Bruce Miles on Roles in the Bullpen

[vance_the_cubs_fan]

I stumbled on this yesterday, and it's a few days old, but I thought it would be a good discussion.

 

Link

 

I think I agree with Bruce here. It's not that I don't have confidence in Wood as a closer, but the way that everyone in the bullpen has to have a pre-ordained assigned role irks me as well.

 

Sometimes it might be prudent when Marmol is throwing well to let him continue on and get the final outs. The same might be said of any pitcher in the pen.

 

And maybe sometimes it would be best to bring Wood into the eighth inning of a game.

[biittner77]

Part of the reason for the defined roles is that a manager has to be mindful of what he'll be able to do the next 2 games. If Marmol pitches 2 innings and saves the game that's great today but will he still be available tomorrow? By pitching him 1 inning in each game he could potentially affect the outcome in 2 games instead of 1.

 

Where I object to the defined roles is, for example, not bringing Marmol in during the 6th inning with the bases loaded because he's the 8th inning guy.

 

Eventually, some team is going to have success with the "closer by committee" model. Old timey baseball types cringe at this prospect but that has more to do with the committee than the concept. Closer by committee worked just fine for the 1990 Reds mostly because they had 3 guys that had great stuff.

[CuseCubFan69]

I think a lot of this would depend on how Wood would feel about it. He seems like a team guy to me but he may be mentally set on being the closer and this might be a tough move for Lou to switch him to another role.

[wvcbxl]

I agree with the reasoning Biittner77 alluded to above: as far as Marmol is concerned, the Cubs should use him as often as possible/needed in his fireman role, i.e. to pitch out of a jam with men on base (in any inning between 4th and 7th), or in the 8th inning of a tight game. I'd leave Wood in the typical closer role he's been growing in to for the past two months.

[Jehrico]

Considering Wood's injury history, he needs to have a specific role. I like the idea that he isn't pitching more than one inning, isn't getting up and down during games like a middle reliever might, or anything like that. Keep the wear on his arm down to a minimum during his first full year as a reliever.

 

But roles for everyone else? I agree, I think they're overblown (other than Marmol's role needs to be to come in and get out of jams during tight games, regardless of whether it's the 6th, 7th, or 8th).

[vance_the_cubs_fan]

This discussion reminds me of something I read in Dierker's book. Dierker was talking more in terms of when to leave a starting pitcher in a game, but I think some of his points may be pertinent.

 

Anyway, Dierker said that often he would let a starting pitcher who had been throwing well work his way out of jams rather than go to the pen because he said there's never any way to know if your pitcher has his best stuff until he's on the mound and with a bullpen pitcher that often can be too late. Dierker used his experience as a pitcher to say that sometimes the best pitchers just don't have it on a given day and sometimes a bad pitcher can have good stuff in an outing.

 

So, once Marmol is in the game and dealing, you know he has his stuff. If he quickly disposes of the three batters in the eighth, then it's clear he's on his game. It might be wise to keep the pitcher who has the good stuff in there to finish it. With Wood, he might be good or he might not. Even if he's lights out 80%-90% of the time, you never know if the next one is his 10%-20% bad night until he takes the mound.

 

This isn't just true of Marmol-Wood though. It's more an options thing. When the roles are defined and set, there isn't any flexibility to let the hot hand continue to get the job done.

[Rob]

Keep Wood exactly where he is.

 

Move Marmol to the rotation.

 

Cut Marquis.

 

Win.

[Jehrico]

This discussion reminds me of something I read in Dierker's book. Dierker was talking more in terms of when to leave a starting pitcher in a game, but I think some of his points may be pertinent.

 

Anyway, Dierker said that often he would let a starting pitcher who had been throwing well work his way out of jams rather than go to the pen because he said there's never any way to know if your pitcher has his best stuff until he's on the mound and with a bullpen pitcher that often can be too late. Dierker used his experience as a pitcher to say that sometimes the best pitchers just don't have it on a given day and sometimes a bad pitcher can have good stuff in an outing.

 

So, once Marmol is in the game and dealing, you know he has his stuff. If he quickly disposes of the three batters in the eighth, then it's clear he's on his game. It might be wise to keep the pitcher who has the good stuff in there to finish it. With Wood, he might be good or he might not. Even if he's lights out 80%-90% of the time, you never know if the next one is his 10%-20% bad night until he takes the mound.

 

This isn't just true of Marmol-Wood though. It's more an options thing. When the roles are defined and set, there isn't any flexibility to let the hot hand continue to get the job done.

=D>

[NCCubFan]

I remember a few years back a manager (I don't remember who) was asked about bullpen roles, and his response was "rolls are for breakfast."

[Brett]

Keep Wood exactly where he is.

 

Move Marmol to the rotation.

 

Cut Marquis.

 

Win.

 

Marmol in his career as a starter - meh.

 

Marmol in his career as a reliever - ridiculoawesome.

 

Shrug.

[Hairyducked Idiot]

Keep Wood exactly where he is.

 

Move Marmol to the rotation.

 

Cut Marquis.

 

Win.

 

Marmol in his career as a starter - meh.

 

Marmol in his career as a reliever - ridiculoawesome.

 

Shrug.

]

 

 

This. Plus:

 

Marmol's Expected Wins Added Above Replacement: 2.516, second in all of baseball behind Brad Lidge.

 

Support Neutral Win-Loss Above Replacement (essentially the equivalent stat for a starter):

Zambrano: 2.6, Dempster: 1.8.

 

A high-leverage reliever is every bit as valuable as an ace starter.

[bukie]

Keep Wood exactly where he is.

 

Move Marmol to the rotation.

 

Cut Marquis.

 

Win.

 

Marmol in his career as a starter - meh.

 

Marmol in his career as a reliever - ridiculoawesome.

 

Shrug.

]

 

 

This. Plus:

 

Marmol's Expected Wins Added Above Replacement: 2.516, second in all of baseball behind Brad Lidge.

 

Support Neutral Win-Loss Above Replacement (essentially the equivalent stat for a starter):

Zambrano: 2.6, Dempster: 1.8.

 

A high-leverage reliever is every bit as valuable as an ace starter.

 

You keep saying this, but you neglect to take into account that a reliever, even the most overworked reliever in baseball, is throwing about 90 innings. A solid starter is generally throwing at least twice that for the year.

 

So, then, to maximize overall team value, the most effective pitchers on the team should be used in situations where they throw the majority of the innings. I'd rather see Marmol get 200 innings and Marquis get 90 than the other way around. Even if a Marmol 90-inning year is hugely effective.

[Hairyducked Idiot]

 

You keep saying this, but you neglect to take into account that a reliever, even the most overworked reliever in baseball, is throwing about 90 innings. A solid starter is generally throwing at least twice that for the year.

 

So, then, to maximize overall team value, the most effective pitchers on the team should be used in situations where they throw the majority of the innings. I'd rather see Marmol get 200 innings and Marquis get 90 than the other way around. Even if a Marmol 90-inning year is hugely effective.

 

I'm completely taking that into account. You are not account that not all innings are created equal.

 

There are two sets of innings to be thrown:

 

90 innings, all late in games, most of them in very close games where a run either way is the deciding factor in a win.

 

200 innnings, a large chunk of which come in games that are not within three runs and essentially already decided.

 

Those two innings packages are of roughly equal value, and I'd rather have Marmol pitching the 90 and Marquis pitching the 200.

[bukie]

I'm completely taking that into account. You are not account that not all innings are created equal.

 

There are two sets of innings to be thrown:

 

90 innings, all late in games, most of them in very close games where a run either way is the deciding factor in a win.

 

200 innnings, a large chunk of which come in games that are not within three runs and essentially already decided.

 

Those two innings packages are of roughly equal value, and I'd rather have Marmol pitching the 90 and Marquis pitching the 200.

 

Wow, if that's not extremely arbitrary and dishonest.

 

Marquis faced 2920 batters last year. 2905 of them were when the game was within 4 runs.

Marmol faced 780 batters last year. 742 of them were when the game was within 4 runs.

 

So, actually, Marmol faced more situations where the game was essentially already decided.

[Hairyducked Idiot]

Although obviously there are no guarantees, I think fans in general underestimate how quickly a game can be "over." Once a team has a four-run lead or greater, or even a three-run lead in the later innings, or a two-run lead in the ninth, they are almost certainly going to win. There will be occassional comebacks, but for the most part the game is in the bag and it doesn't matter who is pitching.

 

These are worthless, dead innings, but they still need to be thrown. Starters throw a lot of them.

 

Cubs pitchers have faced 2028 plate appearances this season, and 28% of them came when the game was not within three runs one way or the other.

 

To be specific, 526 of them did.

 

Marmol has faced 125 batters, only 24 came when the game was not within 3 runs, just 19%. And his innings are generally later in the game, too, where the value of closeness is further multiplied.

 

The importance of a pitcher's innings can be summed up in leverage index. Marmol's innings have had a leverage of 1.89, nearly twice as important as run-of-the-mill innings.

Edited May 28, 2008 by Hairyducked Idiot

[Hairyducked Idiot]

I'm completely taking that into account. You are not account that not all innings are created equal.

 

There are two sets of innings to be thrown:

 

90 innings, all late in games, most of them in very close games where a run either way is the deciding factor in a win.

 

200 innnings, a large chunk of which come in games that are not within three runs and essentially already decided.

 

Those two innings packages are of roughly equal value, and I'd rather have Marmol pitching the 90 and Marquis pitching the 200.

 

Wow, if that's not extremely arbitrary and dishonest.

 

Marquis faced 2920 batters last year. 2905 of them were when the game was within 4 runs.

Marmol faced 780 batters last year. 742 of them were when the game was within 4 runs.

 

So, actually, Marmol faced more situations where the game was essentially already decided.

 

That was last year. Marmol was not a high-leverage reliever all of last season. The choice wasn't "random relief innings" vs. "starting innings." It was 'high-leverage reliever' vs. 'starter.'

 

The WRXL stats I posted account perfectly for the importance of the inning using score and lateness, and they show Marmol being equally valuable with Zambrano, the Cubs' best starter.

[bukie]

Although obviously there are no guarantees, I think fans in general underestimate how quickly a game can be "over." Once a team has a four-run lead or greater, or even a three-run lead in the later innings, or a two-run lead in the ninth, they are almost certainly going to win. There will be occassional comebacks, but for the most part the game is in the bag and it doesn't matter who is pitching.

 

These are worthless, dead innings, but they still need to be thrown. Starters throw a lot of them.

 

Cubs pitchers have faced 2028 plate appearances this season, and almost 30% of them came when the game was not within three runs one way or the other.

 

I think you're underestimating how often a starter is throwing with the game in doubt. If you think about it, every single game a starter pitches, they enter the game tied. If you use your best pitcher as a reliever, you are banking on a lesser pitcher throwing 60-70% of the game well enough to give the Cubs a lead, so that your "crucial" inning(s) can be thrown by Mr. Dependable. However, every inning is crucial, from 1 to 9. It's just as important to pitch well enough in innings 1-6 to give the Cubs a chance to win as it is in inning 8 to maintain that chance. For every slim 2-1 8th inning lead that needs to be held down, there's a game that slipped away in the early innings due to shoddy starting pitching.

[bukie]

That was last year. Marmol was not a high-leverage reliever all of last season. The choice wasn't "random relief innings" vs. "starting innings." It was 'high-leverage reliever' vs. 'starter.'

 

The WRXL stats I posted account perfectly for the importance of the inning using score and lateness, and they show Marmol being equally valuable with Zambrano, the Cubs' best starter.

 

OK, splits from THIS year, then:

 

Marquis has pitched to 822 batters, 810 of which were when the game was within 4 runs.

Marmol has pitched to 393 batters, 379 of which were when the game was within 4 runs.

 

And this is with Marmol being overused at a rate he can't possibly keep up as a reliever, and Marquis being basically god-awful as a starter. Marquis has pitched in more "close" situations than Marmol, at a more frequent rate. Marmol is equally valuable as Zambrano, but he's only getting 1/3 of the innings to be that valuable.

[Hairyducked Idiot]

I think you're underestimating how often a starter is throwing with the game in doubt. If you think about it, every single game a starter pitches, they enter the game tied. If you use your best pitcher as a reliever, you are banking on a lesser pitcher throwing 60-70% of the game well enough to give the Cubs a lead, so that your "crucial" inning(s) can be thrown by Mr. Dependable. However, every inning is crucial, from 1 to 9. It's just as important to pitch well enough in innings 1-6 to give the Cubs a chance to win as it is in inning 8 to maintain that chance. For every slim 2-1 8th inning lead that needs to be held down, there's a game that slipped away in the early innings due to shoddy starting pitching.

 

I'm less worried about the one-run games than I am the ties, but your point in general is still wrong. No matter who the starter is, there are going to be some high-leverage innings at the end and some low-leverage innings at the end.

 

Every inning is not crucial. There are many innings that aren't crucial at all. Toward the beginning of the game, innings are of approximately average crucialness. As the game goes along, the crucialness varies wildly. While the first inning is usual pretty average, the ninth inning can be incredibly crucial or incredibly pointless. There are enough highly crucial innings at the end of games for there to be one or two relief pitchers on the team who are of just as much importance as starters pitching roughly twice as many raw total innings.

 

Instead of speaking in vague, pointless generalities like my first paragraph above or this:

 

"For every slim 2-1 8th inning lead that needs to be held down, there's a game that slipped away in the early innings due to shoddy starting pitching."

 

We can quite precisely measure the relative odds of a start or a relief appearance deciding who wins the game. That is what SNLVAR and WRXL do. They measure the odds of the team winning (based on relative score, baserunners, inning and outs) before and after the pitcher pitched, and credit or debit him with the difference. And Marmol has come up just as valuable as Zambrano, and moreso than Dempster or anyone else in the rotation.

 

Given that most pitchers improve their performance when going to relief, and that Marmol's performance improved dramatically in relief, and that his value to the team is statistically at least as good as any ace, it'd be lunacy to convert him to starting.

[Hairyducked Idiot]

That was last year. Marmol was not a high-leverage reliever all of last season. The choice wasn't "random relief innings" vs. "starting innings." It was 'high-leverage reliever' vs. 'starter.'

 

The WRXL stats I posted account perfectly for the importance of the inning using score and lateness, and they show Marmol being equally valuable with Zambrano, the Cubs' best starter.

 

OK, splits from THIS year, then:

 

Marquis has pitched to 822 batters, 810 of which were when the game was within 4 runs.

Marmol has pitched to 393 batters, 379 of which were when the game was within 4 runs.

 

And this is with Marmol being overused at a rate he can't possibly keep up as a reliever, and Marquis being basically god-awful as a starter. Marquis has pitched in more "close" situations than Marmol, at a more frequent rate. Marmol is equally valuable as Zambrano, but he's only getting 1/3 of the innings to be that valuable.

 

Four runs is a poor cutoff, and you aren't accounting for the lateness of the innings as well. Four runs in the first is different from four runs in the ninth.

 

Again, instead of speaking in generalities, why not use leverage index?

 

And you don't seem to understand what "valuable" means in that context. WRXL and SNLVAR aren't rate stats. It isn't "this valuable per appearance." It's total value contributed.

 

The reason he can be as valuable while pitching fewer innings it that he is more productive as a reliever than he would be as a starter (he's not maintaining a sub 2.00 ERA in the rotation) and that his innings to this point have been 89% more likely to influence the outcome of a game than average innings.

 

He should be pitching fewer relief innings, to be sure, but that's because he's had some pointless blowout innings padded on to his stats. Almost all of his value has come outside those innings.

[Tracer Bullet]

That was last year. Marmol was not a high-leverage reliever all of last season. The choice wasn't "random relief innings" vs. "starting innings." It was 'high-leverage reliever' vs. 'starter.'

 

The WRXL stats I posted account perfectly for the importance of the inning using score and lateness, and they show Marmol being equally valuable with Zambrano, the Cubs' best starter.

 

OK, splits from THIS year, then:

 

Marquis has pitched to 822 batters, 810 of which were when the game was within 4 runs.

Marmol has pitched to 393 batters, 379 of which were when the game was within 4 runs.

 

And this is with Marmol being overused at a rate he can't possibly keep up as a reliever, and Marquis being basically god-awful as a starter. Marquis has pitched in more "close" situations than Marmol, at a more frequent rate. Marmol is equally valuable as Zambrano, but he's only getting 1/3 of the innings to be that valuable.

 

Four runs is a poor cutoff, and you aren't accounting for the lateness of the innings as well. Four runs in the first is different from four runs in the ninth.

 

Again, instead of speaking in generalities, why not use leverage index?

 

And you don't seem to understand what "valuable" means in that context.

 

4 runs is a poor cutoff? Is 3 that much better?

 

And, you're right, lateness matters. A 4-run lead earlier in the game is much less "safe" than a 4-run lead late. So if Marmol pitches in a 4-run game in the 8th, we're much more likely to have already won or lost before he comes in, than if we're up or down 4 in the 2nd.

[bukie]

And you don't seem to understand what "valuable" means in that context. WRXL and SNLVAR aren't rate stats. It isn't "this valuable per appearance." It's total value contributed.

 

I understand perfectly. Total value contributed, over 90 innings. You don't think he'd contribute more value than that over 200 innings? Sure, his value per appearance would drop somewhat, but his overall value would most likely increase significantly.

 

Four runs in the first = four runs in the ninth. In the end, that's still four runs the team has to account for. If you prefer the cutoff to be 3 runs, the stats look the same. Starters pitch in more close situations than relievers. It's not presumptive, it's a fact. Which has been shown to you twice, in very specific ways. If you prefer to keep your head in the sand about it, though, there's no point in discussing it further.

[Hairyducked Idiot]

And you don't seem to understand what "valuable" means in that context.

 

4 runs is a poor cutoff? Is 3 that much better?

 

And, you're right, lateness matters. A 4-run lead earlier in the game is much less "safe" than a 4-run lead late. So if Marmol pitches in a 4-run game in the 8th, we're much more likely to have already won or lost before he comes in, than if we're up or down 4 in the 2nd.

 

Which is a good reason not to use Marmol in a four-run game. A starter cannot be directed to pitch important innings. He's going to go out there on his turn and pitch as many innings as he can, regardless of score. A reliever can be targeted to either high-leverage (if he's great) or low-leverage (if he's not) situations.

 

Instead of using a specific number as a cutoff of importance, let's use leverage index, which measures the average importance of the pitcher's exact and actual appearances.

 

Marmol's leverage index is 1.89. His innings are almost twice as important as average innings. If Piniella wasn't a ninny and used him in a few blowouts, it'd be well over 2.

[Hairyducked Idiot]

I understand perfectly. Total value contributed, over 90 innings. You don't think he'd contribute more value than that over 200 innings? Sure, his value per appearance would drop somewhat, but his overall value would most likely increase significantly.

 

No, it wouldn't. History has shown us that the best high-leverage relievers usually end up with roughly the same value contributed as the best starters.

 

Four runs in the first = four runs in the ninth.

 

That's both wildly wrong and missing the point.

 

I don't have a four-run example easily available, but there was a perfect three-run example last night.

 

Let's say a pitcher comes in to start the top of an inning with a tie game and allows three runs.

 

If that inning was the first, that pitcher has taken his team from a 50% chance to win the game to a 19.8% chance to win the game, using historical results. He's cost his team .302 of a win.

 

If that inning was the ninth, that pitcher has taken his team from a 50% chance to win to a 1.9% chance to win, costing his team .481 of a win. His team has 1/6th the chance of winning that it did before.

 

If we talk about the difference between a shutout inning and giving up a single run, the difference is even more dramatic.

 

A pitcher comes into the top of the inning with the game tied. If he pitches a shutout inning, then his team now has a 55% chance to win the game if that inning was the first, and a 64.6% chance if that inning was the 9th. If he allows exactly one run, that team now has a 42.5% chance of winning if it was the first, and a 12.1% chance to win if it was the ninth.

 

Top of first, tied:

Zero runs = 55% chance of winning, one run = 42.5% chance of winning.

Top of ninth, tied:

Zero runs = 64.6% chance of winning, one run = 12.1% chance of winning.

 

Clearly, the ninth is wildly more important. All innings are not created equal.

 

Starters pitch in more close situations than relievers. It's not presumptive, it's a fact. Which has been shown to you twice, in very specific ways. If you prefer to keep your head in the sand about it, though, there's no point in discussing it further.

 

Starters do pitch, on average, in more close situations than relievers.

 

That does not refute that Carlos Marmol's innings are just as important as a starter's in sum for two reasons. One, close is only half the equation of importance. Lateness is the other. And second, Marmol doesn't have the usage pattern of an average reliever. He has the usage pattern of a reliever targeted for use in close games.

[bukie]

No, it wouldn't. History has shown us that the best high-leverage relievers usually end up with roughly the same value contributed as the best starters.

 

Win Share totals disagree with you. Over the past four years, the best reliever year (2004 Brad Lidge - 17.2 pitching win shares) comes in as the 38th best pitching year over those past four years, the equivalent of 2007 Zambrano/Lilly.

 

A good example of a pitcher that was both an effective starter and reliever is John Smoltz. In 2004, he was the fourth most effective reliever in baseball (Lidge, Gagne, Benitez). He finished the year with 12.1 Pitching Win Shares. The next three years, he was an effective starter, with Win Share totals of 21.1, 19.4, and 17.5.

 

I think your assumptions lean heavily on the concepts of win probability, which are really quite arbitrary and provide little predictive value. It does provide for some fun charts to look at, though.