"As I said, I love to be proven wrong, since that gives me an opportunity to learn something. However, after all the bashing I incurred at BBTF, I think it only fair for them to see this new research. Not that I wouldn’t get bashed again. After all, it is BBTF."

I dunno. Maybe if the first post hadn't had a sneering "If you haven't done the research like I have, you have no right to have an opinion" tone/message to it.

When the pitching team is up by a lot, they pitch real well, at least according to wOBA. When the game is close, they don’t pitch nearly as well. Again, I think we are looking at different approaches by the batting team rather than how “well” the pitchers are pitching. In fact, you would think that pitchers would pitch better when the game is close and are more likely to “groove” pitches when they have a large lead. Certainly batters and pitchers are approaching each PA differently depending on the score differential.

(Bolding added by me)

Not just different approaches by the batting team, but also different approaches by the team in the field. It's pretty clear just from the raw performance data that there are fewer extra-base hits and more walks (even when you account for IBB), and IMO that is at least as much a function of the pitching-and-defense as it is of the hitters.

-- MWE

When the pitching team is up by a lot, they pitch real well, at least according to wOBA. When the game is close, they don’t pitch nearly as well. Again, I think we are looking at different approaches by the batting team rather than how “well” the pitchers are pitching. In fact, you would think that pitchers would pitch better when the game is close and are more likely to “groove” pitches when they have a large lead. Certainly batters and pitchers are approaching each PA differently depending on the score differential.

But can it also be a chicken and egg thing? Do they have a large lead because they are pitching well, or is the offense having a bad day? Are they facing a Dusty Baker getaway day special? Are they facing a bad team? Does the opponent basically give up* and/or replace a bunch of starters with scrubs?

* A classic example of this is game 7 of the 1985 WS. After getting down big, the Cardinal hitters basically gave up. They looked like they didn't give a #### and just wanted it to be over as quickly as possible. Saberhagen used only 24 pitches from the 7th - 9th innings. 5 batters put the first ball in play, 3 the second, and only 2 of 11 batters saw more than 3 pitches.

I don't need to see any research to show me that good starting pitchers who have pitched strongly and efficiently through the first eight innings should be kept in to try and complete the game. I suspect that almost every serious baseball fan in America knows this instinctively.

[1] You missed the introspective portion of his comment:

As well, I think I learned a lot from this research, even though in the end, I think I was vindicated, which is kind of a silly notion anyway.  As I have always said, and Guy put it aptly, my opinions are almost always informed.  Sometimes they are specifically supported by evidence and sometimes they are not.  They are never, however, out of my a**, as most “lay” opinions on sports are.  After all, I am an expert in the field of sports analysis.  You would think I was a lay journalist opining on sabermetrics, like Jayson Stark, Buster Olney, or Murray Chass, if you read the comments on BBTF.

"Sometimes they are specifically supported by evidence and sometimes they are not. They are never, however, out of my a**, as most “lay” opinions on sports are. After all, I am an expert in the field of sports analysis. You would think I was a lay journalist opining on sabermetrics, like Jayson Stark, Buster Olney, or Murray Chass, if you read the comments on BBTF. "

...He said as he pointed to his intercontinental title while the crowd jeered. He then went on to insult the local college sports team and swivel his hips.

See guys I pulled it out of my informed ass while you guys just pull it out of your dumb asses. I didn't run the numbers but my ass is so smart that I knew that eventually I would prove myself right. So why bother fighting against being a dittohead and simply save me the time of doing the research? You know I'll be right.

After all, I am an expert in the field of sports analysis.

Ah, a remix of his greatest hit:

I don’t get a “pass” because I am MGL. My projections are annually in the same league as the best on the planet. That is why I get a “pass.” And because I am considered one of the pre-eminent sabermetricians in the world. You? I didn’t catch your name?

I don't need to see any research to show me that good starting pitchers who have pitched strongly and efficiently through the first eight innings should be kept in to try and complete the game. I suspect that almost every serious baseball fan in America knows this instinctively.

I've long believed that modern bullpen management resembles nothing so much as removing effective pitchers in a continuing search for an ineffective one.

My general rule: if the pitcher is getting outs, leave him in.

Not just different approaches by the batting team, but also different approaches by the team in the field. It's pretty clear just from the raw performance data that there are fewer extra-base hits and more walks (even when you account for IBB), and IMO that is at least as much a function of the pitching-and-defense as it is of the hitters.

Mike, why does this need to be a function of pitching and defense. The hitters on a team losing by 3 or more runs know that their job is to get on base. A walk is literally almost as valuable as a HR. So shouldn't we expect them to draw more BBs at the expense of BA/SLG? I don't see any reason to attribute this to fielders -- indeed, that would make no sense at all, since fielders should be focused on getting outs, not worrying about XBHs.

See guys I pulled it out of my informed ass while you guys just pull it out of your dumb asses. I didn't run the numbers but my ass is so smart that I knew that eventually I would prove myself right. So why bother fighting against being a dittohead and simply save me the time of doing the research? You know I'll be right.

Really, McCoy? MGL does all this additional research, and it turns out he was right all along while your data didn't show what you thought it did, and this is your idea of a gracious response? Very disappointing....

My general rule: if the pitcher is getting outs, leave him in.

"And I'm sticking to this position, even though the research linked to here shows -- once again -- that my position is complete nonsense."

I don't need to see any research to show me that good starting pitchers who have pitched strongly and efficiently through the first eight innings should be kept in to try and complete the game. I suspect that almost every serious baseball fan in America knows this instinctively

Is this sarcasm or self-parody?

Really, McCoy? MGL does all this additional research, and it turns out he was right all along while your data didn't show what you thought it did, and this is your idea of a gracious response? Very disappointing....

Yeah, because MGL has been, is, and always will be so gracious.


It also looks to be an interesting case of moving the goalposts around until one finds the correct distance to make the shot.

I like MGL's work and I like the book they wrote. But that doesn't change the fact that MGL can be an arsehole to be people and thus he shouldn't be surprised if people are arseholes back at him. He shouldn't really be so sensitive to that. Secondly if he doesn't want to be treated like a Lupica or Francesca or whomever they he shouldn't have the "Worst Manager, Ever" article the way he did. The Murray Chasses of the world write articles like that and they get flamed for it. If he didn't want that he could and should have written his posts in a different way.

Mike, why does this need to be a function of pitching and defense.

The defense is usually playing what's now called "no-doubles" defense. Intentional walks contribute a large fraction of the increase in OBP - that's a pure defensive decision. And there are more unintentional walks with RISP (some of which may be semi-intentional) - a situation where the hitter would be focused on converting the runner(s) already on base along with getting on base himself.

-- MWE

It may or may not have been a mistake to leave Carpenter in the game that started this based on his pitching, but why hasn't anyone analyzed this as thoroughly from the POV of whether letting him lead off the 8th was a bad decision. Given that LaRussa played the 8th as if one run was huge subsequent to Carpenter's hit, I'd really like to see an analysis of how likely you are to get one run with Carpenter/Furcal/Jay up vs leading off with Chambers/Theriot/Craig/whoever. I have a hard time believing the pitching difference between Carpenter and 2 innings of the Cards pen, if you somehow calculate that Carpenter was a better bet, could possibly be as big as the batting difference between Carpenter (who's a terrible hitter) and whoever would PH for him.

I don't need to see any research to show me that good starting pitchers who have pitched strongly and efficiently through the first eight innings should be kept in to try and complete the game. I suspect that almost every serious baseball fan in America knows this instinctively

Is this sarcasm or self-parody?

I would like to see the research but I do share some of the same sentiment as Snapper. Part of the reason I would be reluctant to go to a reliever is because there is a chance that the reliever you go to simply doesn't have it that day. I would like to know what the odds of that happening are. Sort of like the old odds of scoring one run vs average scoring per inning discussion.

It may or may not have been a mistake to leave Carpenter in the game that started this based on his pitching, but why hasn't anyone analyzed this as thoroughly from the POV of whether letting him lead off the 8th was a bad decision. Given that LaRussa played the 8th as if one run was huge subsequent to Carpenter's hit, I'd really like to see an analysis of how likely you are to get one run with Carpenter/Furcal/Jay up vs leading off with Chambers/Theriot/Craig/whoever. I have a hard time believing the pitching difference between Carpenter and 2 innings of the Cards pen, if you somehow calculate that Carpenter was a better bet, could possibly be as big as the batting difference between Carpenter (who's a terrible hitter) and whoever would PH for him.

This issue is another issue where MGL kind of talked out of his "informed" ass. While making his bold statements about how horrible a decision it was he completely forgot about leverage. A PH in that situation faced normal leverage while pitching in the bottom of the 8th and 9th with a one run lead had much more leverage. So while it is probable that a PH in that situation would have been slightly better it wasn't really close to being such a horrble call as MGL originally claimed.

Now cue 2 week debate in which a ton of new variables get factored in which proves MGL is correct on his original assumption that wasn't based on any of those variables.

...He said as he pointed to his intercontinental title while the crowd jeered. He then went on to insult the local college sports team and swivel his hips.

I was going to comment that a moneyball/SABR gimmick would be awesome (especially if it manager hitting opponents with a copy of the New Bill James Historical Baseball Abstract that was loaded with a brick), but then I remembered that WCW already did it back in 1990 with the York Foundation.

Ok, I just glanced through it, (will read it in depth later) but does this at any point in time say that a starter pitches worse in the ninth inning relative to his seasonal/expected performance? That was the entire point of the original argument.

The initial discussion was brought about because he claimed that Motte is clearly the better choice for pitching the ninth than Carpenter. Does this article actually attack that? or is it concentrating on relative to how they normally pitch?

Carpenter is a better pitcher than Motte, Carpenter was pitching the ninth inning, MGL's claim was that a pitcher fourth time through the order gets rocked, the counter claim was that Carpenter was both on, and is the better pitcher, does this actually refute the obvious, which is to let Carpenter pitch the ninth?

edit: the last paragraph/conclusion somewhat addresses my comments, but I still find it hard for anyone who has watched Motte pitch, to actually think Motte is a better pitcher than Carpenter.

Good follow-up research by MGL. There's potentially more work to be done here, but I'm not going to fault someone for moving in the right direction.

I also endorse Emeigh in #2 here. It makes sense that both the offense and the defense - all components, not just the pitcher - take a different late-inning approach depending on game state. For the defense... In all cases you need to make outs AND prevent base-progress when outs aren't made. There's a tradeoff on making outs to prevent bigger hits, and what part of that tradeoff a team might take will depend on how big a lead they have. The offense is also varying their tradeoff based on whether they have a large deficit or a small deficit to erase.

Whether they should make that tradeoff, or to what degree, is still open to question (IMO, but I might be under-informed). But it seems obvious that they do.

Is this sarcasm or self-parody?

It was neither, though your existence may be a parody.

LaRussa certainly may have his personal flaws, but he has accomplished more than enough at this point to convince me that he knows what the hell he's doing.

MGL clearly believes that he's a genius, and wants everyone else to believe the same, but he has accomplished nothing that I'm aware of to convince me that I should believe in his towering intellect. I simply don't put much credence into his "research".

Really, McCoy? MGL does all this additional research, and it turns out he was right all along while your data didn't show what you thought it did, and this is your idea of a gracious response? Very disappointing....

Was he right all along? the point was a specific claim he made, and he has now backed off of that specific claim and instead is going with a general thesis.

Again, he claimed that Motte is a clearly better option in the ninth inning over Carpenter who was facing the lineup a fourth time through the order. He has now backed off and is saying that time through the order doesn't seem to matter if it happens in the ninth inning, and is now using generic terms of elite pitchers and elite relievers. And his definition of Motte as an elite reliever is based upon such a small sample size, that it's really silly to take that claim seriously. I know he'll say that Carpenter is no longer an elite pitcher, but he's a better pitcher than Motte and it's not even that close.

If he is going to claim Motte is an elite reliever based upon his performance this year, then he needs to base Carpenters talent level based upon his last 60 innings or so to make it an equal comparison. (Motte still wins but it's a lot closer)

I like MGL's work and I like the book they wrote. But that doesn't change the fact that MGL can be an arsehole to be people and thus he shouldn't be surprised if people are arseholes back at him. He shouldn't really be so sensitive to that. Secondly if he doesn't want to be treated like a Lupica or Francesca or whomever they he shouldn't have the "Worst Manager, Ever" article the way he did. The Murray Chasses of the world write articles like that and they get flamed for it. If he didn't want that he could and should have written his posts in a different way.

Agreed. And on top of that, considering the title of the article, and his claims that he is the most superior statistical researcher around, he should maybe do the studies on other managers and compare the difference. I mean if TLR made the move that every single manager made, and a move that Tango said every manager would have made, then how can it be the worse managing ever?

So he should realize he needs to tone down the title of the article if he is worrying about people nitpicking his saying crap out of his ass, because it's absolutely clear that he did say something out of his ass and got called on it. And the research he's produced hasn't dispelled anything, in fact if anything, it's dispelled the conclusion that pitchers pitch significantly worse the fourth time through the order in the ninth innning.

From the article's conclusion:

Putting in a reliever who is 10 to 20 points in wOBA better than the starter can save a team from .5% to 1% in win expectancy, not considering leverage. If the LI is high, say 2.5, as in a close game (in the 9th inning), the savings can be around 1-3% in WE, not a trivial amount, in my opinion.

The importance of all of this would seem to rely on what one considers a trivial amount. His research here seems to be solid, but in the end, if you're talking about a 1 to 3% change in win expectancy, it's gonna be hard to convince people that's worthwhile.

I also agree with the Bill in this thread. If you could combine this research with the decision to not pull Carpenter for a pinch hitter, I think you'd be getting closer to what could tangibly be called a bad managerial decision. As is, it seems to be very marginal (from my uneducated perspective).

MGL clearly believes that he's a genius,

yes

and wants everyone else to believe the same

while pretending that he doesn't care what other people think

but he has accomplished nothing that I'm aware of to convince me that I should believe in his towering intellect.

he has managed to get hired by an MLB club, and he has fanboys of his own...

There was a time when MGL seemingly made sporadic attempts at being civil and engaging in communication like a normal adult human being, seems he's given that up and is simply presenting himself as an arrogant know it all emotionally stunted forever teenager.

Even when he's right, or at least has something interesting to say - he;s still really not worth it because he surrounds it with such crap

Reading the article a little more in depth.

His conclusion is this, in a close game a reliever pitching the ninth pitches .003 oWOBA allowed worse than his established levels, and ALL starting pitchers pitch .007 worse than his established level. So that fourth time through the lineup, or that inning fatigue basically has no bearing on how they pitch in the ninth inning, a performance drop of around .5% for a guy who is pitching at his ability and that you, as a manager knows is pitching at his ability, vs going to a reliever who you don't know what his ability is that particular day, but who's ability you know is over the course of the year.... and he still thinks going with the uncertainty is the right way to go.


I mean what is the standard deviation of his numbers, is it higher for the relievers?(I would almost bet it is)

at least his comments he acknowledges some of the potential flaws(and one of them I think can pretty much make up the difference, the frequency relievers face same handed batters is higher)

The defense is usually playing what's now called "no-doubles" defense.

Well, that may be true when there are runners on base. But otherwise, with a lead of say 3-4 runs, the defense should be maximizing the chance of recording an out -- right?

Was he right all along?

I was addressing McCoy there, and I meant MGL was right all along in saying that Carpenter's strong performance in innings 1-7 does not mean he's any more likely to be effective in innings 8 and 9 than at any other time (of course, he's Chris Carpenter, so we would expect him to be a good pitcher in any case). McCoy had done some research which he thought showed that starting pitchers left in to pitch the 9th performed exceptionally well. This turned out to be a function only of the fact they tend to have very large leads (more often than relievers), not because their managers know something special. And I think it's interesting that McCoy found the time to post three replies, and yet still hasn't found the time to say "oops, I was wrong about that."

He has now backed off and is saying that time through the order doesn't seem to matter if it happens in the ninth inning,
I don't think that's right. I think MGL's research indicates that the times-thru-order effect does hold in the 9th, but you have to adjust for the fact that teams losing by multiple runs will change their approach in a way that reduces wOBA.

as a manager knows is pitching at his ability, vs going to a reliever who you don't know what his ability is that particular day, but who's ability you know is over the course of the year.... and he still thinks going with the uncertainty is the right way to go.

And I know it's not really his purview, but the human element has to be considered. If you leave the lights out starter in and he blows up, there's a lot less blowback on the manager than if the manager replaces him and the reliever blows up. I suppose it's mgl's and tango's mission to get away from the human element and encourage relying strictly on the numbers, but it will never happen as long as humans make not only the in game decisions, but the decisions about who gets to make the in game decisions.

Well, that may be true when there are runners on base. But otherwise, with a lead of say 3-4 runs, the defense should be maximizing the chance of recording an out -- right?

That's not generally what happens, even with bases empty. Most of the time, the outfield plays deeper and the infield corners protect the lines.

-- MWE

That's not generally what happens, even with bases empty. Most of the time, the outfield plays deeper and the infield corners protect the lines.

You're saying that a team with a 4-run lead in the 9th, no runners on, would typically use that formation? That would be some seriously sub-optimal strategy, if true. I'm not saying you're wrong, but I'd need to see some evidence before accepting that this kind of incompetence is widespread.

I was addressing McCoy there, and I meant MGL was right all along in saying that Carpenter's strong performance in innings 1-7 does not mean he's any more likely to be effective in innings 8 and 9 than at any other time (of course, he's Chris Carpenter, so we would expect him to be a good pitcher in any case). McCoy had done some research which he thought showed that starting pitchers left in to pitch the 9th performed exceptionally well. This turned out to be a function only of the fact they tend to have very large leads (more often than relievers), not because their managers know something special. And I think it's interesting that McCoy found the time to post three replies, and yet still hasn't found the time to say "oops, I was wrong about that."

I'm not sure this study refutes any of that, due to sample size issues, MGL was forced to go all pitchers, while McCoy's original point was very good/elite pitchers. Tango's study did a better job of disputing some of that claim(but not entirely) MGL makes an assumption that elite starters and elite relievers will have the same effect as all starters and all relievers.

My general rule: if the pitcher is getting outs, leave him in.

"And I'm sticking to this position, even though the research linked to here shows -- once again -- that my position is complete nonsense."

Except, it shows nothing of the kind.

MGL's number show SPs pitch as well or better than RPs in each of the score differentials, except when losing by 2 or more.

My principle applies to RPs as well as SPs. I can't stand seeing a RP get a 10-pitch 1-2-3 inning and then get yanked.

I'm not sure this study refutes any of that, due to sample size issues,

If the results among elite starters are too small to be statistically significant, then McCoy's research doesn't need to be "refuted." There is no "finding" to refute. And realistically, there is no reason at all to think elite starters don't also tend to enjoy large leads in the 9th -- given how good they are, they may well have large leads even more frequently.

Except, it shows nothing of the kind.

Then you need to read it again. What MGL's research shows is that the fact a pitcher has been getting batters out does not make him any more likely to keep getting them out. I can't imagine any evidence that would more directly contradict your "rule."

If the results among elite starters are too small to be statistically significant, then McCoy's research doesn't need to be "refuted." There is no "finding" to refute.

But then you've got to deal with the problem of how mgl posed this in the first place. He asserted that not replacing Carpenter was one of a number of decisions that proved La Russa was guilty of the "worst managing ever". As cardsfanboy has pointed out, unless La Russa is doing something unusual for a major-league manager in a must-win game, by definition it can't be 'worst'. I don't think mgl has demonstrated that La Russa did something odd, at all. He didn't even bother.

Let's remind ourselves of mgl's original point:

Surely Motte is a better pitcher than Carpenter the 4th time through the order. Again, it is not even close.

Well, his own research as presented in the linked article seems to suggest it is kind of close. A visiting reliever has a wOBA when up by one run of .342. A starter's wOBA is exactly the same.

But then you've got to deal with the problem of how mgl posed this in the first place.

No I don't. I'm interested in the analytical issue here, which is whether or not (and how much) a starting pitcher's performance over the the first 7 or 8 innings is predictive of what he'll do the rest of the way. That's a really interesting and important question. In contrast, MGL's style, or choice of a hyperbolic headline in a late-night blog post, is not -- to me -- an interesting question. YMMV, of course. But I must say I find it surprising that so many posters here are more interested in MGL's tone than they are in, say, baseball.

YMMV, of course. But I must say I find it surprising that so many posters here are more interested in MGL's tone than they are in, say, baseball.

Then why in your first post did you make sure to register your disappointment in McCoy's "ungracious" response?

Or why is MGL keeping close tabs on the number of posts here mocking him (though without, of course, lowering himself to comment at BTF)?

MGL is welcome to play I'm too smart to get tied up in social niceties card all he wants. But if he does that, he shouldn't bemoan being the victim of ad hominems. Likewise, he can deem BTF a sad husk of its once-great self and not worthy of his time. But then he ought to make a clean break and not pop over to see what the great unwashed are saying about him and commenting within the safety of the Book Blog.

Incidentally, "GuyM" is the same "Guy" referenced in MGL's 'introspective' portion of the post so memorably excerptbed by BobM in #5. So if you're wondering why he's defending MGL so doggedly...fanboy.

Needless to say, my opinion of MGL is already well-known around here. A mildly talented analyst, but an utter failure as a human being.

I must say I find it surprising that so many posters here are more interested in MGL's tone than they are in, say, baseball.

My broad point is that mgl's tone is an obstacle to understanding the baseball analysis. He (and now you) keep talking past this point and pretending it is irrelevant. It is not.

In the original thread I suggested that I thought the topic of 'worst manager ever' would be Charlie Manuel, because in the past mgl has heaped vitriol on Manuel's sabermetrically challenged approach to lineup construction. I was surprised he'd singled out La Russa because La Russa had won the game. Now, reading that original entry, one can see that La Russa did indeed make some dubious decisions, in terms of playing the percentages. And the discussion that followed largely focused on what was the most controversial point made by mgl, on the matter of replacing Carpenter with Motte. The question as mgl originally posed it gradually metamorphosed into the question he answered in the blog post to which this thread links. mgl has provided us with a very good study, but one that undermines his original point as he initially posed it. In this study he shows that, with a one-run lead, a visiting starter has exactly the same wOBA as a visiting reliever.

mgl's original analytical point, that La Russa had no statistical reason to anticipate that a starting pitcher would continue to be as good as he had been through the first eight innings, has been partially obscured by a silly and completely avoidable discussion over his intemperate tone. The easiest way to avoid such silliness is to avoid couching the whole topic in the tone he used originally.

The fact that this verbal brawl has broken out between two of my favourite baseball sites saddens me precisely because it need not have happened.

Incidentally, "GuyM" is the same "Guy" referenced in MGL's 'introspective' portion of the post so memorably excerptbed by BobM in #5.

That's some good stuff. "As has been noted by experts in the field, such as those who worship at the altar of my expertise, my expert opinions are always substantiated by expert knowledge."

Needless to say, my opinion of MGL is already well-known around here. A mildly talented analyst, but an utter failure as a human being.

Does MGL have Asperger's, you reckon? Serious question.

"I've never claimed to be an expert, either; my view is that when you write something it is either true or false, and being an expert or not being an expert really has nothing to do with it."

- Bill James

General comment: Argument via appeal to authority is annoying generally, doubly so when the authority you're appealing to is yourself.

Then why in your first post did you make sure to register your disappointment in McCoy's "ungracious" response?

I said I wasn't interested in MLG's tone (a well-worn subject), not McCoy's. But I don't really care about McCoy's "tone" either -- I care that he failed to acknowledge his error. If people won't acknowledge error, or that someone they disagree with has made a strong argument, then it's very hard to have productive discussions. MGL has admitted his initial claim about Motte vs. Carpenter was likely wrong -- it would be fair to criticize him had he refused to do so.

Incidentally, "GuyM" is the same "Guy" referenced in MGL's 'introspective' portion of the post so memorably excerptbed by BobM in #5. So if you're wondering why he's defending MGL so doggedly...fanboy.

You've really got my number, E. But this will likely come as a surprise to MGL. He and I went about 250 comments once just battling over whether UZR incorporates errors incorrectly in evaluating fielders. And I'm one of the few people who ever took him up on one of his frequent offers to bet on an analytical point (a bet that I won, but he never paid off, now that I think about it!). I would be shocked if any regular reader of The Book Blog thinks that I would agree with MGL (whom I've never met) purely out of some kind of personal loyalty. It is to laugh.....

There are numbers all over the place in this that are hard or impossible to reconcile:

Score differential Starters Relievers
Starters listed first ("not adjusted for the pitching pool")

Up 1 or less (or down) .346 (.339) .335 (.332)

But ...

road starters

Up by 1: 9% .342
Tied: 8% .377

Road relievers pitching the 9th

Up by 1: 17% .342
Tied: 19% .380


So road starters pitch as well or better ... and starters are less likely to pitch in the higher wOBA tied scenario relative to the 1-run lead.


Home team starters pitching the 9th

Score differential Occurrence wOBA against, adjusted for pitchers and batters

Up by 1: 9% .334
Tied: 8% .356
Down 1: 7% .319
Down2: 3% .349
Down 3: 2% .381
Down 4 or more: 1% .350

Home team relievers pitching the 9th

Score differential Occurrence wOBA against, adjusted for pitchers and batters

Up by 1: 9% .334
Tied: 8% .357
Down 1: 7% .347
Down2: 3% .343
Down 3: 2% .348
Down 4 or more: 1% .340


So home team starters pitch better in each scenario except down 2 or more, a situation we don't really care about. How much are MGL's overall numbers skewed by starters pitching crappy when behind 2 or more in the 9th?

Then ...

Home team pitching the 9th (starters and relievers combined)

Up by 1: 9% .337
Tied: 9% .359
Down 1: 9% .349
Down 2: 8% .346
Down 3: 7% .351
Down 4 or more: 20% .343


Home team starters pitching the 9th

Score differential Occurrence wOBA against, adjusted for pitchers and batters

Up by 4 or more: 53% .326
Up by 3: 12% .298
up by 2: 11% .316
Up by 1: 9% .334
Tied: 8% .356
Down 1: 7% .319
Down2: 3% .349
Down 3: 2% .381
Down 4 or more: 1% .350

Home team relievers pitching the 9th

Score differential Occurrence wOBA against, adjusted for pitchers and batters

Up by 4 or more: 19% .324
Up by 3: 12% .323
up by 2: 11% .322
Up by 1: 9% .334
Tied: 8% .357
Down 1: 7% .347
Down2: 3% .343
Down 3: 2% .348
Down 4 or more: 1% .340


Take a look at those numbers. Up by 1, home team pitching has a 337 wOBA against but both starters and relievers have 334 wOBAs. For tied games, overall wOBA is 359 while starters are 356 and relievers 357. Down 1 is 349 overall but 319 for starters and 347 for relievers. You see the same thing in the "up by 1" numbers for road pitchers with both starters and relievers listed at 342 yet the overall wOBA is 351.

And, as Snapper says:

MGL's number show SPs pitch as well or better than RPs in each of the score differentials, except when losing by 2 or more.

MGL also says:

Putting in a reliever who is 10 to 20 points in wOBA better than the starter can save a team from .5% to 1% in win expectancy, not considering leverage.

Except, by his own overall numbers (my first quoted numbers) the gap between the starters and relievers is only 7 points of wOBA on average (in terms of overall performance). You're not going to get a 10 to 20 point gap in wOBA unless you replace an average starter with an elite reliever which is not the scenario under debate.

And MGL appears to continue to be making the huge mistake of believing that we have any decent estimate of the true talent of a reliever which, except in the case of a handful of long-surviving relievers and maybe not even then, we simply don't.

Finally, a 380 wOBA in tied games in the bottom of the 9th? Really? This year Pujols had "just" a 385 wOBA. The average home batter in a tie game is Albert Pujols 2011 and MGL doesn't even bother to explain why this might be? Most likely that higher wOBA is strongly IBB and hits through drawn-in infield related performance.

My broad point is that mgl's tone is an obstacle to understanding the baseball analysis. He (and now you) keep talking past this point and pretending it is irrelevant. It is not.

FP: I didn't "pretend" anything -- I simply said MGL's tone was not a topic that interested me. If you and I had a conversation about MGL's style, and how it affects people's reaction to his work, I'm not sure we'd disagree very much. But what's the point? There's nothing new to say here. MGL is a package deal: the tone and substance are not available a la carte. To me, the analytic insights are worth accepting the tone. But I have no problem at all with someone who reaches the opposite conclusion. What I don't get is the decision by some here to spend their time repeating tired old criticisms of MGL's style. And what I really don't get -- what is pathetic, really -- is those who spend time searching for new opportunities to hyperventilate over the issue.

Or why is MGL keeping close tabs on the number of posts here mocking him (though without, of course, lowering himself to comment at BTF)?

Speaking of mocking (from the original "Worst Manager" thread):

I can’t stand smart-guys like Circle say things like this:

Aside from the data reseach, did anyone watching the game really feel as though either pitcher was going to give up a run after it got past the middle innings? I’m not saying that as a smart-ass, but rather a pitching fanatic that saw two guys essentially toying with very good major league hitters.

I would love, Circle, for you to document every time a starting pitcher enters the 8th inning, and based on your scouting, you will assert that this pitcher is going to have better than expected results.

All of this is just after-the-fact bullsh!tting. There’s a million places around the web for that. Please, don’t make this place a million and one.

Right, the research suggests that managers and coaches who are at the game and have watched every pitch intently and their team’s success and ultimately thir jobs hang on every decision they make and they know these guys better than anyone other than perhaps their mother and wife, can’t tell if they are “on” that particular day, but circle, you can tell?

So, Circle, are you kidding us?

The italics are tango, the bold is mgl.

Most likely that higher wOBA is strongly IBB and hits through drawn-in infield related performance.

And drawn in outfields, where easy fly balls become walk off singles.

You're not going to get a 10 to 20 point gap in wOBA unless you replace an average starter with an elite reliever which is not the scenario under debate.

I was wondering about this as well. Serious question - does Jason Motte have a 10 to 20 point lead in wOBA over Chris Carpenter?

McCoy had done some research which he thought showed that starting pitchers left in to pitch the 9th performed exceptionally well. This turned out to be a function only of the fact they tend to have very large leads (more often than relievers), not because their managers know something special. And I think it's interesting that McCoy found the time to post three replies, and yet still hasn't found the time to say "oops, I was wrong about that."

I was wrong? I'm also not sure why I would say "oops, I was wrong about that" if it turns out that the truth is something different. I wasn't the one making bold declaritive statements and insulting managers. My research was part of the process of the investigation. I looked up one variable and reported my findings. MGL has gone further (I think, I haven't read the whole thing yet) and has accounted for some more variables. Is the investigation over with? I kind of doubt it.

I said I wasn't interested in MLG's tone (a well-worn subject), not McCoy's. But I don't really care about McCoy's "tone" either -- I care that he failed to acknowledge his error. If people won't acknowledge error, or that someone they disagree with has made a strong argument, then it's very hard to have productive discussions. MGL has admitted his initial claim about Motte vs. Carpenter was likely wrong -- it would be fair to criticize him had he refused to do so.

My name popping up all the time is really weird. As others have said MGL hasn't really refuted what I said. I was disputing MGL's notion that elite starters are worse than Motte's of the world and that it wasn't even close. Numerous people, not just my one season's worth of data, posted data that shows that MGL was wrong on this particular question. Which was the only issue I addressed. I never once claimed that all SP were way better or anything like that. I simply doubted that MGL was A) using the correct data to make the right call and b)that it isn't a clear and obvious choice like MGL said.


I will add that the lead part is interesting because I wonder if that also factors in how many runs the pitcher has given up. Under normal circumstance Carpenter should have had a 4-0 or 5-0 lead when he entered into the 9th but he only had a 1-0 lead. I wonder out of all MGL's data how many games does he count that have the SP giving up 0 runs through 8. Or 1 run or 2 runs or so on. I mean it isn't really relevant to our discussion if the bulk of the data for close games is coming from 5-4 or 4-3 games.

Does MGL have Asperger's, you reckon? Serious question.

Don't a goodly percentage of the people who post here (me probably included, albeit I hope to a rather mild degree)? Serious question.

My broad point is that mgl's tone is an obstacle to understanding the baseball analysis.

It certainly is an obstacle to understanding for the masses. However, I don't think Guy is talking about that when he talks about separating MGL's tone from his content. He is talking about serious students of the game wanting to learn more. Putting up with occasional unpleasantness of MGL is the price to be paid for intellectual stimulation in Guy's mind (I could be wrong; if I am I apologize in advance).

I will add that the lead part is interesting because I wonder if that also factors in how many runs the pitcher has given up. Under normal circumstance Carpenter should have had a 4-0 or 5-0 lead when he entered into the 9th but he only had a 1-0 lead. I wonder out of all MGL's data how many games does he count that have the SP giving up 0 runs through 8. Or 1 run or 2 runs or so on. I mean it isn't really relevant to our discussion if the bulk of the data for close games is coming from 5-4 or 4-3 games.

I don't think too many starters are still pitching in the 9th after having given up 3-4 runs. Just to pick 1 example, James Shields pitched 8+ innings 12 times this year, and gave up 2 runs once and the rest 1 or 0.

Except, it shows nothing of the kind.

Then you need to read it again. What MGL's research shows is that the fact a pitcher has been getting batters out does not make him any more likely to keep getting them out. I can't imagine any evidence that would more directly contradict your "rule."

The research he's presenting show that starter are as good or better than relievers in 9th inning OBA.

I wasn't referring to any other research, but does it show that's a pitcher is less likely to keep getting outs? Because if it doesn't (and I'm sure it doesn't) that's not a reason to pull the pitcher.

I don't think too many starters are still pitching in the 9th after having given up 3-4 runs. Just to pick 1 example, James Shields pitched 8+ innings 12 times this year, and gave up 2 runs once and the rest 1 or 0.

I don't doubt you are correct on this but I think since we are dealing with such small samples that even a handful of guys that are maybe just doing enough through 8 to get to the 9th and then put up a couple of hard hits in the 9th might be playing with the numbers a bit more than they should.

For the most part we are trying to answer if whether or not letting Carpenter pitch the 9th was a good idea and to do that we need to create a bucket that most closely resemble the actual situation. I don't really think some of the data MGL has obtained to answer that question really reflects the actual situation. Granted that will happen and perhaps he has moved to trying to answer other questions but at the moment that is the question I'm looking at.

Incidentally, "GuyM" is the same "Guy" referenced in MGL's 'introspective' portion of the post so memorably excerptbed by BobM in #5.

The M stands for Mincognito.

So home team starters pitch better in each scenario except down 2 or more, a situation we don't really care about.

No, Walt -- that's true only if you ignore the talent of the pitchers. In fact, the relievers will often be better pitchers (in terms of wOBA against).

Except, by his own overall numbers (my first quoted numbers) the gap between the starters and relievers is only 7 points of wOBA on average (in terms of overall performance).

That's true only when the pitcher leads by one, is tied, or trails -- a sample that includes some pretty weak relievers, but only the best of the best starters. Oddly, you left out the data for when a pitcher is up by 2 or more: .326 (.343) .314 (.327), a 17-point advantage for relievers. Teams will often have an opportunity to replace a starter late in the game with a reliever who is "10 to 15 points" better. This is not in any way an unrealistic scenario.

Putting up with occasional unpleasantness of MGL is the price to be paid for intellectual stimulation in Guy's mind (I could be wrong; if I am I apologize in advance).
That's true for me.
Anyway, he's produced a lot of good work. I don't take his conclusions at face value - and certainly not his initial assertions (put bluntly, I don't trust his instincts or self-assuredness), but he's out there doing interesting research and I, by and large, am not. He's decidedly a net positive.

***

I haven't had time to catch up on the last few missives in this debate - outside of this thread I guess ... have MLG's refinements accounted for handedness?

For the most part we are trying to answer if whether or not letting Carpenter pitch the 9th was a good idea

No, that is the question you are still trying to answer. Lord knows why. MGL is examining whether starters who are pitching well in a game will tend to continue pitching well (relative to their general ability, in both cases). That is, is the "Snapper rule" a good one? The evidence says "no."

Serious question - does Jason Motte have a 10 to 20 point lead in wOBA over Chris Carpenter?

Probably. OPS against last two seasons:
JM: .558, .618
CC: .669, .680

Don't a goodly percentage of the people who post here (me probably included, albeit I hope to a rather mild degree)? Serious question.

Oh, I wouldn't doubt it, but MGL's unique combination of apparent intelligence juxtaposed with a startling lack of awareness of both how socially inept he is AND how that adversely impacts his ability to communicate what he's trying to communicate (and his apparent astonishment that anyone could react to him on a personal level in such a way) seems particularly symptomatic. For what my completely amateur armchair diagnosis is worth.

No, that is the question you are still trying to answer. Lord knows why.

Because I find it interesting. How dare I. I didn't realize this about moving the goalposts and declaring victory. I've never once claimed to be doing anything else yet you've tried to move my statements onto a different subject and then tsk-tsked me for not saying I was wrong on a subject I wasn't part of.

Secondly MGL keeps referring back to his original statement so I think he is still trying to answer it as well despite what you claim.

I will add that the lead part is interesting because I wonder if that also factors in how many runs the pitcher has given up. Under normal circumstance Carpenter should have had a 4-0 or 5-0 lead when he entered into the 9th but he only had a 1-0 lead. I wonder out of all MGL's data how many games does he count that have the SP giving up 0 runs through 8. Or 1 run or 2 runs or so on. I mean it isn't really relevant to our discussion if the bulk of the data for close games is coming from 5-4 or 4-3 games.

Bingo. You'd certainly hope that some context was applied to these numbers, but there's no way to tell from what's being presented (one of the lingering deficiencies that Da Book folks just don't seem motivated to address). The "lumpage" of run differential categories is a bit suspicious.

Here's what I can tell you. When we look at all the games from 1996-2011 where pitchers threw at least 8.1 IP and allowed at least one run, their records in those games are as follows: 1634-233 (.875). There are also 202 no-decisions for those starters, which split more than a bit better than 50/50 toward the pitcher's team (110-92, .544).

In the no-decisions--the games where the outcome is muddied and almost always goes into extra innings, which represent about 10% of the games--pitchers who go the full nine have an ERA of 1.55, whereas the pitchers who have to be taken out in the middle of the inning (presumably because they are giving up hits/walks if not also runs) wind up with an ERA a full run higher (2.58). Note, though, that we still don't know in what actual inning that extra run was surrendered.

So teams flat out win 88% of these games, and in the 10% of the games where the trouble that the starter gets into during the ninth results in a starter no-decision, they win only 54%, which winds up with an overall WPCT for the full set of games of .842 (1744-325). That's a 4% overall drop in WPCT. Is that enough to make us declare an inviolate rule that all starters get relieved after eight innings?

Note that this may include a few games where the pitcher has given up no runs through eight innings...you'd have to go check all those games to get that info.

What it doesn't include are the games where pitchers were taken out for relievers after pitching eight scoreless innings. When we look at the results of those games-where the reliever come in "clean" in the ninth--we find that the result clearly depends on the score of the game. We also discover that the relievers don't create a "lock" for their teams. Overall in these games, the teams who relieve their shutout-pitching starters with the score 0-0 compiled a record of 77-87 over that time span. (BTW, the average number of games a year from 1996-2011 that fall into this category is just over ten a year.)

Now 50 of these games are ones where the team failed to score a single run. When we pull those out, that makes things look a lot better--77-37 (.675). So the relievers, in what amounts to less than 5% of the total sample, do better in clean innings than they do when they have to come in with men on base (110-92, .544).

But the vast, vast majority of these games (over 90% when we factor in the actual 9 inning shutouts thrown) are victories. The winning percentage for these games is 892-87 (.911). Interestingly, relievers allowed at least one run in 158 of these games, a little more than sixteen percent of the time. Relievers were responsible for blowing leads in 34 of these games; the other 53 were lost by relievers because the team couldn't score at all.

The difference between a) 8 IP and 0 R with "clean ninth" reliever and b) 8+ IP with at least one run (followed by reliever) is the difference between .911 and .875. It's right in that same 4% region. Sample sizes: 979 games (61 times a year, about twice a year per team), and 2110 games (131 games a year, a bit more than four times a year per team). That "decay" looks to be about a tenth of a win a year per team.

I would suggest that someone go look at the 145 games in the past sixteen years where the pitchers threw nine innings and lost the game, to see how many of those runs were surrendered in the ninth inning. What I can tell you from the first small sample (2009-2011, a total of eleven games), runs were scored in the ninth only twice. And in only one of those cases did those runs directly cause the loss. If 90% of these games are lost already, then we can certainly question why the manager is leaving the pitcher in that long (saving the bullpen by using a starter with a low pitch count? just generally stretching Roy Halladay? etc.) but we really can't assign this to the specific purported strategic deficiency that MGL is apparently fixated on.

The figure for the starters to beat is 16%. Do they give up at least one run in one of every six ninth innings that they are allowed to appear in? Someone (else) can go into the bb-ref game logs and count that for us...

As wonderful as wOBA may be, it's really not a substitute for looking at how many actual runs are scored in the inning(s) in question.

Does MGL have Asperger's, you reckon? Serious question.

I don't think the aggressive obnoxiousness evidenced by MGL is characteristic of Asperger's, however, his general social clumsiness/cluelessness coupled with his obsession over analyzing numbers might be- but as you note the same can be said of a great many Primates here (as well as dovetailing with the basement dwelling geek stereotype).

MGL seems to be a highlevel stahead who can also be a bit of an arsehole (as someone else put it up above)- on paper sop to speak, what he is like in real life I have no idea.

MGL is examining whether starters who are pitching well in a game will tend to continue pitching well (relative to their general ability, in both cases). That is, is the "Snapper rule" a good one? The evidence says "no."

I don't understand how what MGL presents here addresses this question nor how it answers the question with a "no". As I read the results, it acknowledges that starting pitchers' 9th-inning performance is better than you'd expect purely from the "4th time thru the batting order" effect (which, if anything, would support snapper's point, not refute it), and that the difference in performance between starters and relievers in the 9th inning is small and of indeterminate sign.

Does MGL have Asperger's, you reckon? Serious question.

I find this to be incredibly unlikely for two reasons: 1) tossing out Asperger's is something done a little too readily, and 2) according to people who've actually met MGL in real life, his social skills come off fine.

Rabies more likely.

"He who truly knows has no occasion to shout."

- One of the original sabermetricians

I don't understand how what MGL presents here addresses this question nor how it answers the question with a "no". As I read the results, it acknowledges that starting pitchers' 9th-inning performance is better than you'd expect purely from the "4th time thru the batting order" effect (which, if anything, would support snapper's point, not refute it), and that the difference in performance between starters and relievers in the 9th inning is small and of indeterminate sign.

MGL is saying that these starters, despite having pitched incredibly well (on average) over the first 8 innings, basically only match their season average in the 9th. That does appear to be better than the times-thru-order effect would predict, except that other research has shown that lower temperatures in the late innings produce a drop in offense that roughly matches the 4th-time-thru-order effect (as MGL references in the post). But whatever the various offsetting factors may be that produce this net result, I don't think that many fans -- or perhaps many managers -- think that a pitcher will just put up his average stats after 8 great innings. But that's what the data suggests.

And no, the starter/reliever difference is not small. If both starters and relievers will basically put up their own seasonal performance in the 9th, that will almost always argue in favor of a reliever. Few teams have bad enough relievers, or great enough starters, that they won't have a better option in the pen.

Maybe I am misreading it as I am skimming it but MGL's data says that road starters in the 9th have a .342 wOBA against while road relievers in the 9th have a .342 wOBA against as well. Both relievers and starters at home have a .334 wOBA against.

I find this to be incredibly unlikely for two reasons: 1) tossing out Asperger's is something done a little too readily, and 2) according to people who've actually met MGL in real life, his social skills come off fine.

In which case, I feel less compunction than I otherwise might in trashing his social skills.

MGL is saying that these starters, despite having pitched incredibly well (on average) over the first 8 innings, basically only match their season average in the 9th. That does appear to be better than the times-thru-order effect would predict, except that other research has shown that lower temperatures in the late innings produce a drop in offense that roughly matches the 4th-time-thru-order effect (as MGL references in the post). But whatever the various offsetting factors may be that produce this net result, I don't think that many fans -- or perhaps many managers -- think that a pitcher will just put up his average stats after 8 great innings. But that's what the data suggests.

And no, the starter/reliever difference is not small. If both starters and relievers will basically put up their own seasonal performance in the 9th, that will almost always argue in favor of a reliever. Few teams have bad enough relievers, or great enough starters, that they won't have a better option in the pen.

I'm not seeing it in these data:

Road team starters pitching the 9th

Score differential Occurrence wOBA against, adjusted for pitchers and batters

Up by 4 or more: 60% .336
Up by 3: 12% .326
up by 2: 11% .321
Up by 1: 9% .342
Tied: 8% .377

Road team relievers pitching the 9th

Score differential Occurrence wOBA against, adjusted for pitchers and batters

Up by 4 or more: 36% .339
Up by 3: 13% .340
up by 2: 15% .339
Up by 1: 17% .342
Tied: 19% .380

Home team starters pitching the 9th

Score differential Occurrence wOBA against, adjusted for pitchers and batters

Up by 4 or more: 53% .326
Up by 3: 12% .298
up by 2: 11% .316
Up by 1: 9% .334
Tied: 8% .356
Down 1: 7% .319
Down2: 3% .349
Down 3: 2% .381
Down 4 or more: 1% .350

Home team relievers pitching the 9th

Score differential Occurrence wOBA against, adjusted for pitchers and batters

Up by 4 or more: 19% .324
Up by 3: 12% .323
up by 2: 11% .322
Up by 1: 9% .334
Tied: 8% .357
Down 1: 7% .347
Down2: 3% .343
Down 3: 2% .348
Down 4 or more: 1% .340


Score differential Starters Relievers

Up 2 or more .326 (.343) .314 (.327)
Up 1 or less (or down) .346 (.339) .335 (.332)

The numbers in parentheses are the pitchers’ overall seasonal wOBA against, similar to their true talent.

Starting at the bottom, he shows starters who pitch the 9th with a lead of two or more have a wOBA of .326 vs. a true-talent wOBA of .343. These starters have a higher wOBA than "true-talent" when pitching with a lead of 1 or less, but he elsewhere noted that overall wOBA is much higher than expected the closer the run differential, so you'd have to control for that.

For home teams, I see a lower wOBA for starters than for relievers when pitching with a lead of 2-3 runs or when trailing by 1-2 runs with the results basically a tie for leads of 4+, leads of 1, and ties. Relievers out-pitch starters when the team is trailing by two or more runs, but those are almost certainly cases where the pitcher has NOT pitched excellent thru the first 8 given that his team's trailing.

For road teams, I see lower wOBAs for starters than relievers at every score differential.

What am I missing?

The bandwagon.

To me, the analytic insights are worth accepting the tone. But I have no problem at all with someone who reaches the opposite conclusion.

For the record (not that anyone cares, but w/e), this is where I am. I really like the Book Blog (have it bookmarked actually, have read it every day for a few years now, and a copy of "The Book" is currently staring at me from across the room, right next to Jonathan Wilson's "Inverting the Pyramid"), but I ordinarily only read Tango's posts. The only time I read MGL is when when I'm fairly sure he's going to just be analytical/when he deals directly with an issue I find very interesting. I get that I miss information occasionally, but I just find it difficult to wade through abrasive and dismissive tone along with interesting analytical insight. I have the same problem with many commentators here, though, which is why I post on very few threads.

Just my two cents :)

Probably. OPS against last two seasons:
JM: .558, .618
CC: .669, .680

A couple of questions - first, why would you only use the past two seasons? Motte pitched 56.2 innings in 2009 where he had an .804 OPS against. Also, in 2009 Carpenter threw 192.2 innings and had a .581 OPS against. While I certainly understand giving more weight to recent seasons, cutting it off at two really harms Carpenter and aids Motte. Second, Motte has thrown 188 innings in his career. Carpenter has thrown 192.2, 235, and 237.1 innings in the past 3 years. Over the last 3 years, Carpenter has an OPS against of .647 in 665 innings, while Motte has an OPS against of .654 in 177 innings pitched. How confident are we that Motte is better than Carpenter?

What am I missing?

Kiko: What the data show in aggregate is that starters do slightly better in the 9th than you'd expect, given their talent and 4th-time-thru effect. However, relievers also perform slightly better than expected in the 9th. So it's almost certainly not the case that the 9th inning performance is a function of the starters' strong performance in innings 1-8. MGL has a good comment up now that combines the data into more robust cells.

Also, just look at your last two lines: the relievers are significantly outperforming the starters -- by 11-12 points of wOBA -- regardless of score. That's the bottom line: relievers outpitch the starters, even though these are mainly good to great starters pitching at the top of their game.

MGL has a good comment up now that combines the data into more robust cells.

Yes 1 up is now combined with tied and losing. I'm not really sure why those should all be combined.

I've followed this ongoing debate from the start and it's been a lot of fun, eye-opening challenge my assumptions stuff. I can't say I understand the significance of all the data and fully trust the conclusions, but that has more to do with my deficiencies. One question I have, but I'm afraid to ask lest I look foolish, is how much do things like quick hooks skew the data? (For all I know this has been discussed and I've missed it, or it's completely insignificant). For instance, a starter is more likely to be lifted from the game after giving up a hit in the 9th inning. Reliever who give up hits, especially closers, are allowed to finish the inning. So you have maybe not an insignificant number of starters who start out the inning, give up a hit to the lead-off batter and are taken out, or give up 1 or more hits with 1 out and are taken out. On those cases, doesn't the wOBA of the hitters they've faced skew up since no outs were recorded. Additionally, do things like lefty-lefty match-ups skew the numbers in favor of relievers? Sorry if this is super stupid. Just trying to follow.

Also, just look at your last two lines: the relievers are significantly outperforming the starters -- by 11-12 points of wOBA -- regardless of score.

But when you split it out by score in the previous tables, the starters basically always out-perform relievers. So which is it? It seems to me that the swings in average wOBA by score differential and the differences in the mix of starter v. reliever performance across score differentials (starters rarely pitch the 9th inning of games their team is losing) mean that (a) you have to look at the results by score differential, and (b) wOBA's probably not the right metric to use here.

What the data show in aggregate is that starters do slightly better in the 9th than you'd expect, given their talent and 4th-time-thru effect. However, relievers also perform slightly better than expected in the 9th.

If everybody out-performs their expectations, then that tells me that your expectations are wrong. Which goes back to my point (b): I don't think wOBA's the right tool for the job here.

Sorry if this is super stupid. Just trying to follow.

Om the contrary, I think wondering about the effect of these things, particularly the quick hook, is a worthwhile line of inquiry.

But when you split it out by score in the previous tables, the starters basically always out-perform relievers.

The differences are small in most cases, and in the aggregate. And remember that the relievers are being compared to their seasonal rates, which are heavily influenced by their 9th inning stats (while starters' are not). So the reduced offense in the 8th/9th innings will tend to look a little bigger for starters, just because of the methodology.

I don't think wOBA's the right tool for the job here.

There's probably something better than wOBA, such as WE, given the peculiarities of the 9th. But who cares, in terms of the starter/reliever decision? A different metric won't change the story at all, which is that both relievers and good starters (at least those allowed to pitch the 9th) will do about their seasonal average in the 9th. A truly elite starter may be as good a choice as your closer (if you don't have to let him hit). In all other cases, the reliever would be expected to pitch better.

But when you split it out by score in the previous tables, the starters basically always out-perform relievers. So which is it?

Those results are adjusted for pitcher talent. The 12-point spread is not -- it reflects the reality that relievers tend to perform better.

There's probably something better than wOBA, such as WE, given the peculiarities of the 9th. But who cares, in terms of the starter/reliever decision?

I would think anybody who actually cares about finding the right answer.

good starters (at least those allowed to pitch the 9th) will do about their seasonal average in the 9th.

But I don't see where the data says this. A wOBA of .326 strikes me as meaningfully better than .343. Or at least, I haven't seen anything from MGL that shows that those differences aren't statistically significant.

No, that is the question you are still trying to answer. Lord knows why. MGL is examining whether starters who are pitching well in a game will tend to continue pitching well (relative to their general ability, in both cases). That is, is the "Snapper rule" a good one? The evidence says "no.

If that is the case, then it's a case of moving goal posts, and there is no reason for MGL to be ridiculing BBTF for not falling down on their knees in worship of him. The fact that he has changed his original assertion DOES NOT vindicate him. His original assertion, the one that got people worked up, was a silly notion that Motte is clearly a better choice to pitch the ninth than Carpenter who was cruising at that point.



As I mentioned his study doesn't remotely talk about the suredness of a performance, what I mean what is the liklihood of a pitcher who is pitching normal to continue pitching normal, versus bringing in a reliever who you haven't seen throw a ball that day? Heck this own study helps to some extent solidify the decision to keep Carpenter in, it says basically that a pitcher is very likely to pitch close to his norm even going into his ninth inning of pitching, regardless of how well he is pitching. To me, sticking with the pitcher who you know is pitching solidly/well and has a good chance(because the numbers say so) of pitching to his talent level in the ninth inning is by far the better choice than dealing with the uncertainty of a reliever who you haven't seen throw against major league pitching that day.

Those results are adjusted for pitcher talent. The 12-point spread is not -- it reflects the reality that relievers tend to perform better.

and the managers know the quality of the relievers. Again the decision wasn't whether to pitch Westbrook in the ninth vs Motte, but Carpenter vs Motte. The manager already knows the quality of his pitchers(you would assume---but of course, according to MGL, only his expertness is allowed to assume because he knows more than everyone else including a hof manager, who according to at least one well researched book, is probably the second greatest manager of all time)

I mean we again are talking about Motte's numbers, who according to MGL represent a well defined talent level, but according to TLR/Duncan it's due to putting him in situations where he'll succeed. (another point not brought up, but TLR is one of the better managers of matching up players, and his pitchers numbers possibly don't accurately reflect their true talent level, but instead reflects their talent level against the matchups that TLR/Duncan think they can succeed at)

don't accurately reflect their true talent level, but instead reflects their talent level against the matchups that TLR/Duncan think they can succeed at)

Interesting theory, sounds like a study that is right up Tango and MGL's alley.

For the record, after all my bashing of MGL, I'm looking forward to his BP article that he is working on on pitcher usage, pinch hitter usage, and WE, that he references in the comments of the linked article. I think he brings a ton to the table as an analyst and is willing to tackle some really minor issues with a fine tooth comb(and that is a compliment btw) to try and find out as much as he can. The problem is that I think he goes into these studies just like the Murray Chass's of the world, with an opinion formed(his opinion is that everything can be broken down by numbers, and a right answer can be close to 100% best answer everytime) and that even if he is shown his numbers are wrong, it's that the numbers are the only thing that matters.

I think the human element is of course a major part of the game, and I always thought it was silly, even when Bill James said it, to think that a careful observer couldn't tell a .250 hitter from a .300 hitter if they didn't track the numbers. People can see when a person is doing good or not, scouts/coaches can tell when a person has their head screwed on straight, managers can tell whether that 8 inning 3 hit performance was masterful or lucky. Things might equal out in the end, or on the macro level, but on the micro level, the numbers cannot possibly answer to the point that you can answer with certainty that pulling Carpenter for Motte is a +3% WE. It's utterly silly to think something like that is even possible unless your name is jehovah(assuming you believe in him/her/it)

It hasn't been subtle that TLR/DD have been trying to shield Motte from LHB. Or maybe had been trying, since he's now the de facto closer, perhaps because he's incorporated a cutter. Anyway, over the last three seasons Carpenter's faced 54/46 RHB/LHB. Motte is at 63/37.

It hasn't been subtle that TLR/DD have been trying to shield Motte from LHB. Or maybe had been trying, since he's now the de facto closer, perhaps because he's incorporated a cutter. Anyway, over the last three seasons Carpenter's faced 54/46 RHB/LHB. Motte is at 63/37.

Good point. Ultey and Howard were up in the 9th and even in 2011 Motte was not good against lefties. Another huge variable that hasn't really been addressed by MGL's data.

#39 That's not exactly what James wrote:

"[...] as I've tried to explain to so many people, that when you write facts they are either true or false, and being an expert really has absolutely nothing to do with it."

Bill James, Historical Baseball Abstract

(the insertion of "facts" is important to James' point I think. Expert opinion has some value)

Interestingly this comment was written in response to his wife's note that she wasn't an expert on the topic at hand (uniform IIRC)

#86 The other thing that I found puzzling about this analysis is that there doesn't seem to be any adjustment for the specific relief pitchers. Sometimes the "closer" is Jose Mesa in one of his bad years. Or more recently Joe Borowski.

Shouldn't one of the steps be to normalize for the expected results?

Also, the pinch-hitting opportunity was in the 8th. I'm honestly not sure what Allen Craig's true talent level is (though I doubt it's a 153 OPS+ -- his 2011 numbers -- or even a 131 -- his career numbers) and this makes it tricky to estimate what the pinch-hitting opportunity was worth and balance this with trying to mix and match for 6 outs.

I think though that any sim will show you gaining by pinch-hitting in the 8th. (Particularly since the broadcaster have not tired of shouting the virtues of the Cardinal bullpen this year) I have little doubt that Craig is a pretty good hitter and that the bullpen had a pretty good chance of holding the lead.

Remember that no one here is attempting to correlate one-inning wOBA with actual run scoring in actual games. And even if the calculations here are correct and not subject to skewing, the results here indicate that the number of times such a decision actually matters in a ballgame is extremely limited. One of those times, of course, has to do with the Carpenter game that produced the hyperbolic statements in the first place.

The sample size over the past sixteen years for games where it would even be possible for the Carpenter scenario is exceptionally low--from looking through the data, it's a total of 189 games. That's a dozen a year for both leagues. So whatever else you want to say about it, these situations don't affect the outcome of very many games.

Of these games, 87 are ones where the starter lost a one-run game (either 2-1, 3-2, or 4-3).

In those games, starters gave up go-ahead runs in the ninth a total of twelve times, or 13.8% of the time. On the other side of the ledger, there were ten starters who pitched ninth innings with a one-run lead and won the game (2-1, 3-2, and 1-0 games, including five shutouts). Those are just a small subset of 1-0 shutouts over the time frame: the total number of 1-0 complete game shutouts over the time is 102.

Relievers gave up runs in the ninth inning of games games that were still won eight times out of seventy-seven actual game opportunities, or 10.4% of the time. Note that this doesn't count games where there are blown saves/losses (hard to parse from the bb-ref data), or the 19 games where the starter threw nine shutout innings only to have the reliever lose the game in extra innings.

There were only two games where a starter took a 1-0 lead into the ninth and gave up two runs to lose the game on his own.

The other question that isn't addressed is the probabilities for all of this in the eighth inning, or the seventh. The starters who lost complete games also gave up go-ahead runs in these innings--seven times in the seventh, and twice in the eighth. Clearly teams are not using closers in the seventh, and only rarely in the eighth--so is bullpen usage indicated there as well? If one attempts to calculate that, do you use the full data set for what was done in those innings or only slice out the instances where the bullpen pitches with these exact scoring differentials? Methodologies can get murky here rather quickly...

The advantage here--if it is actually an advantage when we look at actual runs scored and not the misty, abstract aggregation of performance that is wOBA--is at best very slight, and hardly worth all the effort to defend as some sort of iron-clad precept as it was first represented. There were 226 1-0 and 2-0 complete game shutouts thrown from 1996-2011...that's 226 wins as opposed to a total of twelve losses where the go-ahead run was scored in the ninth inning. As noted before, we are in a context where winning is highly likely either way.

Good point. Utley and Howard were up in the 9th and even in 2011 Motte was not good against lefties. Another huge variable that hasn't really been addressed by MGL's data.

Not much data for TLR to use in assessing Motte vs. Carpenter for those guys (fewer than five PAs vs. Motte for each)...but Carpenter has had success against Howard, not so much against Utley. Motte's OPS vs. lefties in 2011: .738; Carpenter: .665.

But yes, this is a significant variable for which there has been no adjustment.

Also, the pinch-hitting opportunity was in the 8th. I'm honestly not sure what Allen Craig's true talent level is (though I doubt it's a 153 OPS+ -- his 2011 numbers -- or even a 131 -- his career numbers) and this makes it tricky to estimate what the pinch-hitting opportunity was worth and balance this with trying to mix and match for 6 outs.

Makes for an entirely different calculation, Ron. The data presented by MGL has to be recalculated to factor this in. Looks like you've opened another can of worms entirely. For my data set (2-1 losses), you'd add two more in, bringing us back to the 16% total that we'd discussed earlier. When all is said and done, it's looking more and more like a wash.

And don't forget to try to take into account the extremely unusual fact that Carpenter actually got four outs in the eighth inning--he had a K that reached base on a catcher error!

Yes 1 up is now combined with tied and losing. I'm not really sure why those should all be combined.

This has been bothering me as well, but mostly for the reason that I remained focused on the particular situation that started this discussion: visiting team up by one, elite starter pitching a shut-out. Surely up by one needs to be linked to the other situations where the visitors are winning.

What he says is that

starters tend to pitch when their team is ahead by 2 or more runs and relievers tend to pitch when the game is close or their team is losing

If you look at his data, however, there is no evidence that starters tend to pitch in the 9th, which makes me think he's referring more generally to any inning. On the contrary, starters are rarely used to pitch the 9th. His data on usage patterns show that in only two cases does a starter pitch on more than 10 per cent of the occasions, and even then not by much.

It's fair to say that MGL has accepted that his statement about it being clear that Motte was the better option was wrong. In this case, though, it looks like he's created an artificial distinction that simply aids his argument. What he's basically done is to group the results by outcome, on the principle that in a close game, the hitters have a significantly better wOBA.

I'm somewhat 'statistically challenged', so my ignorance may be at work here, but I follow MGL's argument thusly:
1) When using all games, the starters appear to do better by wOBA than relievers.
2) However, in close games both starters and relievers do worse.
3) If we make two separate categories based on wOBA outcomes, we see the gap between starters' success and relievers narrows. Q.E.D.!

Shouldn't we in fact be seeking to make our distinction based on usage patterns? So isn't the obvious distinction, when controlling for score, to be made at the 'up by 4' versus the 'up by 3 or worse' (when we get 91%+ reliever usage) categories?

However, in close games both starters and relievers do worse.

This makes me think the study is flawed.

Close games tend to see better RPs pitch. wOBA shouldn't be higher.

This needs to be looked at to see if factors like IBB, semi-intentional BBs, infield in, outfield in (with fielders not bothering to even catch routine flyballs that are walk-offs) etc. are skewing the #'s.

This makes me think the study is flawed.

Close games tend to see better RPs pitch.

I'm guessing, but it looks to me as if the figures for relievers are dragged down by tied games, when the manager might not bring the closer in, because it is not yet a save situation. So if the closer is the 'best' reliever, we see their effect more when the team has a lead by one.

#87: I was quoting the New Historical.

good starters (at least those allowed to pitch the 9th) will do about their seasonal average in the 9th.
But I don't see where the data says this. A wOBA of .326 strikes me as meaningfully better than .343.

Of course .326 is better than .343. That's the 9th inning reduction in offense by teams losing by 2+ runs that MGL is reporting, and is perhaps the most interesting finding here. But the point is that this can't be a function of the starters' good performance in innings 1-8, since relievers show the same effect: .314 (.327). If both starters and relievers enjoy this 9th inning bonus, then you're still back to the same place: you expect the pitcher with the better seasonal rate to do better, and by about the same amount.

Yes 1 up is now combined with tied and losing. I'm not really sure why those should all be combined.....
In this case, though, it looks like he's created an artificial distinction that simply aids his argument. What he's basically done is to group the results by outcome, on the principle that in a close game, the hitters have a significantly better wOBA.

FP/McCoy: There's nothing "artificial" about this. MGL needs to create larger categories to deal with sample size limitations. Below is his data. Choosing to break the data at leads of 2+ vs. 1-or-less was entirely reasonable. Plus, he gives you enough data to figure out the result for any different categorization you prefer! This carping is pathetic....

ROAD PITCHERS
Up by 4 or more: 43% .337
Up by 3: 14% .338
up by 2: 17% .336
Up by 1: 19% .351
Tied: 20% .380

HOME PITCHERS:
Up by 4 or more: 21% .327
Up by 3: 7% .323
up by 2: 9% .324
Up by 1: 9% .337
Tied: 9% .359

If that is the case, then it's a case of moving goal posts,

CFB and McCoy, if you want to continue debating the Carpenter/Motte decision, knock yourselves out (though I'm not sure who you will be debating, exactly). But this "moving goalposts" thing makes no sense at all. MGL wrote a post of over 2,200 words here. The name "Motte" appears zero times. Same for "LaRussa" and "Carpenter." This post isn't about your pet issue.

And even if you want to discuss LaRussa's handling of pitchers, it seems a little weird to focus on keeping Carpenter in when LaRussa has spent the past week changing the future of postseason baseball by winning with 2-5 inning starters. My guess is that except for the truly elite starters, this is increasingly what post-season and very late-season baseball is going to look like. That's way more interesting, IMO, than the coin toss Motte/Carpenter call....

It's just MGL being MGL. And I mean that in both the positive and negative senses.

This carping is pathetic....

Do you want a conversation or do you simply want to come off as a fanboy? You spent numerous posts trying to call me out on something I wasn't even part of and then when you were shown you were wrong on it do you say "oops, I was wrong"? Nope. For someone who keeps saying you want to talk about the numbers and baseball you sure do spend a lot of time trying to divert the conversation away from the number and baseball.

My issue with the groupings was that it isn't relevant to the question that was being discussed. It was goalpost moving in the clearest and most obvious sense. MGL wasn't starting a new topic but continuing an old conversation and even used the new data to try and show that his first statements about this were not incorrect. So it is quite kosher to then question why the other scoring situations in the grouping are relevant. Tie ballgames produce Albert Pujols like production from hitters and up 1 environments produce below average (I believe, I'm recalling from memory-it might have been average) hitters. How are the two related? And yes I can create my own groupings and did-Up 1 run. So I don't see the need for you to write what you did other than to shoot out a broadside on something not relevant.


MGL wrote a post of over 2,200 words here. The name "Motte" appears zero times. Same for "LaRussa" and "Carpenter." This post isn't about your pet issue.

Like I said above if this is a different topic then he and YOU shouldn't try to use it to say you guys were right all along which he and YOU did.


My guess is that except for the truly elite starters, this is increasingly what post-season and very late-season baseball is going to look like. That's way more interesting, IMO, than the coin toss Motte/Carpenter call....



There is a thread on BTF about LaRussa's reliever usage. Despite what MGL thinks about us we're capable of talking about more than one thing at a time.

And even if you want to discuss LaRussa's handling of pitchers, it seems a little weird to focus on keeping Carpenter in when LaRussa has spent the past week changing the future of postseason baseball by winning with 2-5 inning starters. My guess is that except for the truly elite starters, this is increasingly what post-season and very late-season baseball is going to look like. That's way more interesting, IMO, than the coin toss Motte/Carpenter call....

Wash gets an assist, but yes, that is the really interesting development in these playoffs. I wonder if it's a one-year aberration or a playoff likelihood in the future. I also wonder if it seeps at all into the regular season. Probably not, but worth watching.

I think the difference between this year and other years is that Tony pulled it off this year. It generally doesn't work. It wasn't like Tony was pulling startes in the 4th or 5th while they were throwing shutouts. He pulled the starters because they were getting rocked which generally leads to losses since even if relievers are stingy the offense is still in a hole. This year that didn't really happen because the Brewers couldn't keep the Cardinals from scoring and yes Tony's pen could stop the Brewers.

I don't think this will change anything. How do you plan for something like this? I'll let my starters give 4 or 5 runs in 5 innings or so and then let my pen pitch 4 or 5 innings each game and allow virtually no runs to score?

I wonder if it's a one-year aberration or a playoff likelihood in the future.

I think it's the latter. And then we will see starters on very short leashes in Sept. on contending teams. And eventually, some small market team will realize they shouldn't spend money on a 5th starter, and when they have 5 consecutive games on the schedule they will just do a bullpen game in which a long reliever pitches the first 3 innings.

It's easy to forget how big the reliever advantage is. Here are the top 50 pitchers in OPS against the past two seasons (100 IP min.). Only 10 are starters. Basically, closers are more effective than all but the most elite starters, and even a good setup man is more effective than the vast majority of starters.

Player OPS
Mike Adams 0.511
Mariano Rivera 0.514
Joaquin Benoit 0.52
Jonny Venters 0.53
Sergio Romo 0.54
Daniel Bard 0.543
Neftali Feliz 0.555
Joel Peralta 0.558
Koji Uehara 0.559
Scott Downs 0.563
Rafael Soriano 0.564
Sean Marshall 0.567
Santiago Casilla 0.568
John Axford 0.57
Carlos Marmol 0.57
J.J. Putz 0.57
Jose Valverde 0.582
Josh Johnson 0.583
Clayton Kershaw 0.583
Jason Motte 0.585
Joe Smith 0.586
Heath Bell 0.587
Ryan Madson 0.588
Justin Verlander 0.591
Rafael Betancourt 0.593
Luke Gregerson 0.593
Tyler Clippard 0.594
Joel Hanrahan 0.597
Evan Meek 0.597
Eric O'Flaherty 0.6
Matt Thornton 0.6
Adam Wainwright 0.604
Jered Weaver 0.61
Grant Balfour 0.612
Cliff Lee 0.612
Jonathan Papelbon 0.612
David Robertson 0.613
Ramon Ramirez 0.614
Roy Halladay 0.615
Chris Perez 0.615
Alfredo Aceves 0.616
Brandon League 0.618
Matt Cain 0.621
Javier Lopez 0.621
Michael Pineda 0.621
Kyle Farnsworth 0.622
Felix Hernandez 0.622
Fernando Salas 0.622
Drew Storen 0.623
Darren Oliver 0.624

And if they had to pitch more they wouldn't be as effective.