I think he has some idea, he's just being coy.

Or, he likes fish, has some idea, and is just being koi.

each year, on the average, 1-2 teams would win or lose almost 15 games more than they should given their talent, another 1-2 will do 10 games better or worse, etc. That is guaranteed (again, on the average). Guaranteed. And that is with a manager having exactly zero influence on the way the players perform, given their talent. In fact, that is without anything having any influence, positive or negative, on the players’ or team’s performance above or below their true talent.

So if that is the case in our hypothetical league with no manager influence whatsoever, how can we have any idea whether a manager deserves credit or blame for a team’s over or under-performance? We can’t. It’s that simple.

This is certainly correct, like many of MGL's pronouncements, from an outside perspective. I would liken it to handicapping advice. Let's say a horse won its last race, under a brand-new jockey. Jockey's the reason, right? But at the same time you notice that the horse moved down in class from a $30K allowance race to a $20K claiming race. And carried three more pounds. And went onto the turf from dirt, and moved up from six furlongs to a mile. Who's to say at this point how much the change of rider factored in?

From the inside perspective, of course, the horse has to have a rider, and you have to be sure you are giving the horse to a competent jockey, that the trainer and jockey communicate well, that the jockey knows the peculiarities of the track, the starting gate, the other jockeys, and numerous other factors. But at the same time, you can hire a new rider with excellent qualities and the horse still lose for many, many reasons, prime among them being that it's not a very good horse.

But I liken MGL's edicts to handicapping advice because they are not really about how things are run from the inside, but about how we should interpret them from the outside. When he said in a recent entry linked in a different thread that a manager should never look at what a batter had done lately, this was obvious nonsense on one level (and hence was blamed by me and others as confusing). But it was brilliant advice to bettors. If a true-.300 hitter has hit .500 over the last week, his chance of a hit in the next AB is 30%, and it's a sucker bet to take shorter odds on the chance of a hit. But if you are a pitching coach, catcher, pitcher, or manager, your adaptation to that recent .500 streak is part of what brings the hitter back down to .300.

And so with managers. From the outside, it absolutely looks like all they do is keep their teams close to their true-talent level of performance, and that is the way you should bet. From the inside, just treading water like that is a set of balancing acts, magic tricks, psychiatric interventions, strongarming, and PR spinmeistering on a daily basis. For all we know, Lou Piniella did his best managing in Tampa Bay, and Gene Mauch was the greatest manager ever. There's just no way (and may never be a way) of quantifying their work to prove it (as MGL very rightly says here).

I generally agree that we don't know how much or how little a manager affects his team's success and failure. However...

It is ridiculous because, one, he or anyone else has no idea how good or bad Willie was at managing his bullpen

It's not that hard to find out if you either take notes while watching games or peruse the box scores. Does he over-use and under-use his pitchers? A rough look at the Mets' bullpen reveals that it doesn't look like he does. Heilman has pitched 5 more innings than anyone else (2nd is Jorge Sosa), but it's Heilman's job to pitch 70+ innings. Does he not leverage his bullpen correctly (ignoring Sabermetric wisdom that it's wise to use the "closer" before the ninth inning in high leverage situations sometimes)? This would take some research but is totally reasonable to hypothesize. Does he give his pitchers adequate rest? Another easy one to hypothesize.

two, what does that even mean

I don't know how MGL is confused by this, but given the prominence of bullpens, good bullpen management will translate to a significant advantage for a team.

three, he offers no evidence to support that claim

Valid criticism, though he urges his readers not to read the article so it's moot anyway.

and I doubt there is any

There can be, it's just not stuff that people log on a regular basis.

and four, how many wins can a good or bad “managing” of a bullpen in “may, June, and July” be worth?

As much as any other game... a game in April = a game in September. They're worth one win or one loss in the standings.

I have to agree with MGL on this one. He's right about Sheehan making claims he has no firm basis to make.

Especially the "put the division away" comment. The division was "put away."

And that happened.

And that is with a manager having exactly zero influence on the way the players perform, given their talent.

This premise is completely unsupportable. Just because you can't measure the manager's influence doesn't mean there isn't one.

I mean, as long as we're saying "we don't know."

Fire him anyway!

(It is ridiculous because, one, he or anyone else has no idea how good or bad Willie was at managing his bullpen, two, what does that even mean, three, he offers no evidence to support that claim, and I doubt there is any, and four, how many wins can a good or bad “managing” of a bullpen in “may, June, and July” be worth? .5 wins? 1 win? 1.5 wins? “Put the division away?” Please!

Obviously you don;'t watch the Mets on a regular basis
Willie repeatedly uses poor pitchers in high leverage slots
Willie uses guys who should be loogies against RHPs
For example: Show is a LOOGY, LHBs hit .204/.308/.247 against him in 2007, .236/.333/.292 in 2006 and .188/.260/.241 in 2005.
LAst year RHBs hit .316/.390/.574 against Show, for his career RHBs have hit him to the tune of .293/.364/.466 whereas LHBs have managed only .225/.300/.294

In the 2 years prior to coming to the Mets, Show's managers had made an effort to limit his exposure to RHBS- last year? Willie let him face RHBs 50% more often than LHBs

Willie let Mota pitch repeatedly in high leverage situations last year, contributing to the "collapse".

Willie's bullpen [mis]management last year was an outlier, ordinarily I don't think manager's use or misuse of his pen has much of an effect, but Willie is in a league of his own

That is guaranteed (again, on the average).

As guarantees go, that one's not exactly Namathian.

This premise is completely unsupportable. Just because you can't measure the manager's influence doesn't mean there isn't one.

I mean, as long as we're saying "we don't know."

He is saying that if you completely remove the manager's influence, you would still have those kinds of swings.

Oh, please change the spelling of his first name in the blog. It's correct elsewhere, but the title has it wrong.

Willie's bullpen [mis]management last year was an outlier, ordinarily I don't think manager's use or misuse of his pen has much of an effect, but Willie is in a league of his own

I think MGL's argument would be that it's impossible to say that he mismanaged the bullpen, because the information that you have about relievers is almost always based on tiny sample sizes. L/R splits are particularly problematic. If a good reliever has a few bad outings in a row, people want the manager to stop using him, even if his previous track record is solid. This is the kind of thing for which managers are unfairly put in the pillory.

L/R splits are particularly problematic. If a good reliever has a few bad outings in a row, people want the manager to stop using him, even if his previous track record is solid. This is the kind of thing for which managers are unfairly put in the pillory.

In your opinion, should Scott Schoeneweiss be used as a LOOGY, or not?

On an abstract level, I entirely agree with MGL. You can't look at L/R splits, or games lost by relievers, or Pythag record, or anything else so abstracted from the decisions themselves. It's like trying to judge a salon manager's personal management skills by amount of hair left on the floor at the end of shift.

However, on a non-abstracted level, it is possible to gauge a manager's ability to handle his bullpen. I listened to the Red Sox - Mariners game the other night, and that there was no-one warming in the pen when King Felix suffered a mini-implosion was definite mismanagement, in my opinion. The infamous Grady Little/Pedro stuff was an infamous example of high-profile bullpen mismanagement, again in my opinion. Every one of us has probably shook our head when a certain reliever was brought in or not brought in at a certain point, regardless of the results that followed (more on results later). Now, if one was to collate instances of bad handling of pitchers over a 162-game season, and then compared them with a similar list collated for other managers (presuming a certain level of objectivity in the coding process, and ability on behalf of the collaters to make rational judgements on bad usage), we could absolutely arrive at a reckoning of who had made more 'errors' in bullpen management, in the view of a given set of observers. It goes without saying the observers would need to be skilled in their art, and how you find them is another question, but it is certainly possible.

That said, the key is that this couldn't, by its nature, be a results-based analysis, only an analysis of the manager's process. If the manager chooses Jim Parque over Papelbon in the bottom of the 8th with bases loaded, game tied, and Jim Parque strikes out the side, that may have worked, but the process behind making the decision may well be flawed (of course, there may be reasons why you'd go with x over y, but that should be part of the analysis). Just like normal work situations, a staff can be mismanaged and still produce good results. But a non-abstracted analysis of a manager's bullpen management is possible, it's just that it would be a ton of work for potentially little reward. By focussing on results, and ignoring process, most commentators confuse bullpen success with bullpen management skills, and, by extension, team success with team management skills (which is, of course, pretty common in real-life). The correlation of these factors, especially considering the manager has little input in the personnel side of the business, and that baseball is essentially an individual game, is never going to be that strong.

1.5 wins? “Put the division away?” Please!)

Well, 1.5 more wins would have "out the division away". That matters.

In fact, each year, on the average, 1-2 teams would win or lose almost 15 games more than they should given their talent, another 1-2 will do 10 games better or worse, etc. That is guaranteed (again, on the average). Guaranteed.

I'll need some demonstration, not just hypotheticals. Not that they *could*, but that they, in fact, do. This is just as made up as whatever he's ######## at Sheehan about.

In your opinion, should Scott Schoeneweiss be used as a LOOGY, or not?

Sure he should. That's not the question. The question is if he should be used against a right-hander in certain situations. There are a number of factors outside of Schoenweis' L/R splits that go into the individual decisions to use him in specific situations, specifically, the alternatives.

I'm not arguing that Randolph didn't misuse Schoenweis, but one has to go a lot further than that to support Sheehan's indictments. What I agree with is MGL's fundamental assertion that it is a disservice to baseball analysis when you make statements like the one he is criticizing.

(162*.5*.5)^0.5 = 6.36
6.36 * 2.2 = 13.99

(162*.5*.5)^0.5 = 6.36
6.36 * 2.2 = 13.99

Is that an answer to my post?

Let's think about this for real. Let's take three pitchers. Billy Wagner is the Mets' best reliever. He is not used as often as the Mets' other relievers. Jorge Sosa was the worst Met reliever. He was also the most often used Met reliever at the time of his release. Jorge Sosa has demonstrated a massive and statistically significant platoon split over the course of his career. Willie Randolph repeatedly used Sosa against LHB in high-leverage situations. Scott Schoeneweis has demonstrated a massive and statistically significant platoon split over the course of his career. Willie Randolph has repeatedly used Schoeneweis against RHB in high-leverage situations.

MGL is taking something way out of context (that is, the context of actually studying the subject of Willie Randolph's bullpen usage) and made a sweeping renunciation of any possible conclusion because the effect of the subject (which he clearly has not studied) is currently immeasurable. But just because it can't be measured accurately right now doesn't mean it's not real.

#### sabermetrics. Fire Willie.

"Willy"? The movie-star orca?

Is Willie responsible for this? Who knows? MGL is right. That ain't the point. Willie is the manager and is therefore accountable. If his team wins, he gets a parade. If his team loses, he gets fired. That's the job he took.

I'm not arguing that Randolph didn't misuse Schoenweis, but one has to go a lot further than that to support Sheehan's indictments.

Schoeneweiss, Sosa, Feliciano, Mota.

I'll need some demonstration, not just hypotheticals. Not that they *could*, but that they, in fact, do. This is just as made up as whatever he's ######## at Sheehan about.

It's not that hard to look up. Here's a list of all the teams in recent years that have deviated from their Pythagorean record (which is what I assume he's referring to) by ten or more games:

2007, Arizona: +11
2006, Cleveland: -11
2005, Arizona: +11
2004, New York Yankees, +12

That's it, going back to 2001. From 2001 to 2003, there were no teams missing their Pythag by ten games.

It's possible, as I said, that he's defining "talent" some other way, in which case what he's saying has no meaning, because the "true" record for a given group of talent could be anything you want it to be. But if he's defining true talent by Pythag record, he's wrong. No team in recent history has missed its true talent by 15 games, and less than one per year misses it by ten.

But if he's defining true talent by Pythag record, he's wrong.

Almost by definition, he isn't. Using pitchers incorrectly will result in more runs allowed.

It's possible, as I said, that he's defining "talent" some other way, in which case what he's saying has no meaning, because the "true" record for a given group of talent could be anything you want it to be. But if he's defining true talent by Pythag record, he's wrong. No team in recent history has missed its true talent by 15 games, and less than one per year misses it by ten.

I don't think that's what he's doing. I think he's definining it by what a team's established level of performance projects them to do compared to what they actually do.

So then what he's saying is, "Every year my projection system misses at least one team by 15 games." Right? And this is supposed to be why we can't fully evaluate managers?

That's it, going back to 2001. From 2001 to 2003, there were no teams missing their Pythag by ten games.

It's possible, as I said, that he's defining "talent" some other way, in which case what he's saying has no meaning, because the "true" record for a given group of talent could be anything you want it to be.

I think if you use BPro's 3rd order WP% you'll get pretty close to the same resulte that pythag does.
MY off the top of my head guess was that typically the most any team would deviate from it's "true" level is 12 game or so, and that one team out of 30 only deviates that much once every 3-5 years... My off the cuff assumption is probably no less accurate than his apparent assumption that every year 1-2 teams deviate by 15.

I don't think that's what he's doing. I think he's definining it by what a team's established level of performance projects them to do compared to what they actually do.

This is a problem with MGL's writing that was addressed in earlier threads- it is quite possible that what he means by "true talent level" is different than what someone else would define as same, and how he measures it is different than hat someone else would measure it as.

I suppose that if you take a team's roster, and use a run estimator to determine each players established level of ability over a 3 year period, then use that to generate runs scored and allowed by that roster, and then use pythag to determine the team's "true talent" WP- that such "true talent" WPs would deviate from actual records a bit more than simply using pythag would.

He is saying that if you completely remove the manager's influence, you would still have those kinds of swings.

And I'm saying that since all players have had managers, there is no basis for this observation.

Schoeneweiss, Sosa, Feliciano, Mota

Okay, what I said originally was too broad, so I can see why you're calling me on this. One can probably fairly assess that overall Randolph used the bullpen suboptimally. The problem is turning that into saying that it caused the Mets to lose the division. To whatever degree it's measurable in terms of its expected effects, Sheehan didn't do the math. And doing the math you have to consider the constraints on the roster in terms of, say, using someone else against a particular righty on a particular day. While we can see broad trends (e.g., Schoeneweis used too much against RH batters), managers do not always have the optimal reliever available. How often did Randolph have a better alternative in what was a pretty lefty-heavy bullpen? I'm guessing that he was stuck with using Schoeneweis against righties more than you would want, although he may well have gone too far with it.

How often did Randolph have a better alternative in what was a pretty lefty-heavy bullpen? I'm guessing that he was stuck with using Schoeneweis against righties more than you would want, although he may well have gone too far with it.

Your guess is wrong. In fact THIS year he is using Schoenweiss more appropriately, specifically, he's now decided that Show is loogy, whereas last year he used Show as more of a long reliever because that was the role he decided beforehand that Show would fill, come hell or high water, because, as he himself said, Show had once been a starter and there fore it would be a waste just to use him for a batter or two.

However there is a reason Show was no longer a starter- if you let RHB's hit .293/.364/.466 (career average in 2725 PAs- in recent years it's been worse) YOU CAN'T START, period end of story. In fact Show's career number have him giving up .280/.352/.435 as a starter and only .256 /.332/.373.

He is much more effective in relief because he's been platooned (loogied) for the most part- RHBs hit him just as hard as a reliever as they did when he started- in some- every warm body in the Mets bullpen was a better option to face RHBs than Show, and Willie let him face 157 RHB PAs- the most he has ever seen in the relief role- and every Mets blog was screaming about it from day one.

Show, with one of the most pronounced L/R splits you will ever see saw 50% more RHBs than LHBs.
Sosa, who kills righties and gets mauled by lefties, saw a greater % of LHBs than Show did.

How is that possible?
1: Willie mismanaged the pen.
2: Willie was outmanaged by his opposite number in game critical situations.
3: A combination of 1 & 2.

I vote 3

You have to remember that in Willie's rookie year he kept an ineffective Looper (who was injured as it turned out) closing games while he had a better alternative available (Heilman), because Willie said Looper "deserved" to close, because Looper had the most saves.

I may be ranting, but no one who watches a lot of Met games, or even follows the boxscores, defends Willie's bullpen management. I don't care that his impact is not quantifiable down to a decimal point- he's terrible, he's subjectively terrible, and when you start analysing his usage of specific players and situations you quickly see stuff like the Show/Sosa numbers up above.

He is saying that if you completely remove the manager's influence, you would still have those kinds of swings.

Nit: MGL's not saying "remove the manager's influence", he's saying "assume the manager's influence has the same net effect on every team." He's right as far as he goes - you WOULD see many of the same types of swings. What MGL doesn't know (nor does anyone) is whether the "same" teams would show those performance variations, or whether different teams would show the swings when you DID account for manager's influence. Accounting for manager influence might reduce the overall variance - or might redistribute it instead, so that different teams would underperform or overperform.

That said, you COULD build a model of expected performance, using something close to an optimal distribution of appearances in high/medium/low leverage situations for your best pitchers and your worst pitchers, and evaluate the team's actual performance against that model. It's not an EASY thing to do, mind you - and it might not be especially worthwhile, because you might very well find that the variance is too small to matter in all but a few extreme cases - but we might all learn something from the attempt. Most of the complaints about bullpen usage fall along the lines of (a) why can't the closer pitch MORE? and (b) why can't everyone else pitch LESS? Outside of a holistic model of bullpen usage - incorporating workload distribution and leverage - I don't see that either question can be easily answered.

-- MWE

Worth pointing out that Sheehan's criticism of Randolph's bullpen management was a very small part of the column MGL references and was close to a throwaway line. His point was largely that Randolph is getting all the heat but his level of culpability in this mess (if two games under five hundred with an in-the-black run margin can be called a "mess") is debatable at best.

Happy Base Ball

You have to remember that in Willie's rookie year he kept an ineffective Looper (who was injured as it turned out) closing games while he had a better alternative available (Heilman), because Willie said Looper "deserved" to close, because Looper had the most saves.

This really isn't fair to Willie. Heilman's ERA as a reliever in his first 22 games as a reliever was 4.40. He put up some superb peripherals but the actual results were quite mediocre. Heilman was dominant in his last 37.1 IP though, posting a 0.48 ERA.

If anyone should have been moved into the closer's role that year, it should have been Kool-Aid.

Yeah, the thing with Sosa seeing so many lefties was weird. What was that about?

Yeah, the thing with Sosa seeing so many lefties was weird. What was that about?

Maybe Willie was confused. "Is the lefty Soweiss, and the righty Schoensa? or the other way around?"

We should also consider Lastings Milledge, who was clearly in Randolph's doghouse. Certainly it should be some part of a manager's responsibility to handle a clubhouse and ease a rookie with adjustment issues into success. Or at least to pretend like he cares to.

Makes it hard to get fair value when you ultimately trade him.

This really isn't fair to Willie.

Yes it is, and I should have mentioned Kool-aid too.
Looper was ineffective, and was visibly pitching with diminished stuff- were you surprised when it turned out he'd been pitching while injured for the whole year?

AND, more importantly, Willie really did say that Looper "DESERVED" to close because he had the most saves.

#31. Agree completely.

AND, more importantly, Willie really did say that Looper "DESERVED" to close because he had the most saves.

Looper started off poorly and he was criticized a great deal but from April 11th to Auguest 2nd, he posted a 2.18 ERA, certainly not worthy of a demotion and his ERA on August 2nd was 2.98. He collapsed at the end pf the season and that made his numbers look pretty bad. In reality, Willie probably used his bullpen in a very SABR-friendly manner. He used his relief ace (Kool-Aid, later Heilman) in the most important situations for most of the year and his less effective pitcher in the "closer role".

just checking is MGL basically saying that there is no way to prove that Randolph is hurting his team, so the answer is not to fire him? And then didn't he basically go a little further and say there is no way to prove anything about a managers effect on a team? does that mean that all managers should be guaranteed a lifetime job because there is no way to prove anything about them?

I honestly think there is nothing wrong with fans, saber friendly writers etc for voicing their opinion on this issue. I honestly think that Randolph and Yost have to go, just because. Heck if there was evidence they were doing well, I would still argue that a change has to be made for change sake. At some point in time the team needs to reach expectations or prove that it's not the managers fault by getting rid of him. It's a cold, callous way to look at it, but it's the nature of the game. Reasoning is not mandatory for making a decision.

MY off the top of my head guess was that typically the most any team would deviate from it's "true" level is 12 game or so, and that one team out of 30 only deviates that much once every 3-5 years... My off the cuff assumption is probably no less accurate than his apparent assumption that every year 1-2 teams deviate by 15.

As fate would have it, I built a little database of Diamond Mind projections with actual results along with that. While this isn't the same thing as "true talent", any model of true talent runs into the same problems as DM, so I'll use it as a proxy. Anyway, here's a breakdown of deviation from expected wins over the last ten seasons:

0 14
1 27
2 26
3 21
4 30
5 22
6 19
7 21
8 16
9 17
10 12
11 18
12 6
13 10
14 9
15 5
16 5
17 5
18 3
19 4
20 2
21 1
22 2
23 2
24 1
27 1
31 1

The 31 is, of course, the 2001 Mariners. They gave 24 of those games back in 2004.

As fate would have it, I built a little database of Diamond Mind projections with actual results along with that.

I love Diamond Mind, but I have to tell you, when running multiple seasons you see "fluke" seasons occurring much more frequently than they do in real life.

What I mean is this:
Adrian Beltre: career 107 hitter, 2nd-4th best seasons: 114, 112, 111
but in 2004 he posted a 163

that's extraordinary
but in DMB if you run a full league, every year 2-3 players will outhit their "true talent" level by 50 OPS+ points or more, and 2-3 will under hit by a like amount.

Anyway, depending upon your definition of "true talent", my pick for the recent team that most overperformed was the 2003 Royals- likely, imho, a "true talent" team of 60-65 wins, yet they won 83. Their pythag was 78-84, but even that was too high
Angel Berroa, a truly awful player, actually played well that year.
Beltran had his best year as a Royal
Aaron Guiel gave them 350 productive at bats
Joe Randa had one of his better years
Darrell May, 210 ip, ERA+ of 130 (career BABIP of .292, it was .251 that year as space aliens were apparently directing every ball hit off him into gloves)
They got 126 ip of 125 era+ pitching from Affeldt- he' snever been as valuable before or since
Jose Lima, 73 ip, ERA+ of 100
The Royals ran out scads and scads of pitchers, but except for Chris George's 94ip, no one truly SUCKED
and they got up to 83 wins...

2001 Mariners? maybe, They really were a true talent 95-100 win team at least, so I didn't think of them, but man oh man 116 wins...
In hindsight, only Brett Boone's individual performance stands out
[actually Sele's season seems to stand out, 116 ERA+ in 200+ ip?, but his 2001 was right in line with his career up to that point, it was after 2001 that he sucked every year... I'd totally forgotten that he really was a good pitcher once upon a time)

I love Diamond Mind, but I have to tell you, when running multiple seasons you see "fluke" seasons occurring much more frequently than they do in real life.

This is true if you're running the ACTUALLY played seasons, but not if you're running the PROJECTED seasons, which I think is what he meant he used.