In 2013, the average starting pitcher lasted a bit under six innings (5.89, to be exact), and starters as a whole had an ERA of 4.01 (and an earned-or-not run average of 4.17). Collectively, the guys who began the game posted an xFIP of 3.91. They did so by using what has become a standard rotation setup. Five starters take their turns in sequence and throw roughly 100 pitches in an attempt to make it through six or seven innings (the 5-6-7 model.)

Over the past few weeks, I’ve been examining the history of how we came to this particular (and peculiar) setup, and in doing so have found some interesting things about the evolution in starting rotation structure, including some things that don’t appear to be backed up by any evidence. It’s one thing to say that there’s a problem with the way things are. As my uncle (who was an engineer) was fond of saying, “Any idiot can make noise. It takes an engineer to make it quiet.” So, a question: Knowing what we know, how can we build a better starting rotation?

A few rules. First of all, we have to use parts that are already on the shelf or that can be bought at the local Radio Shack. “Sign Clayton Kershaw, Felix Hernandez, Justin Verlander, Yu Darvish, and Dan Brooks” would make a great rotation, but we are not working in science fiction or sports-talk radio (same thing). No suggesting players who don’t actually exist right now, and whatever plan is out there, the plan to staff it needs to be realistic. Second, we have to take into account real issues like how the new proposal would affect roster construction, salary concerns, and the need to have spare parts available. Third, it has to work both on the game level and the season level. We need to make sure that innings are covered, that the plan won’t overtake the whole payroll, that the plan could be sustained for six months, and that we’re not changing for the sake of changing. It actually needs to make sense.

Let’s look at a few options, shall we?

Option #1: The Status Quo (The 5-6-7 model)
Advantages: Well, there would be no transition problems, and we already know how well it works. I’ve previously shown that one major advantage of using the traditional 5-6-7 system is that it provides bulk innings and conserves roster spaces. The entire baseball infrastructure is set up in service of this model, and you don’t have to put up with those pesky critics wondering why you’re doing something crazy.

Disadvantages: Maybe the biggest disadvantage of the 5-6-7 model is the fifth starter. On a recent episode of Effectively Wild, Sam Miller made the suggestion that maybe the next big thing in baseball is not the all-bullpen approach, but the understanding that not all pitchers are made to go six or seven innings, and while some are, there aren’t really enough to go around who are good at it. Instead of trying to force five guys into a role for which they are not qualified, why not think of ways to integrate the Clayton Kershaws of the world with some alternate method for covering innings? Let’s first look at whether there are pitchers who just should not be allowed to start.

I looked at all pitchers who started at least 10 games (there were 185 such pitchers in 2013) and how well they did at going deep into games. I totaled the percentage of starts each pitcher made in which he made it to at least the fifth inning. On that list, Aaron Harang is 135th (“the median fifth starter”?) with an 85 percent success rate (89 percent is the league average). The 135th pitcher on the “getting into the sixth inning” list is Dan Haren. Haren made it into the sixth inning two thirds of the time (60 percent vs. 71 percent for the league). The 135th guy on the “made it into the seventh inning” list was Bruce Chen (28 percent vs. 40 percent for the league). If we define a quality start as going six or more innings and giving up three runs or fewer, 135th place belongs to Tom Milone at 42 percent (vs. 52 percent for the league).

Those differences in percentages may not seem like much, but they have a pretty big effect. Teams win 58 percent of their games when the starter pitches into the sixth vs. 29 when he does not. For getting into the seventh inning, the spread is 65 to 39 percent. If you give your average fifth starter 30 spins on the wheel, and in 10 percent fewer of those (three starts) he won’t meet one of these milestones, that negates roughly one win’s worth of probability, compared to the league average. Maybe we can find a way to do better than that.

Option #2: The Four-Man Rotation
In theory, there was never a good reason for teams to have abandoned the idea of four men in a rotation; they could have abandoned only the extreme pitch counts that starters had to endure during those decades. But abandon it they did, and we’re still stuck with the consequences.

Advantages: As Earl Weaver pointed out, it’s easier to find four good starters than five.

Disadvantages: The evidence suggests that, while pitching on three days’ rest on a regular basis had little effect on injuries when this practice was common, this is no longer the case. Evidence does suggest that the results of pitching on three days’ rest are no better or worse than we would otherwise expect of a pitcher now, but that the injury risk is so large that it makes the idea impractical.

The Rockies did try a four-man rotation for a while, with a 75 pitch limit per starter. Because of the small sample size, it’s hard to figure out whether this “worked” or not, or whether the 75 pitch limit was a good idea. We know that in a more standard five-man rotation, there is little pitch count hangover from start to start in performance until you get past 115 or so. We also know that historically, there is little evidence to suggest that the previous outing’s pitch count affected a pitcher differently based on whether the start was made on three or four days’ rest. Letting a pitcher throw 100 pitches every four days will increase his overall pitch totals, which will in turn make him a more likely candidate for an injury, so maybe keeping pitchers down to 75 had more to do with keeping them off the DL.

Verdict: It ain’t coming back. Maybe it should, but now, everything in baseball runs on five beats instead of four. A team implementing the four-man rotation would have no place on the free agent market to shop for spare parts and would face an added injury risk that would likely wipe out the gains they would reap from not having to put up with a fifth starter.

Option #3: The Four-and-a-Half Man Rotation
In this model, teams assemble five starters, but the fifth guy pitches only when his team plays five games in a row. If there’s an off-day, the fifth starter gets bumped, and the first four guys pitch every fifth day regardless. Teams sometimes implement a version of this plan in April when arms are fresh and off-days are more plentiful. You’ll often hear that a team doesn’t “need” a fifth starter until such-and-such date, and so the fifth starter will pitch out of the bullpen, or the team will carry another short-inning/situational reliever or position player and stash the fifth starter at Triple-A for a few weeks.

Advantages: In 2013, 52 percent of starts were made on exactly five days of rest. 38 percent were made on either six or seven days of rest. That’s roughly 60 starts per team on “extended” rest. If we assume that there were likely cases where an off-day pushed the entire five-man rotation back (so five cases of pitching on extended rest) we’re talking about 10-12 times per year where the fifth starter could be skipped, and 2-3 more starts could be given to each of the top four starters. Again, if we replace those starts with league average, it might end up being worth a third of a win.

But there’s one thing we need to clear up first. When we speak of the five-man rotation, we assume that pitchers throw every fifth day. We see above that this is not always the case. Perhaps having that fifth day of rest once in a while helps keeps arms healthy? I went back to the Cox regression method I used previously and counted up how many times over the course of a season a pitcher had started on five or more days of rest. Using data from 2000-2012, I looked for evidence that having these extra bonus days offered some sort of protection against injuries. Turns out that the answer is no. If anything, it actually makes pitchers more likely to suffer an injury. (Note: that could just be picking up on the fact that when pitchers are held back more often, it’s because a coach noticed something that would make them more prone to injury, and the team decides to be more careful.)

There are also some ancillary benefits to having an extra roster spot to play with on occasion. I’ve previously found that the ability to construct a good platoon or to have a better replacement around for injuries or to keep more players fresh is worth a couple runs over the course of a season. It wouldn’t be much, but it wouldn’t be nothing. Also, the fifth starter spot could be something of a transitional spot for developing pitchers. Young pitchers in this spot would get some starts, but could also get their feet wet in a long-relief/mop up role, and could bounce a bit between Triple-A and the majors as the needs of the team dictated. If a young pitcher was just that good, he could simply be converted into a full-time starter and the team could go with a standard five-man rotation, or another struggling starter could be given a chance to rest a bit.

Disadvantages: There don’t seem to be a lot. The biggest one seems to be sociological. The five-man rotation comes with a bit of unspoken assumption that “You are all equal in our eyes.” Elsewise, why would teams use all five spots in rotation when they don’t really need to?

This plan would be formalizing a role that could be seen as one down from a “real” starter. Teams might have some trouble finding people to sign up for it. The advantages of being able to send the fifth starter back and forth to Triple-A suggest that a young pitcher might be best suited for it, and teams may want to fully immerse their young pitchers in a full-time rotation spot. But even a veteran who is hoping to hang on to his MLB career might be happy simply to have an MLB uniform to pull on.

Verdict: Teams probably should skip their fifth starter more often. It won’t make them world-beaters, but it’s a more effective strategy that comes at minimal cost and with low barriers to implementation.

Option #4: Turning the Fifth Starter into a Tandem
We’ve now entered the world of tandem starters. The idea of a tandem starter, for those not familiar, is that if we expect a traditional starter to throw 100 pitches, we might instead spread that workload across two pitchers who would throw 50 pitches each. In my previous work, I found that there were a group of starting pitchers who were much better in their first 50 pitches than they were in their second half-century (at least they were in 2012). The idea was that the market was undervaluing these men because they were “failed starters” rather than seeing them as men who might excel in a different role, if we bothered to create that role.

I ran the same analyses for 2013, looking for pitchers who had an xFIP of 4.25 or lower in batters whom they faced when their pitch count was south of 50. (The 135th-best xFIP—which belonged to R.A. Dickey—among starters with 50 IP last year was 4.23.) I also looked for those who had an xFIP that was at least a quarter of a run higher in the second half of the start, so that they might be perceived as damaged goods. The list yielded 17 names (Justin Grimm, Johnny Cueto, Chris Tillman, Eric Stults, Alexi Ogando, Jeff Locke, Mike Pelfrey, Tommy Hanson, Hector Santiago, Jordan Lyles, Ubaldo Jimenez, Anibal Sanchez, Brendan Morrow, Tyler Chatwood, Dan Haren, Scott Kazmir, and Mark Buehrle). Perhaps telling, there was no little overlap between the 2012 and 2013 lists.

When I originally ran this list, it was meant as a proof-of-concept. There are pitchers out there who fit the Platonic ideal of where we might look for a tandem starter, who are good for three innings but not for six. I hypothesized that there might be pitchers in Triple-A who would also do well in this role, but in a world where pitchers throw either six innings or one inning, and there isn’t a name for the three-inning pitcher, they fall through the cracks and could perhaps be obtained for a small investment. The 2013 list has the names of some fading veterans and guys “who should be better” on it, but given the ever-popular shortage of starting pitching, they are pitchers who have or will get multi-million dollar deals. A team might be able to identify pitchers in their own farm system (or others, whom they would have to quietly acquire) to staff such a tandem spot.

Advantages: IF… IF… IF… a team could find two pitchers who would be good in 50-pitch spurts, who would agree to the role, and who could bring that spot up to league-average performance, it could likely cover the innings that a fifth starter would commonly throw, and would be worth a little shy of one win over the course of a season. (If you take 180 innings and convert them from a 4.23 to a 3.91—league-average xFIP—that’s worth 6.4 runs per year. You could make the case that the tandem starters could go a little harder with each pitch, since they don’t have to pace themselves for 100 pitches, and that they wouldn’t suffer as much from a times through the order penalty.

One minor benefit is that teams in the National League would have the ability to squeeze out one extra plate appearance from a “real” hitter, rather than a pitcher. In the fourth inning, when the first half of the tandem is about to exit, you might as well hit for him. In the 1,215 games played under National League rules in 2013, a team had its pitcher come up in either the bottom of the third or top of the fourth (after he had finished his third inning) about 46 percent of the time. Call it 15 fewer pitcher plate appearances, which would probably be worth around a run to a team.

Disadvantages: It shortens the roster. You now have two pitchers doing the job that one used to do. True, the tandem starters could probably do some relief work in between “starts,” and so could provide value that way, and this strategy could be combined with the “skip the fifth” strategy above. However, that extra spot is worth something like half a win, so in the end, maybe you eke out a net profit of a couple of runs.

Verdict: It’s not that this couldn’t be properly executed. It’s that many of the guys who would be well-suited to the role of tandem starter will probably get better offers from teams and the chance to be a “real starter.” So, you either have to go with pitchers who are probably less desirable (which cuts into the net profit) or more expensive. Or you have to grow your own. The Astros did use a tandem starter approach in the minors last year (see below)—not to develop tandem starters, but to get a better look at more players. However, if they end up discovering that some guys are better in three-inning bursts, then bully for them.

It seems that using a tandem strategy could work, but the obstacles to actually doing it are bigger than we might think, and the pay-off isn’t amazingly huge. Still, it brings up an interesting point about baseball not having a good name for a guy who can turn a lineup over (and do it well) once, but not twice or thrice. Maybe teams are neglecting a resource that they already have. A three-inning guy could have some good uses, even if he’s just a bullpen guy.

Option #5: The All-Tandem/All-Bullpen Approach
I covered this one at length last year. The idea is that a team would recruit six pitchers, pair them up, and run tandem starters all the time. Pitchers would throw every third day and each would throw 50 pitches. The Astros tried a wrinkle on this model in their minor league system, gathering four pairs of pitchers, and having them alternate throwing 75 and 50 pitches per outing.

Advantages: Well, in theory, the math works great in that there are theoretically a group of pitchers out there who would rock this arrangement. But are there enough of them to bring in six of them that can do better than league average? Teams could probably gain some additional advantage by playing some platoon games. (Pair a lefty with a righty… now what does the other manager do?) National League teams would get the chance to use an extra pinch hitter about 80 times during a year. That’s worth something around a win there unto itself, maybe more depending on the strength of the pinch hitter.

Disadvantages: One major problem with the 50-50 timeshare is that it’s a lower-risk, lower-reward strategy as far as covering innings. You’re less likely to have your “starter(s)” go 2 1/3, but less likely to have them go seven innings as well, and if a single starter goes 2 1/3, you can simply bring in a long reliever and then demote him later that night. Having six “starters” also eats up an extra roster spot, and the cascade effects would probably require that teams have a couple of back-end relievers who could go two innings regularly, and not do as much matching up for platoon splits. Two-inning relievers are something of a rare breed in 2014. The Astros’ 75-50 model covers more innings (good thing, because the “starters” eat up eight roster spots!), but in order for the program to work, you need to squeeze all of those pitches out (the bullpen will be small), even if it’s the eighth inning and you’d rather bring in a matchup guy. You lose the ability to do the one thing that the modern bullpen is good at, and that’s allocate more of the good relievers’ time into high-leverage situations, and to use the platoon advantage liberally.

Additionally, there’s very little agreement on how a pitcher would react to a regular schedule of 50 pitches, then two (or three) days off. Teams and pitchers who tried to do something like this would be in uncharted waters. On top of that, it isn’t likely that one team could gather the 6-8 pitchers who would be needed to make this work and make it work better than the league average. And if your minor league system just happened to develop the next Justin Verlander, you wouldn’t want to make him a three-inning pitcher.

Verdict: On paper, where you control everything, this works great. In the real world, there are too many unknowns and too many things that need to go exactly right for this to be effective, and it requires changing over the organization’s entire structure for pitcher development. There’s probably something to be gained from the exercise in thinking about pitchers’ roles, but little more.

Option #6: The hybrid real starter/tandem starter model
There are two of these models. One would be a team having three regular starters (1, 2, 3), and two guys who formed a tandem (4, 5), and making them rotate in the form of 1, 2, 4/5, 3, 5/4. This model was first proposed by Tom Tango in The Book.

Another model is a team with two regular starters (1, 2) and then two tandem pairs (3/4 and 5/6). The starters rotate 1, 3/4, 2, 5/6, 3/4, 1, 5/6, 2, 3/4, 5/6.

Advantages: The team doesn’t need to find nearly as many tandem partners (two in the first scenario, four in the second), and if you happen to have Justin Verlander on your team, he can pitch as a normal starter.

Maybe the biggest endorsement that this system has going for it is the fact that teams seem to use it a lot in the playoffs. Of course, in the playoffs, the point is that you don’t need to worry about tomorrow, or three months from now. There’s also the National League advantage of being able to pinch hit more often, something that, again, could be worth a win or so.

Disadvantages: It’s one thing to try to replace the performance of a fifth starter. But a third or fourth starter? The 105th-best xFIP among starters in 2013 belonged to Todd Redmond at 4.07. Suddenly, the list of guys who would be better than that, willing to go into the role, and cheap enough to make it worth the while starts to dwindle.

Verdict: It would take a lot of courage and just the right roster to make a plan like this work. Because a starter must go five innings to get a win, and because pitchers, like it or not, still think in terms of wins, staffing this might be harder than one might think. It would probably be done only by a team that looked up one day and realized that all the right pieces had accidentally fallen into its lap.

So What’s the Best?
It turns out that the five-man rotation, much like democracy, is the worst system in the world, except for all the other ones that we might try. The four-man rotation (sans massive pitch counts) is philosophically better, but we’ve reached a point of no return on that one. Systems involving tandem starters require exactly the right roster, and rely on types of pitchers who may not exist in enough abundance for even one team to make it work. Mathematically, you could make the case for some of these systems being better if everything is just so, but it’s rare that everything is just so. The gains that they promise might not be worth the cost of re-tooling an entire developmental system and the risk that things won’t work out perfectly. Maybe the biggest thing that the five-man rotation has going for it is that the spare and upgrade parts are plenty, and with five open spots, there’s some room to be modular. That’s a nice way of saying the five-man rotation works because everyone else is doing it.

Still, I think we’ve learned a few interesting things along the way. Teams should skip over their fifth starter more often. There might be individual games where using two guys, rather than one, makes sense. Teams might benefit from trying their failed starters as bridge/multi-inning relievers, rather than just skipping to the one-inning reliever. But for those of you who were hoping that I’d suggest a new era for starting rotations, my apologies. The five-man rotation may not be pretty, but it’s the rational product of its history. And it’s the best that we’re going to get.

Thank you for reading

This is a free article. If you enjoyed it, consider subscribing to Baseball Prospectus. Subscriptions support ongoing public baseball research and analysis in an increasingly proprietary environment.

Subscribe now
You need to be logged in to comment. Login or Subscribe
Have loved this series/streak, Russell. A lot of these things, I knew, because I share your nerdy obsession with this particular facet of roster construction and deployment, but I still love the work-through.

Tom House, who knows his biomechanics, used to say that you needed five in the rotation because the arm absolutely must have 72 hours of rest between starts. I found that hilarious, since even on three day's rest, a pitcher has that, but that's what he said.

I have a wrinkle to offer, a model I've tinkered with over the last six months or so. In it, five starters would still rotate, although it would need to be every fifth day, as you say, not every fifth game. They could each even pitch a vbery little bit less. The twist is, they wouldn't start. Instead, there would be a three-man "starting" rotation of short-burst relievers, the second- through fourth-best of that breed in your organization. They would pitch an inning, two maximum, then give way to the no-longer-a-starter, who would not only be three-to-six outs closer to the end of the game, but (for TTOP purposes) would be facing an opposing lineup whose best hitters batted last. It'd make it tougher to platoon against you. It'd turn the progression of the game more in your favor. And if you did it right, you could probably get by with 11 pitchers on the staff, a coup these days.

Starters would hate it, of course. They'd be unable to accumulate wins, they wouldn't know exactly the time at which they'd begin their outing and they'd be less elevated above relievers. But screw the starters, I say. If you can afford to, anyway.

One last thing: My perception is that the last year or two has actually seen the reemergence of more varied bullpen roles, a first step toward using pitchers more to their exact skill set and optimal workload. Craig Stammen is a minor example; Josh Collmenter is a major one. And whether he did it to protect a young arm, or because he had 119 relievers stuffed into his bullpen in September, or in a nod to a specific player's abilities, Terry Francona made excellent use of Danny Salazar.
To be fair to Tom House, since games last for 3+ hours, needing a full 72 hours of rest would make it necessary for 4 off days assuming evening starts. The clock starts when a pitcher leaves the game, not starts it.

I have no idea if this was actually what he meant (so a starter could start on 3 days rest if the first start was a matinee and the next was an evening one) but it is plausible.
"Still, it brings up an interesting point about baseball not having a good name for a guy who can turn a lineup over (and do it well) once, but not twice or thrice."

I believe that used to be called a relief pitcher. /rimshot

Thanks for these articles, they're great.
"Those differences in percentages may not seem like much, but they have a pretty big effect. [Data]. If you give your average fifth starter 30 spins on the wheel, and in 10 percent fewer of those (three starts) he won’t meet one of these milestones, that negates roughly one win’s worth of probability"

Do you really think the causality implied by this statement exists? That meeting the milestone causes the odds of winning? Maybe I'm misreading the text, but I'd argue that the drop in winning% is due to the pitcher not performing well in the early innings [resulting in him getting pulled earlier], rather than just him not making it to a certain threshold. But maybe I'm being too much of a literalist with what you wrote.
You are correct in your interpretation. I interpreted those data in roughly the same way in my head. Pitching into the sixth usually means he's doing well enough not to have been taken out. The fact that he made the sixth is a quick marker. That's all I intended for it to be.
Currently, most rosters have five starters and seven guys in the bullpen. How disruptive would it be to have four full time starters, two tandem starters and six guys in the bullpen, especially when a tandem guy could once in a while act like a long man out of the bullpen.

Seems like the tandem approach should work. You would need a strong manager / GM to implement it, since it's not the way we've always done it.

The problem is that if you're going to use a tandem 5th starter (model 4) you have to settle on what day the two men would pitch and what their rest schedule would be. So, on some days, you'd effectively have a 6 man bullpen.

You could say that maybe if you did it on the 3rd day ("start" on day 1, rest 2 and 3, available on 4, rest on 5, "start" on 6.) that if he was available for light duty on day 4 that on day 3, the manager could deploy his pitchers differently. That's reasonable. But on 2 and 5, you're a roster spot short.
"One minor benefit is that teams in the National League would have the ability to squeeze out one extra plate appearance from a “real” hitter, rather than a pinch hitter."

Didn't you mean "pitcher" not "pinch hitter"?
Yeah, I did. I may have written that line with a 4 month old wriggling in my arms.
When Bobby Cox managed the Braves, he used something like your option #3. Whenever an off day allowed, he would usually skip his fifth starter, at least in the days when Maddux, Glavine, and Smoltz headed his rotation. That seemed to work rather well. However, I do recall one writer (maybe even in a Baseball Prospectus annual) claiming that this retarded the development of the fifth starter, who was usually a young pitcher being introduced to the majors. I'm not sure what evidence there may have been for that.
Strikes me as unlikely that it would be significantly damaging to come through in the fifth starter role. We regularly read about the Cardinals and others introducing their starters to the bigs as relievers without anyone seriously suggesting that this retards their development. I mean, obviously it helps if the manager is intelligent and looks to get the young guy enough work when the fifth starter isn't required, but that can be done either in the majors or in AAA. It could also be used for pitchers coming back from significant injuries, to limit innings.
For the tandem experiments, we presumably won't only be drawing on the marginal starter pool. If we accept that relievers are mostly failed starters, there are probably a bunch of guys who are currently working in short relief roles who could be semi-stretched out to manage 3-4 innings even if they can't be effective for 6. Then again, the question as to whether a good short reliever is more valuable than a tandem starter may actually be a real issue.
This is an example of why I renew my subscription to BP every year. I have to go to the dentist, because that's some sweet brain candy!

I have an image in my head of a GM making copies of these types of articles and handing them to their head coaches who immediately throw them in the garbage; but you never know when one will actually read it.

The 5th starter getting skipped more often will definitely become prominent very soon. I feel like that movement has already quietly begun.

As for tandem starters, Do you feel a lot of those mediocre 4th & 5th starters would get a boost in stats from being able to let it all hang out, instead of saving a little in the tank for those unlikely late innings?
I think you could quietly try the tandem starter thing just by saving a long reliever for the 5th starter day and then having a very quick hook. If your 5th starter is pitching a perfect game (or even just very effectively) then you let him keep pitching. But once you are past 60 pitches or whatever the next base runner brings in the hook and your ready long reliever. You don't need to announce to the world that your 5th starter is lesser, and your tandem long reliever just goes to work.