Troy Percival (astonishingly) held opponents to a .230 BABIP over 700+ innings of relief. Considering his great (fortune?/skill?) on balls in play, and that he additionally struck out guys at a pretty good clip, and that he didn’t give up an insane amount of walks/homers (though a fair amount of each), one would think he’d have managed an ERA even better than his career 3.17 mark.

Surely you don’t really mean there is no evidence that pitchers have control over the type of contact they induce? GB/FB rates certainly have some stability, even over the course of a career. For example, Hoffman was consistently a flyball pitcher on balls in play.

A pitcher’s park won’t even out over the course of a career (a big park would certainly have played to Hoffman’s strengths). Nor will the contextual effects of league and era (In this case, though their careers overlapped, Smith played in a weaker offensive era than Hoffman, no?). But I was careful to say that “some” things would even out—namely the pure luck aspects.

Basically, my goal here is for neither of us to throw out the baby with the bathwater. Because WAR takes into account things like ballpark effects, it should not be dismissed. Then again, if you’ve got the time (say, you’re voting for the HOF), why even limit yourself to such a coarse summary statistic, right?

Anyway, I hope we both got something to mull over from this exchange.

I believe that is still accurate. All implementations of the framework are going to be better at some things than others. I’m not going to argue that rivera is underrated according to his WAR. He is an outlier. The rest of the relievers cited don’t have those same issues. Some stats may even out over the course of a career, but if a pitcher pitches half of all his innings in a pitcher’s park, that will not even out. If there is evidence that pitchers have control over the type of contact they induce and that they can sustain over the course of a career, I have not seen it. There are going to be outliers going both ways over the course of a career with great pitchers on either side of it. I feel more comfortable relying on ks, bbs, and homers not just because it is predictive, but also because it is descriptive of the events that the pitcher has the must control over. How much control a pitcher has over the type of contact I’m not sure about, but I’m very confident when there is no contact.

As in the case of rating HOFers, we want to *describe* what happened in a given season, rather than infer the player’s underlying skill. We give credit for what a player did, just like we award the World Series trophy to the team in the series that gets the most wins, not the most Pythagorean wins.

Going with WAR, I think, can really undermine the spirit of the MVP. Consider two players: Player A is an absolute terror at the plate, but only when the game is entirely out of reach in either direction. In situations where good hitting can actually effect the outcome of the game, he strikes out virtually every time. Player B strikes out virtually every time in meaningless situations, but is an absolute terror when it counts. Let’s say both these players finish with identical stats. WAR will say both these players are equally valuable, but it is evident that player B has contributed much more to the success of his team.

WPA tends to be highly correlated with WAR—you don’t have to be superhumanly clutch to have a very high WPA—but I think it adds something highly desirable.

2. WAR takes into account ks, bbs, and homers, but it doesn’t take *everything* meaningful into account. It makes assumptions that pitchers aren’t able to to systematically induce weak contact (leading to lower BABIP) and that “clutchness” doesn’t exist. It doesn’t account for the pitcher’s own defensive abilities, nor his ability to keep runners from advancing on the basepaths via steal or wild pitch. You’ll notice that Mariano Rivera, who has had all of these fine qualities, routinely outperforms his FIP, and therefore is undervalued by WAR.

You can make the argument that ks, bbs, and homers are important because they tend to correlate well year-to-year, while other stats don’t. Therefore, as tools for *projection* people have had success with them.

That doesn’t mean that when a career’s player is finished you aren’t allowed to consider anything else, especially considering that a lot of that pure luck is going to even out over a 15 year career. WPA is more of a descriptive statistic than a projective statistic, but when we’re looking back at a player’s career we’re trying to describe, not project. You’re allowed to entertain the possibility that a player outperforming his K, BB, and HR rates might not have just gotten lucky for 15 years, and that maybe give that player credit for at least some of the other variance contributing to his performance.

You make several very good points, especially in relation to the highly leveraged nature of relievers. I also agree that looking at the lowest level of hall of famer for comparison is generally not a good idea, although with hoffman, smith, and jones, we are dealing with non-hofers. A couple of points though.

1. WAR, I believe, does take into account, leverage index for relievers. Those relievers do get credit for pitching in more important situations.

2. WPA is a good tool, but it doesn’t tell you how the outs are made or how the runs are given up. WAR takes the most important factors, ks, bbs, and homers. WPA is not aware, nor does it consider park factors or defense that the pitcher has no control over.

Lee Smith may be a marginal HOF case statistically, as many players are, for his regular season performance. But If I’m on-the-fence, I’d rather give the nod to a player like Curt Schilling, or Chipper Jones on the basis of what they’ve truly mean to the game of baseball over the last couple of decades. I’m sure Lee Smith had his moments, but he’s no icon in any of the 8 cities he played in, and he never left his mark in the postseason.

But on the other hand, a properly-utilized ace reliever will be injected into games in the most highly-leveraged situations, where the difference between allowing 0 and 1 runs can be the difference between winning or losing. I don’t think it’s naive, archaic, or sabermetrically-challenged for me to make the claim that superior performance in these situations provides value not captured by WAR.

Since a reliever’s perceived value is almost entirely tied up in performance in highly-leveraged situations, if you use a metric that is intentionally agnostic to this context, of course it’s not going to capture the reliever’s full value.

I encourage you to at least consider WPA when evaluating relievers. I’m pretty sure WPA only takes into account a position player’s offense, so take this with a grain of salt, but Lou Whitaker’s career WPA of 28.01 now seems like a saner comparison with the likes of Goose Gossage (31.40) and Trevor Hoffman (32.98).

Being highly effective for 1-2 highly leveraged innings is the hallmark of modern reliever effectiveness. WAR doesn’t care at all about leverage, it just cares about quality (performance, specifically FIP) and quantity (innings).

If you look at career WPA, Hoffman (32.98) and Gossage (31.40) soundly defeat Lee Smith (23.97). Rivera trounces them both (54.70).

Sutter sits at 19.61 career WPA. I think that this stat, combined with the WAR data you have presented, makes for a pretty compelling case that perhaps he does not belong in the HOF. But in terms of Lee Smith’s consideration, I think it’s a big mistake when you are thinking about the HOF to always be comparing a player to the *worst* current HOFer at his position. If you don’t think player A deserved to be in the HOF, you shouldn’t vote for player B just because player A wound up getting in.