Before I read the article, i wondered, will he distinguish between man on 3rd w/ 0 outs, and man on 1st w/ 2 outs?

Nope, just % Stranded.

Use that matrix of expected base/out run expectancies, and I'm with you.

Cant you just look at a relievers whip along with the ERA, if he is letting a lot of inherited runners score then he must be letting lots of folks on base to score them. Not perfect, but currently one can easily do this.

Doesn't quite seem fair to list Sisco, a rule v pick, along with Gordon and Timlin. He may well have the statistical profile; he may not be as good as his ERA says he is, but calling him a "fraud" seems a bit harsh.

Timlin? fire away...

Given enough innings, ERA (adjusted for park and league) tells us pretty much everything about a reliever we need to know. Their strand rate, even adjusted for bases/outs, over and above ERA, again, in the long run, is simpy random variation (i.e., luck). In the short run, of course, strand rate helps us to determine how "effective" a releiver actually was (in terms of helping his team actually win games or not). It doesn't really tell us how effective a pitcher he was (in a context neutral sense). For example, if a pitcher comes into a game with runners on first and second and two outs and allows a bases clearing doublem, he was not very "effective." If he comes in with 2 outs and no one on and allows a double, no problem. Both pitchers pitches the "same" but in terms of value in retrospect, one was way more valuable than the other. The problem with articles and stats (strand rate) like this is that they serve to further blur the line between past value and true talent/future value. People too often equate the two. For example, if pitcher A has a 3.00 ERA and a great strand rate and pitcher B has a 2.50 and a terrible strand rate, while pitcher A was probably much more valuable than pitcher B in retrospect, the pitcher you would want pitching for you in the present or future is clearly pitcher B. How much confidence you have that pitcher B will perform better than pitcher A in the future depends on how many innings those ERA's are based on, but in no way shape or form, does it depend on either pitcher's strand rate. It is disheartening (and inconsistent) to me to see sabermetric oriented people and knowledgeable fans clamoring for "inhereted runner" (strand rate) type stats for relievers and VORP-like stats for batters (for example to help determine the MVP). They are completely different types of stats. One is 100% context neutral (VORP) and does not necessarily have anything to do with how much value a batter actually produced for his team (how much he actually helped them win), and the other (an inhereted runner stat for pitchers) is just the opposite in that it gives you an idea as to how much value a reliever created for his team, at least in terms of runs allowed (although not necessarily in terms of wins). The equivalent stat to VORP for pitchers (relievers and starters) is an ERC (component ERA) turned into a runs above average or replacement saved. Again, VORP for batters or for pitchers or an ERC turned into a VOPR-like stat is strictly a context neutral "theoretical value" type of stat and nothing necessarily to do with actual value produced (although it will obviously correlate well with actual value produced, especially as you increase sample size). This is not the right thread for this, but using a theoretical, context-neutral stat like VORP or linear weights (absolute or rate) is ridiculous for an MVP award, for obvious (to me at least)reasons, yet, again, so many saber-types try and do this. Enough of my late-night rant...

At least during the 1st half of the season, Kent Mercker seemed to let every inherited runner score.

I'm surprised more GM's, mangagers, beat writers, and fans don't pick up on this.

Gagne, a few years back, was incredible. He seemed to come in in the 8th with runners on base all of the time. The Dodgers awful offense would provide him no margin of error; yet he racked up a ton of saves, a tiny era, & couldn't allowed many of the inherited runners to score. To me, he's one of the few closers with save totals that are somewhat meaningful.

Given enough innings, ERA (adjusted for park and league) tells us pretty much everything about a reliever we need to know.

Is this true? I think there are two big problems with ERA for relievers:

(a) partial innings. A reliever who enters a game needing only one or two outs to end an inning will have a lower ERA than a starter who begins every inning, assuming equal ability.

(b) end of game bias. A reliever who takes a game ending loss on the raod will typically leave a base-out state which would result in additional runs if the inning were played to conclusion.

Because of an a & b, I think you have to use ERC for an ability or context neutral stat for relief pitchers, as a counterpart to ARP or APR or whatever it is for RP that takes into account inherited runners scored.

DCA, sure, even in a large number of innings, if there were biases in terms of a and b, then ERA would not be as good as ERC, whereas with starters, ERA and ERC should always converge (as number of innings become larger)....

For example, if a pitcher comes into a game with runners on first and second and two outs and allows a bases clearing doublem, he was not very "effective." If he comes in with 2 outs and no one on and allows a double, no problem. Both pitchers pitches the "same" but in terms of value in retrospect, one was way more valuable than the other.

Except if there is reason to believe that the first pitcher in question is more likely to give up the double with the runners on than he is with the bases empty (or is more likely to give up that double than the second pitcher). There is a physical difference here that might result in performance differences in the two situations: the fact that the pitcher throws from the stretch w/ runners on base and from the wind-up with none on.

I suspect MGL won't like or agree with this, but I think that we should require relatively less statistical proof of something when we begin with a plausible physical explanation. Assume we have two observed statistical outcomes. For the first, all we have are the stats, and they seem to show that Slobodan Whifferoosky doesn't hit well w/ RISP. Since there is nothing we can observe in his swing that would suggest there might well be a difference, it makes sense to demand a LOT of data before we can attribute any significance to the statistical picture. The assumption that it's just a sample size fluke is relatively strong.

But when it comes to allowing inherited runners to score, there IS a difference between pitching w/ runners on and pitching with the bases empty. I would need relatively less data of Horatio Gameblower's torch-like outings with runners on base to believe that there is a difference there, and that he is having a problem being effective from the stretch.

(Obviously, this applies only to a pitcher who pitches from the wind-up w/ the bases empty. I know some relievers opt to pitch from the stretch all the time.)

It also depends on what you're trying to do with the data. Probably the only thing you might do if/when you drew a conclusion about Whifferoosky is to get rid of him or reduce his playing time. But with Gameblower, I'd task my pitching coach to work with him on trying to improve what he's doing from the stretch, or at least to explore whether that might be the explanation. And I certainly wouldn't wait until I had enough data (if I ever did) to reach statistically significance to do it.

Not only is there the question of pitching from the stretch, but there is the question of the type of outs that are generated, which are going to affect advancement of runners.

mgl -

how many innings do you need for statistical significance before you agree that a pitcher lets fewer runs score if he comes in with bases empty?

(let's suppose he always pitches from the stretch so there's no question about windup vs stretch)

The first comment says it all, IMO. Unless we know the situations in which the reliever inherited runners, just looking at inherited runners scored is pretty pointless.

Personally, I'd like to see the rules with regard to who is charged for a run changed to reflect who is most to blame, not who let the guy on. If a reliever comes in with two outs, anyone on first or second should be charged to him. If he comes in with one out, anyone on first should be "his" run. That would make both the starter's and the reliever's ERA more accurate reflections of their contributions.

Oh, and for those worried about effectiveness with runners on vs. bases empty, it's pretty easy to find those splits (ERA and WHIP with runners on, vs. bases empty), isn't it? Given a large enough sample size, it should be pretty easy to determine if a guy is more or less effective with runners on - and it shouldn't make a difference whether it's "his" runner or someone else's.

Personally, I'd like to see the rules with regard to who is charged for a run changed to reflect who is most to blame, not who let the guy on.

God knows that I don't know all your alphabet soup of stats, but doesn't BPro do this with one of their stats.

BL

yes it's either ARP or APR I think. I really like BP's reliever stats, but they cuold use some better branding.

Given enough innings, ERA (adjusted for park and league) tells us pretty much everything about a reliever we need to know.

In fact, the more I think about this, the more I think it is false in a number of ways. In addition to what is written above, consider Joe Slug and Billy Base, two relief pitchers. Joe puts very few guys on base, but gives up a lot of home runs. His "true" ERA is 4.00. Billy keeps the ball in the park, but allows rather more baserunners than Joe, so his "true" ERA is also 4.00.

Not only are Joe and Billy very different pitchers, but their differences are crucial. Bases loaded, tie game, top of the fifth, you don't bring in Joe. The very reason Joe has a 4.00 ERA is because he doesn't "naturally" face a lot of bases loaded situations like this, where his tendency to give up the home run becomes fatal. Billy is the choice in this situation. In fact, Billy is generally a better relief pitcher, because he is more effective with runners on base. Joe might be useful though in do-or-die situations.

There is no magical number of innings or PA's for "significance". The more the merrier.

Yes, pitching from the stretch versus the windup is in fact a "skill" (there are tru variations among pitchers, like the L/R platoon advantage) There is a section in our book that looks at that.

However, like many things that are a skill, you cannot "find" it by looking at small subsets of data. Sure, if you have oodles of innings (thousands) for a pitcher, his men on/no one on splits will reflect his unique "pitching from the stretch versus from the windup" skill. However, one season of an "inhereted runners" stat is going to tell you NOTHING more (actually less)than an ERA does about that pitcher's true skill.

Of course, since nearly all (I don't know any that don't) relievers (at least the short ones) always pitch from the stretch, the windup/stretch skill thing is not relevant. Whether pitchers have unique splits with respect to pitching with men on or without men on (both from the stretch), I don't know. I doubt it...

Two Alous, sure, ERA does not tell you everything you want to know in terms of deciding whom to bring in and when. I mispoke. As you suggest, you can certainly leverage your relievers to take advantage of their "types" of ERA's (as you say, is it HR heavy, walk heavy, etc.). That is something that managers probably do not do enough...

Smoltz used to use the windup when he closed.

Two Alous, sure, ERA does not tell you everything you want to know in terms of deciding whom to bring in and when.

But what about from this angle - what if the reliever who allows the inherited runner to score is replaced by a more effective pitcher? Say a middle reliever being replaced by a setup man or closer? The more effective pitcher may keep the crappier pitcher from being charged with any additional earned runs.

I would think this issue would come into play quite often, and would especially effect relievers since they naturally have much smaller sample sizes than starters.

Any thoughts?

To clean up what I said above -

Middle reliever comes in, allows inherited runner to score. Leaves game with runners on base (presumably one or more of them were put there by said reliever).

Next reliever comes in, does not allow any more runs to score.

How often, given how there seem to be a lot of bullpen mgmt issues in MLB, would the "next reliever" be a better reliever? I guess a lot of that would depend on leverage (if the game was close, chances are a better reliever would be brought in).

I find it hard to believe that a pitcher couldn't consistently have his ass saved by having a better reliever brought in behind him. This WOULD impact his ERA positively, and given the significantly smaller # of innings that relievers pile up, could greatly skew things for quite a while.