March 25, 2013

Baseball Therapy

Could the All-Bullpen Approach Actually Work?

by Russell A. Carleton

the archives are now free. All Baseball Prospectus Premium and Fantasy articles more than a year old are now free as a thank you to the entire Internet for making our work possible. Not a subscriber? Get exclusive content like this delivered hot to your inbox every weekday. Click here for more information on Baseball Prospectus subscriptions or use the buttons to the right to subscribe and get instant access to the best baseball content on the web.

Subscribe for $4.95 per month Recurring subscription - cancel anytime. Subscribe for $39.95 for one year a 33% savings over the monthly price! Purchase a $39.95 gift subscription a 33% savings over the monthly price! Already a subscriber? Click here and use the blue login bar to log in.

Baseball games come with built-in subtitles. Dwight Gooden vs. Roger Clemens in the 1986 World Series. Bob Gibson vs. Denny McClain in the 1968 World Series. Kyle Lohse vs. Ross Detwiler in Game 4 of the NLDS last year. It's one thing to see a game between the Yankees and Tigers, but it's an entirely different game if CC Sabathia is pitching against Justin Verlander. And no one ever subtitles the game A-Rod vs. Miguel Cabrera.

Baseball loves its starting pitchers. And given the way the game is currently constructed, that makes sense. The schedule lists the probable starters for each night's games for a reason. It's the one factor that changes night to night for a team, and it has a major impact on the team’s likelihood of success. After all, the starter will hopefully be pitching six or seven innings and be responsible for a good deal of the run prevention picture. In the National League, you even get to watch him bat! It's always been this way, since the days when a complete game was the default. The idea of one man for one job is so deeply ingrained in the game that it's hard to imagine a world in which it’s structured differently. From the little leagues to the big leagues, each team has a starter who will pitch as long as he can, and then some relievers who will finish things off. That is the order of things.

Here's a question: Suppose that baseball strategy were being re-invented from the ground up, with full knowledge of how the game works. Would the starting rotation be re-introduced in its current form? How much does the existence of the "We have five starters who pitch six or seven innings each, and then relievers who take care of the rest" paradigm owe to good logical sense, and how much to the fact that it's always been this way? (More cynically, how much of it is due to the rule that says that a "winning" pitcher must pitch for at least five innings?)

One thing that's always instructive is to look at the language that we, as a culture, use to describe things. In modern baseball, there's a word for a guy who is conditioned to throw 6-7 innings every fifth day ("starter"), and there's a word for the guy who throws an inning on three days out of a week ("reliever"). What do you call a pitcher who is purposefully tasked with throwing three innings a couple times a week?

What if we tried something a little different? Well, I wouldn't be the first to propose such a system. I wouldn't even be the first to implement it. In 1993, Tony La Russa briefly (as in, "for one week") tried a system in which he pulled together three groups of three pitchers each. Each man was tasked with getting through 2-3 innings, and there were four more "traditional" relievers to pick up the slack. The A's went 1-5 that week and went back to a more standard rotation afterward. There has been talk of tandem starters in other places before, and the idea has even been floated on Effectively Wild.

I propose to really dig deep into the numbers here to see if this is a feasible method. Is there an alternative to the traditional five-man/6-7 innings rotation that at least makes sense on paper? More than that, would it work in reality (and yes, there's a difference)? Let's find out.

What Does a Starter Do?
The 5-6-7 rotation does have its advantages. It's an efficient way to fill innings. By using only five roster spots, a team can cover roughly 2/3 of the innings that it needs to fill. This is the somewhat-hidden reason why "innings eaters" (pitchers who will throw 200 mediocre innings in a season) seem to have such great value. The results might not be great, but someone has to pitch those innings, and it's easier if that person occupies just one roster spot. Leaving aside for a moment any issues of performance, we'll need to make the case that we can deliver the bulk innings that would be lost from eliminating starting pitchers.

Let's look at how deep into games starters actually went in 2012:

Recorded 0-15 outs	30.6%
Recorded 16-18 outs	30.4%
Recorded 19-21 outs	27.0%
Recorded 22-24 outs	9.2%
Recorded 25 or more outs	2.8%

The mean number of outs recorded is 17.4, with a standard deviation of 4.3. The median number of outs recorded is 18, which is six innings, so on average, we need to replace 5-7 innings per night. That sounds about right. An interesting factoid: In 2012, more starts lasted four innings or fewer (12.7%) than lasted past the seventh inning (12.0%). You're slightly more likely to see a bullpen-taxing implosion than a gem in modern baseball.

Someone Does Have to Pitch...
My proposal isn't exactly new: tandem starters. Six men, split into three pairs. Our brave and innovative team uses a three-pair rotation. Each pitcher in the pair is expected to throw 50 pitches, gets two days off, and then comes back and do it again. On the downside, it does require one extra roster spot more than the 5-6-7 rotation, but at least in theory, it covers the 100 pitches that the starter usually throws.

Let's look at how many outs 50 pitches might "buy" a starter. In 2012, among those starters who made it to 50 pitches, by their 50th pitch they had:

Recorded 0-3 outs	0.5%
Recorded 4-6 outs	13.3%
Recorded 7-9 outs	49.2%
Recorded 10-12 outs	32.3%
Recorded 13 or more outs	4.7%

Starters failed to record the 18th out in 42.8 percent of their outings in 2012. They failed to record the ninth out within 50 pitches 44.6 percent of the time. It's not unreasonable to expect that roughly 50 pitches will buy a pitcher three innings a good amount of the time, and it will fail to do so at roughly the same rate that 100 pitches fails to buy a starter more than six innings. I also looked at how many pitches it took to record nine outs among starters in 2012. The mean and median were both 50, with a standard deviation of 10.

We can likely cover the bulk innings needed in line with what starters currently generate, but at the cost of an extra roster spot.

There's one other thing to account for: Pitchers vary greatly from one outing to the next, particularly starters. There are nights when a guy just doesn't have it, and he hits the showers after 2 1/3 innings. In that case, the bullpen is on the hook for filling in the rest of the game. There are also nights when a guy is lights out through eight and gives the bullpen a break. Because we're limiting our pitchers to 50 pitches each, if they're really on that day, they won't have as much chance to eat up a few extra outs, and their replacement is most likely to be his usual average self, meaning that it's less likely that you get the benefit of a long outing. On the flip side, if the first pitcher comes out and lasts only 1 1/3, his replacement is likely to be at his usual average and could go three innings, so it saves the bullpen from having to pick up as much of a workload. The tandem approach is a low-risk, low-reward strategy from the perspective of filling innings.

This isn't unimportant. The way the current major-league roster compensates for a starter flaming out is to have a guy designated as the "long reliever." He's usually an expendable arm who's trying to prove that he's not expendable, and usually it's 9-2 or so at the point that he comes in, so it doesn't really matter who he is or what he does as long as he gets the team through the sixth with his arm attached. Teams also use the fact that, if the long man has to go four innings and leaves the bullpen short the next night, he can be sent to Triple-A and some other fresh expendable arm can be called up to take his place. The ability to arbitrage roster shuffling in this manner reinforces a role that fits well to handle one of the shortcomings of the 5-6-7 rotation. Because teams can get another long reliever on short notice, they can "afford" to use a strategy that is higher risk in terms of filling innings. It's interesting to think that one of the reasons that the five-man rotation exists is because it's easy to find bad pitchers. Using the tandem starter approach doesn't really leverage that possibility.

If we figure that the job of the starting tandem is to get through the sixth inning, there will be some nights where they fail to do that, in which case some sort of bridge reliever will be necessary. On some nights, one member might have to be pulled early, or maybe both men have to fight through some jams and their combined allotment of 100 pitches is used by the fifth. It happens. But since we have a lower risk of an early-inning flameout, it's a little more likely to be deeper into the game, and not due to a collapse, meaning not as much length is needed and the game might still be in the balance.

Still, I don't think we could dispense with the traditional long man. One member of the tandem might go his three innings, but his partner might have nothing, and suddenly it's the fourth and someone needs to go out there. Our team will likely have to carry a pitcher who can fill the long man role (so, no roster spot savings there), but also have enough arms in the bullpen that they can trust in high-leverage situations to cover some extra frames beyond what they do now. (Or it can find a long man who can be trusted in tough situations.) Suddenly, we've created a cascade effect on the rest of the roster.

Names Please?
Our pitching staff can't consist of abstractions. There have to be real bodies inside those uniforms. I've created a theoretical role in which a pitcher throws three innings at regular intervals, and frankly, there's no one in the game of baseball who has that job description right now. About the closest thing in recent memory was the role that Mariano Rivera had in 1996, prior to becoming a closer. He would often throw 2-3 innings to bridge the gap from starter to John Wetteland.

However, I think that there are players who might just be good for the role, and to make things better, they might represent a glaring inefficiency. There are certain starters who seem to have enough "stuff" to get through the lineup once, but who struggle with making it through the lineup the second and third times around. Usually, they're pretty tantalizing guys. They clearly have something, and once in a while they have a nice seven-inning, two-run outing, but why is it that most days they can't sustain the magic? What if we just didn't ask them to?

I looked at all starters in 2012, specifically what they did in the first three innings and then everything after that. I calculated each pitcher's xFIP for both time periods. I then filtered the list a bit. I looked for those who had an xFIP in the first three innings of 3.75 or less. A seasonal 3.75 xFIP would put the pitcher in the top 30 xFIPs among qualifying pitchers in 2012 (Wade Miley had a 3.75), and an ERA of 3.75 would be good enough for no.41 on the list between Bronson Arroyo and Wandy Rodriguez. Not bad.

Then, to figure out which guys might have some hidden value, I sorted the list by the difference between xFIP in the first three innings and in subsequent frames. Their overall ERAs would be higher, but it might be masking the problem that they might be pretty good prior to the fourth. After all, guys like CC Sabathia and R.A. Dickey had good opening frames last year, and both were somewhat worse in the subsequent innings. Neither is available cheaply. But, if we can find a guy who has the tag of "bad starter," but can put him (and a friend) into a situation where their talents can shine in a new role, the stinky parts can be hidden.

Here’s the list, sorted by the guys who would have the most to gain. The numbers given are their xFIPs in the first three innings, followed by subsequent innings. There was a minimum of 100 PA in each bin.

1. Carl Pavano (3.62, 5.24)
2. Jake Arrieta (3.03, 4.46)
3. Tommy Hunter (3.63, 5.01)
4. Colby Lewis (2.96, 4.33)
5. Justin Germano (3.61, 4.94)
6. Drew Hutchison (3.43, 4.73)
7. Brandon Beachy (3.31, 4.60)
8. Bud Norris (3.37, 4.56)
9. Hisashi Iwakuma (2.86, 3.99)
10. Charlie Morton (3.64, 4.71)
11. Ryan Vogelsong (3.41, 4.41)
12. Vance Worley (3.47, 4.44)
13. Jered Weaver (3.60, 4.49)
14. Joe Kelly (3.68, 4.54)
15. Mike Minor (3.43, 4.23)

This is a pretty good list of guys who aren't highly valued. It's not that they could be had for a song, but it's not like other teams are holding on for dear life to most of these guys. A team that was committed to this approach might pick a few of these pitchers up, and probably not for a lot of money. And the fact that they exist at the major-league level makes one wonder how many of their brethren are in the minors, toiling away as "failed starters" who might thrive in a role like this, except for the fact that one does not really exist for them. (Note: I also looked to identify relievers who might be extended beyond one inning, and would thrive in a role where they pitched three instead. They might be out there, but while starters, even bad ones, go well beyond three innings, there were very few relief appearances that lasted three innings in 2012. In fact, only 20 percent of relief appearances in 2012 involved the pitcher getting a fourth out.)

If the trend held, a pitching staff composed of these guys, who had to go only three innings each time they went out, could deliver a combined performance consistent with the top or second starters on most teams, and could do so every night (though only for about six innings, give or take, rather than the seven or eight innings we normally associate with an "ace" outing). Consider what that sort of performance would cost a team on the free agent market, and compare it to what a pitching staff composed of the list above might cost. The savings might be useful in upgrading other parts of the team.

Aren't You Making Some Major Assumptions Here...
We probably need to talk about whether 50 pitches, two days off is a workable model from a physical standpoint. When the Colorado Rockies went to a model of 75 pitches, three days off last year, it smacked of someone who said "Well, we usually do 100 pitches, then four days off, so that's 25 pitches per day..." It's not clear that it's nice and linear like that. BP's Ken Funck, when he looked at the issue a few years ago, found that there was little consensus on the matter because the idea hadn't really been tried on any sort of large scale.

At the MLB level, there are almost no examples of pitchers being used in a 50 pitches, two days off system. In fact, there aren't even that many individual case events. Starters who throw only around 50 pitches are usually either hurt or ineffective. If they're hurt, they might make their next appearance after 15 days on the DL. If they're ineffective, they likely just take their next turn in the rotation. Relievers who throw that many pitches are usually pitching in long relief, and then next pitch when they are needed. No one does this on a regular basis, and it's not clear whether this would wear pitchers down or whether it would be a wonderfully revolutionary idea.

It seems like 50 pitches followed by two days off could work, at least in that it passes the smell test. If anything, this would require a complete re-conceptualization of how pitchers train and structure their workloads, and maybe it wouldn't work unless you've been pitching that way since the age of 12. And maybe the changes would affect the pitchers and make them worse (or better.). The reality is that most pitchers train along one of two tracks, and so if a team wanted to implement this tandem model, it couldn't count on finding guys who were cast off from other organizations but already had trained in the same basic role the way that they can find six-inning starters. That makes searching for reinforcements a lot harder. Basically, a team would have to produce all of its pitchers internally, or it would have to be shown that the conversion to the “50 and two” model could be quick and easy. Otherwise, there just wouldn't be enough pitchers out there to make things work.

Possible Extra Benefits

• For one, the tandem starters wouldn't have to pace themselves as much, nor worry as much about facing a hitter a third time, or in some cases, even a second time. In the same way that starters who become relievers pick up a little extra mustard, our tandem starters might see similar effects.
   
• If the team can pair a left-handed pitcher with a right-handed pitcher, it does present some interesting platoon issues for its opponent. Does the other manager start his "We're facing a lefty today" lineup or his “versus a righty” lineup? And when the other half of the tandem comes in, does he pinch hit? It could create a few extra platoon advantages for the pitching team. In 2012, the platoon advantage was worth 20 points of OBP for right-handed hitters and 38 points of OBP for left-handed hitters. Converting one walk into an out is worth half a run, and 20 points of OBP is a two percent swing. For every extra 50 platoon advantages you gain, you gain at least half a run. Small effect, but it can add up.
   
• In the National League, when it's time for the first half of the tandem to come out, you might as well pinch hit for him, whereas you don't have that luxury with the traditional starter. Since pitchers hit .129/.162/.166 as a whole, even sending up the worst qualifying OBP from 2012 (Drew Stubbs, who hit .213/.277/.333) represents an increase of more than 10 percent in the chances that you won't make an out. If the NL team is able to use this effect 60 times a year, and assuming that we're talking only about changing an out into a walk, it could leverage an extra three runs. That's assuming the worst-possible pinch hitter and that all you do is add a few extra walks (rather than maybe a home run?). With a decent pinch hitter, whom they would have incentive to sign since they'd have plenty of chances to use him, the effect might be enough to be worth a win or two.

Possible Problems

• That extra roster spot is going to have to come from somewhere. In all this I'm assuming that back-end bullpen usage will not change. Since most teams now employ a 12-man pitching staff, this would take them from seven relievers to six. Or to a 13-man staff, leaving only three bench players in the AL. Or relievers will have to become more comfortable with slightly longer outings. Something's gotta give. Teams have gotten by on short benches before (but suddenly, your bench options are limited only to gifted multi-positionalists). Maybe it's just a matter of a different model of reliever usage.
   
• In the event of a doubleheader in which a team would normally have to call someone up to make a spot start, the team might be faced with a situation in which all of its reserves are tandem starters too, and would thus have to make two moves to cover the games. Or, they'd have to keep a traditional starter in the minors for such instances so that they could get by with moving one body up and down. Odd that even going to a tandem approach necessitates keeping a 6-7 inning starter around.

Would Anyone Willingly Sign Up for This?
Then there's the biggest problem of all. Unless baseball changed the way in which pitching wins are awarded (or got rid of wins altogether), this one might be a tough sell to free agents. The first guy out would likely not get a win, but he could get tagged with the loss. Like it or not (and I am definitely in the "not" camp), won-loss record still means a lot in baseball. Even among potential draftees and international signings, a team couldn't keep secret the fact that they were switching to such a plan. Would pitchers steer away from our experimental team (as much as they were able) if they knew that they'd be pigeonholed into this three-inning role?

Then there's the ego bruise that would come from being told, "Well, you're not good enough to be a real starter..." Finding actual human beings to play in this system would be a challenge, and it's not like minor-league systems are churning out guys ready to step into this role. In fact, the primary thing that the tandem model has going for it is that it would be staffed by guys who were chosen because they weren't good enough to be real starters. Often, we hear of guys who take lesser contracts "because it gives me the chance to start." Maybe that's stupid male pride, but while the system might make sense on paper, baseball is still played by humans. Maybe if guys had grown up throwing in this role and idolizing the great starter tandems of all time, there might be a line of guys who would be willing to take the role. But that's not the way it works.

In Sum...
If we make an assumption that switching pitchers to a 50 pitches, two days of rest schedule would actually work and not affect a pitcher's abilities (or at least not make them worse), there's evidence that there's a hidden pool of talent out there that are floundering in the current system, but who might thrive in a new role. Given enough bargain hunting, a team could probably find a group of pitchers who could produce value equivalent to having a starting rotation filled with guys from the top two spots in most team's rotations and still provide the bulk of innings that starters currently produce. The production would probably come at bargain-level prices. Those cost savings could used be for upgrades elsewhere on the team, and there might be a few other benefits that they could squeeze out.

However, #ThereIsNoUnicorn. The types of pitchers that are called for in this system are not naturally provided by the baseball talent training system, and it would be hard to recruit people to the cause. Plus, there would be roster repercussions all across the team, from relievers who would need to work longer to bench players who might have to be a little more multi-talented. It's not impossible, but it's a lot tougher than just saying "Okay, this is how we're going to operate from here on out."

In theory, this would work great and would probably net a team some really nice value. But, as my father is fond of saying, "Everything works in theory." The real-world structural obstacles that would get in the way of converting to this type of program seem like they would be too great to overcome. It would certainly take overhauling an organizational philosophy, and that's hard to do. Maybe the game will evolve a little and the ingredients that would be needed for this to work will be more readily available. Maybe there's a team that could luck into exactly the right roster and give it a whirl. Maybe this works great in video games where you can do a fantasy draft with exactly the right players, and players who don't complain and just trot in and out of the game when you hit the right button. Just don't expect to see it in reality any time soon.

Russell A. Carleton is an author of Baseball Prospectus. Follow @pizzacutter4
Click here to see Russell's other articles. You can contact Russell by clicking here