitching when the LI is 2.0, someone else will. So, rather than give him the full impact of the 2.0, he gets something in-between 2.0 and 1.0.

However, none of this takes away from WPA or REW. Those numbers are what they are, just as a HR hit in Coors is what it is. You need to apply the necessary adjustments if you want to make comparisons.

If you want to take the context out of of WPA, you can look at BRAA for pitchers (or REW, same thing really). That’s how many runs/wins they “saved” over the course of the season.

David Gassko suggested a middle ground to me where I raise the run environment for relievers, and drop it for starting pitchers. Just for grins, I’m messing around with that today.

Is there anyway to fix this “flaw”? Take out run support from WPA?

[For the record, I don't think it could/should be touched, but there's no way you can use WPA for comparing a SP to an RP or a pitcher to a hitter.]