Tuesday, April 20, 2010

Back-of-the-Napkin Analysis: Have Bullpens Become Better?

Sometimes it’s what you don’t see that’s important.

Monday night I was in my first live chat session, hosted by Phil Mackey of ESPN 1500. Semi-predictably, I settleD into the role of elderly curmudgeon, ala Patrick Reusse, except without the charm. That became especially apparent when the topic of bullpen usage came up.

In 1985, Bill James tackled the changing role of relievers in Bill James Historical Baseball Abstract. He noted the changing roleS and predicted further changes. And he was right, as roles became more defined and we saw the rise of the one-inning closer by the 90s. No longer did we see the best relief pitcher enter the game at a critical time early in the game. Their new role was to protect a lead in the ninth inning.

It’s become a hotly debated topic, especially when a game is lost, and I find myself on the opposite side of it than many of my peers. I have no problem with stricter roles in a bullpen, even if that means "saving" the closer for the end of the game. I suspect there are long-term benefits that equal the short-term gain in using them in a more flexible manner:

TwinsGeek: But Mackey, managing a bullpen could be similar to managing any other group of people - they perform best when expectations are set.

Aaron Gleeman: Plenty of setup men seem to do just fine being thrown into all kinds of spots without knowing exactly when and where.

Phil Mackey: Right TwinsGeek, but they'd get used to it.

But Monday night was the first time I really asked myself the question:

TwinsGeek: So do we thing that bullpen's blow MORE games now than they did before 1985? Do we have any evidence of this?

Aaron Gleeman: John, bullpens now also have 3-4 extra pitchers and specialized roles. So you'd expect them to be better, in general.

So let’s do a little back-of-the-napkin analysis today around this. I’m hoping to get a sense whether:
  1. Have teams become better at winning games with close scores, and did this happen at about the same time as the rise of the closer?
  2. Is it possible that while saving the "closer" for late innings helps the team preserve late leads that it has had a similar negative impact on leads earlier?
I’m going to start by using this Win Expectancy Finder to record, year-by-year, what percentage of the time the home team was able to preserve a one-run lead in the ninth inning between 1977 and 2006. Then I’ll do the same for the seventh and eighth innings. And then I’ll chart it and see if I see any trends.

(On caveat – I’m going to skip two years. I’m going to skip 1999 because the site that I’m using doesn’t have any data for 1999. And I’m going to skip 1994 because it was the strike-shortened year. Because of that, I'm not comfortable with the sample data size, because I'm seeing some goofy results. For instance, in 1994 fewer teams could hold a one-run lead going into the eighth than could hold a one-run lead going into the seventh. Think about that for a minute.)

So, how does it look? For year to year comparisons, it’s a little bit of a mess:

Yikes, that's like my EKG during Game 163. It’s hard to see any trends here because the data jumps around a lot. So let’s smooth it a bit by taking the average for every five-year period instead…..

That’s odd. During the late 70s and 80s, bullpens didn’t change very much in protecting a ninth inning lead – that green line is pretty level at first. However, teams got about 5% better at preserving a win in the seventh or eighth inning between 1982 and 1990. That’s a very strange result if, as the theory goes, the best relievers were being moved to work strictly the ninth inning. I’d expect exactly the opposite – a slight uptick for that green line and declines in the red and blue lines.

I’d also feel a lot more comfortable if those lines evened off a bit in the 90s, but instead they decline slightly back to previous levels, except for the eighth inning guy who had a couple of very good years in 2001 and 2002. So there are obvious plenty of other factors impacting these lines. It might be that the previous increase in middle-relief effectiveness had nothing at all to do with the change in usage.

But I think one thing that I don’t see here is relevant. I don’t see any evidence that at any macro-level, the change in bullpen usage that swept through the 80s hurt team’s ability to hold onto one-run leads. We can claim that utilizing a team’s best relievers in this new modern way isn’t particularly smart, but it sure doesn’t look like it spectacularly stupid, either.


Dave Nelson said...

Why even try draw a conclusion about the best way to utilize a bullpen from this study? Its clearly way too limited. Even if there was some shift the ability to maintain a 1 run lead and bullpen management are not independently related. The twins currently have 3 loogys in their bullpen and lots of teams carry ground ball specialists, not to mention added emphasis on quality power arms in setup roles ect. Bullpens are now full of specialist guys that didnt really exist 30 years ago. Better bullpen construction probably plays a bigger role in preserving leads than bullpen management does. Its hard to argue against a team in theory being better off having the best pitcher pitch in high leverage situation instead of saving him for lower leverage save situation simply to acquire a meaningless stat. Saving krod for the 19 inning while lesser pitcher could have lost the game at any point before that without krod even pitching is bad management. I remember 2 years ago at fenway baker pitched 7 shutout innings or something and it was 0-0 when he left. We didnt have the lead so we could put setup man in. Wasnt a save situation so nathan was out. We ended up losing the game when brian bass gave up 8th runs. In theory there is no way have a role exclusive bullpen should produce better results than a flexible bullpen that allows for good pitchers to pitch in higher leverage situations. Theres probably some mental component to knowing a role but youd think players could adjust.

TT said...

Part of the problem is factual errors that get repeated by bloggers who have a point of view and twist the fact to fit it.

Bullpens don't have 3 or 4 more pitchers than they used to. For a long time pitching staffs generally had ten pitchers - 4 starters and 6 relievers.

As they went to five pitcher rotations, teams generally moved to 11 pitchers, with the same six relievers. Now, many have 7 relievers and occasionally 8. But the number of outs the bullpen is expected to get has also increased as starters don't go as deep into the game.

Second, where is the evidence that relievers were EVER used in the manner described. I think this is also mostly invention, based on no actual analysis but supportive of the preferred conclusion.

In fact, before the current patterns, bullpen management was far more seat-of-the pants with different managers having different approaches depending on the makeup of their rotation and bullpen. One manager might be quick to relieve a starter, another might tend to let them pitch out of trouble. And if a manager had Hoyt Whilhelm or Mike Marshall available, they might go to them when they otherwise would let the starters pitch out of trouble. And that isn't even considering the impact of the DH rule.

Second. There seems to be an ignorance of how bullpens are managed even today. In fact, most pitchers have a pretty good idea of when they are likely to be called upon. This time of the year those roles are still being worked out, but by July pitchers will all have roles.

By then the mopup guy will know if the game is close, they aren't likely to pitch. And the top setup guys will know that they are likely to be called on in the 7th or 8th inning to pitch, but will have to go no longer.

Which is why its not hard to argue that using your best reliever to preserve games you are already winning, is a better idea than using them to try to win games that are still in doubt. When one pitcher is going handles the ninth, the rest of the staff know it only has to handle the first 8 innings.

The other thing is that the so-called "high leverage" situations earlier in the game are mostly illusions. This is the problem with your analysis. The question is not only how many one run leads are lost, but how many of those leads are extended to 3 or 4 run leads by the time the game ends.

If you look, there are three times when teams score runs in bunches. One is at the start of the game when the other team's starter is having a bad day. The second is once the other team's starter starts to tire and hitters have seen them a couple times. The third(fourth and fifth) is when the bullpen pitchers come in and you have another chance to hit against a guy who is having a bad day.

If you have a one run lead in the 7th against a team with a crappy bullpen, it is not a high leverage situation just because there are a couple runners in scoring position. Your hitters are going to have a lot of opportunities to extend or recover that lead. Once you get to the ninth inning, your best reliever it putting the finishing touches on a game the rest of his teammates have already won. He is closing the deal.

In short, there is absolutely no evidence that baseball's managers are mishandling their bullpens. Its mostly an interesting invention of writers with more English skills than baseball knowledge. And, like any good fiction writer, they make up the facts to create a good story.

SoCalTwinsfan said...

Another thing to consider is that in the past, teams generally had one closer and then the rest of the bullpen was for failed starters. The addition of the setup men and LOOGYs should improve the bullpen.

Also, in the past when closers were called on to pitch 2+ innings to preserve a lead, what would happen the next night if the team needed to preserve another lead? Would the closer be available for one inning, if at all?

And just because a guy is labeled a closer, doesn't mean he is the bullpen's best reliever. For all the Nathans and Riveras in baseball, there are plenty of Joe Borowskis or Jon Rauchs, guys that have the "guts" to close, but not the best stuff.

Jack Ungerleider said...

A quick question, shouldn't 1981 be removed as well since it was a strike shortened season as well?
(The split season)

A quick check on Wikipedia shows that the 1981 strike resulted in 713 lost games, and the 1994 strike resulted in between 931-948 games including the post season.

Now that looks like a significant difference in the number of games but remember that there were 2 fewer teams in 1981. So that translates to 162 fewer games in the overall season. So if you look at regular season games you are probably dealing in the same sort of time lost. So if the 1994 data is removed than the 1981 data should also be removed.

Also there were two rounds of expansion in the 1990's that probably impacts your win percentage more than the specialization of the bullpens. Which of course begs the question is bullpen specialization a side effect of expansion?

Matthew Brock said...

TT some of your logic is pretty sketchy. First almost all teams have 12 man staffs and more and more are carrying 13. Teams are carry an extra pitcher then they did in the past. And while I dont have evidence that bull pens arent more specialized than they used to be im confident its the case. Starters dont pitch as many innings, teams are using high draft choices to get elite college relievers, teams use lefty righty matchups every game ect. And youre argument against high leverage situation doesnt strike me as well reasoned. You shouldnt worry about giving up a 1 run lead with guys in scoring position and an inferior pitcher on the mound in the 7th because your team will just score more runs? What if you dont score more runs and your best pitcher goes unused? And if your team does score more runs that just makes the 9th inning less important. The goal should be to give up the fewest runs, not to give up the fewest 9th inning runs. John Rauch should be example 1 on average relievers being able to get through an inning without giving up a run most of the time. Why should teams by convention use their best reliever to pitch that inning. If bases are loaded in the 8th with 1 out and my team up 1 run id want my best reliever to try and get out of that inning without giving up the lead. Lots of the time the 9th inning would be an appropriate time to pitch your best reliever but sometimes a situation would dictate that pitcher be used at a more important time. While there might not be a ton of evidence that suggests managers are mismanaging their bullpen, in theory you can make a strong argument they arent managing them optimally.

Its a lot like lineup construction. Ive personally always thought that the way gardy and most managers make their lines ups is not optimal. The construct them based on silly convention and the accumulation of statistics. Why does a fast guy have to be your leadoff hitter? Carlos gomez was fast so he was a leadoff hitter. Why is nick punto batting second? Because hes a slap hitting middle infielder that handles the bat well. The twins used that formula and were successful but that doesnt mean it was the best lineup. I think joe mauer should have been the leadoff hitter because he gets on base 43% of the time and if you stack all your best players after him you are likely to score more runs than you are giving away 2 outs with inferior hitters because a guy is fast and therefore should be a leadoff hitter.

JK said...

John, Tom Tango wrote an interesting article in the 2010 Hardball Times about relievers. Basically, current patterns lead to improved effectiveness and possibly less injuries. The tradeoff is not maximizing the leverage of your "relief ace". Fits nicely into your worse today, better over the long haul idea about current usage patterns.

BeefMaster said...

I'd point out another complicating factor in this analysis - the fact that offense levels in baseball tend to fluctuate. For example, I would expect a 1-run 9th-inning lead in 1998 to generally be much less safe than the same lead in 1988, because there was so much more offense in 1998. Not all 1-run leads are created equal.

Taking that into account might explain some of the 7th and 8th-inning curves on the smoothed graph. The hold percentages in those inning increase steadily beginning in the early/mid-80s as specialization takes hold, then slowly descends as offense increases through the 90s. I was all set to attribute the rise in the 8th-inning numbers to the "post-steroid era", but the shift seems too early for that - it starts going back up in 1997-2001, when offense was just about at its apex.

I don't think that is any sort of thorough explanation of the data, of course, but it might account for some of the general trends.

John H said...

That chart is pretty small but looks like the end results are almost the same as the beginning results. What I get out of this 30 plus chart is that for some years the hitters trend is better and other years the pitchers trend is better. Probably says more about the player's abilities than anything else.

TT said...

Matt -

"First almost all teams have 12 man staffs and more and more are carrying 13."

is almost exactly the same thing I said:

"Now, many have 7 relievers and occasionally 8." with five starters that is 12 or 13 pitchers.

"What if you dont score more runs and your best pitcher goes unused? "

Then he will be available the next day when there really is a "high leverage" situation. Pitchers don't pitch every day and at some point the more they pitch, the less effective they are going to be.

"John Rauch should be example 1 on average relievers being able to get through an inning without giving up a run most of the time."

Rauch is not an average reliever. That is why the Twins traded a pretty good prospect for him. But the same argument applies to a one run lead with runners on base, most good relievers can protect it.

"in theory you can make a strong argument they arent managing them optimally."

Sure, in theory. But you have to have some real evidence, not speculation based on misinformation. And you can't just make up history that fits your argument.

"I think joe mauer should have been the leadoff hitter because he gets on base 43% of the time "

Because you really want your best hitter at the plate with no one on base?

Anonymous said...

"Because you really want your best hitter at the plate with no one on base?"
This is the exact logic that drives me nuts. Carlos gomez and nick punto werent ever getting on base. Batting mauer 3rd is just managing to accumulate RBI which is stupid. Also you control who leads off once. If you had mauer or any other high obp guys leading off all you are doing is ensuring they get the most PAs. And im fine with mauer bating 3rd behind span and hudson because they are both good obp guys but as of last year we were putting out machines like orlando cabrerra, alexi casilla and nick punto in the 2 spot and it hurt our ability to score runs. Teams always bat their best hitters 3 and 4 to accumulate rbi even if they dont have competent hitter hitting 1 and 2. Its bad management and best case scenario it reduces the number of ABs your best players get and worst case it puts likely outs in the middle of your good hitters and makes it much more difficult to score.

Anonymous said...

Evidence that a bullpen with good relievers and a set of specialty relievers doesnt exist because no one has tried it. Matching reliever skill to situation just makes sense. TT you wouldnt demand concrete evidence from something you believed to be sensible and logical. Show me evidence that knowing their role in the BP allows them to pitch better. Its impossible. Should we just discard the idea completely or are we capable of rationally thinking about the idea and deciding if it has merit?

TT said...

Anonymous -

"Carlos gomez and nick punto werent ever getting on base. "

Carlos Gomez and nick punto weren't batting at the top of the order and Punto actually does get on base at above league average.

"If you had mauer or any other high obp guys leading off all you are doing is ensuring they get the most PAs. "

You are also guaranteeing that a quarter of those PA will be with no one on base. In those 162 PA, Mauer is going to get on base about 7 more times than Punto. He's also going to hit about 7 solo leadoff home runs. If he had been batting third, there would have been 4 or 5 runners on base just for those 7 home runs.

I don't see how getting on base 7 more times is going to make up for that. And that doesn't count the impact of any of his other hits, much less the fact that even his walks advance runners ahead of him. Belief that it will is just evidence of an on-base fetish.

"Show me evidence that knowing their role in the BP allows them to pitch better. "

The pitchers and manager both say it does. Show me evidence they are wrong.payelesc

Anonymous said...

"Pitcher are best used in proper leverage situations" Me. Theres your evidence prove me wrong. TT, thats some of the dumbest logic ive ever read. Pitchers saying something is most definitely not concrete evidence, its at best anecdotal. Matthew is right by the way. Lineups are best constructed with all the best hitters clumped together to try maximize the chances of bunching hits together. If joe mauer were to always hit with no one on base and the teams best hitters behind him that would be a fine because joe mauer would be on base a lot giving the high slg hitters many chances to drive him in. Even if puntos obp is average he doesnt belong anywhere near the top of the order because hes a bad hitter. The key is to get high obp guys up in front of better slg guys. Just because mauer is both a high obp and slg guys doesnt mean he has to be a middle of the order guy. He would have been better used in the 2 hole behind span last year instead of filling that spot with punto, cabrerra, casilla, ect.

TT said...

"Pitchers saying something is most definitely not concrete evidence, its at best anecdotal."

I'm sorry, but it is not anecdotal when it is backed experience. Its called observation. And there is no evidence to the contrary.

"if joe mauer were to always hit with no one on base and the teams best hitters behind him "

The Twins best hitters hit behind him whether he is batting first or third in the lineup.

"He would have been better used in the 2 hole behind span last year instead of filling that spot with punto, cabrerra, casilla, ect."

That isn't what the numbers say, nor logic. Having Mauer hit more solo home runs is not a good strategy. And having Mauer bat second and Morneau third would actually reduce both of their opportunities to drive in runs. The guys who would gain opportunities would be mostly the 6th and 7th hitters. They are not the "team's best hitters".

Anonymous said...

TT you are wrong. About everything. Observation not supported with relatively independent data is not concrete evidence. You ask for solid verifiable evidence to refute your claim but present none to support it.

TT said...

"Observation not supported with relatively independent data is not concrete evidence."

Right. So if catchers say they get tired after catching a 12 inning game, there is no reason to believe them if you decide it isn't true?

Anonymous said...

Its certainly not concrete evidence. But youre starting to get to the real shortcoming in your argument. Demanding concrete evidence to show that using your best pitchers in high leverage situations is a more optimum while using anecdotal evidence to support what you believe is unfair. Why is a very logical argument about high leverage pitcher use a creation of the media with no basis in reality because there are no numbers to support it but a less logical method of bullpen management supported by the perspectiveless opinion of a very small sample of pitchers is a slam dunk argument. You dont have any information about how the vast majority of pitchers feel about the topic. And if you did theres still no information on howd theyd respond to prolonged use in a dynamic bullpen.