Would a six-man rotation work?

A few weeks ago, C.J. Nitkowski asked a simple question. Has the time come for the six-man rotation in baseball? Not for every team, but could there be a team out there that might benefit from spreading the workload around a little bit more? Starters would get more rest and stay fresher, and while they wouldn’€™t get as many starts, they’€™d theoretically get better results during those starts. As my father is fond of saying "€œEverything works in theory."

Mr. Nitkowski mentioned a couple of teams who might benefit from this sort of arrangement, but for different reasons. The Yankees have a packed schedule and several pitchers who have struggled with injuries (CC Sabathia, Masahiro Tanaka). The Nationals have six really good starters. The Mets have six guys who are starters, including Matt Harvey, who is coming back from Tommy John surgery. Remember the Tommy John epidemic? The rash of Tommy John surgeries that happened in 2014 lead some to think that maybe we need to be a little bit nicer to pitchers to protect their arms from blowing out. In the same way that baseball embraced the move from a four-man to a five-man rotation, maybe it’s time to again add another member to the rotation?

Teams occasionally use a six-man rotation in September if they are out of the race and are trying to give a couple of prospects a chance, but we’€™ve never really seen a team commit to it long-term. Still, we do have some evidence that we can take a look at to see whether there’s evidence that a six-man rotation would actually work. Let’s dive in to see what we can find.

Warning! Gory Mathematical Details Ahead!

Do better-rested pitchers get better results?

One nice thing that we have going for us is that baseball teams get the occasional off-day, meaning that sometimes a pitcher is actually starting on five days’€™ rest rather than the normal four. It’€™s tempting to just grab the overall stats for those starts and compare them, but there’€™s a problem. The Dodgers might figure that instead of waiting for Clayton Kershaw’s turn in line to come around, they should simply start him on the fifth day rather than the fifth game. In addition, a pitcher might have been pushed back to six days of rest because he is hurt or he just isn’€™t that good to begin with. We need to be able to control for that before we can answer the question.

Fortunately, with some fancy math — something called the log-odds ratio method –€“ we can. The idea is pretty simple. Let’€™s say that a hitter has a .300 OBP. We’€™d expect that in any given at-bat, he’€™s got a 30 percent chance of getting on base. But let’€™s say we also look at whether it’s night or day during those plate appearances. What if he gets on base in 60 percent of his night time PA’€™s and never during the day? He’€™s probably a vampire. But we also need to account for the fact that sometimes he’€™s facing a pitcher who is extremely good as well; the odds ratio method takes both batter and pitcher into account. (For the initiated, I created a control variable using an odds ratio expectation given the batter’€™s and pitcher’€™s seasonal stats, took the natural log and shoved it into a binary logistic regression.)

Once we controlled for the batter and pitcher, I took a look as to whether those expectations differed from actual reality more or less based on whether the starter (and I only looked at plate appearance where the starter was throwing from 2010-2014) had last started four days ago or five days ago. Did the extra rest make a difference? No. I looked at strikeouts and walks and home runs and singles and a few other outcomes. Pitchers pitching with extra rest don’€™t actually show any extra ability. They don’€™t suffer for it either. They just kinda pitch like we would normally expect.

Do better-rested pitchers get hurt less often?

Again, this one is hard to observe directly. One could make the case that extra rest on a consistent basis leads to extra health. I’€™ve previously studied what happened when teams moved from a four-man rotation to a five-man rotation. We don’€™t have good injury data until about 10 years ago (and by that point, everyone used a five-man rotation), but we can at least look back to see if a pitcher suddenly disappeared from making starts for an extended period of time and we can make some assumptions that he might have been on the shelf for a bit.

Using that method, we can go all the way back to the 1950s. What I found was that during the heyday of the four-man rotation (up until about the 1970s) a pitcher more often going on three days’€™ rest was actually less likely to suffer a mysterious disappearance than was a pitcher who more often pitched with the now-standard four days of rest. (For the initiated, I was using a Cox cumulative risk regression model.) By the 1980s though, it was no longer a positive (but not a negative) to go on three days’€™ rest. By the 1990s and 2000s, pitchers who went on three days’€™ rest were more likely to injure themselves afterward. It seems that it’€™s more about what you’€™re used to than the actual number of days of rest. Going to a six-man rotation wouldn’t necessarily make pitchers less susceptible to the injury bug.

On top of that, it seems that the two biggest contributors to injury risk are previous injury and one-game pitch totals that are insanely high. So, the evidence here suggests (and I’€™m happy to hear a rebuttal) that adding an extra day of rest will not do a lot to protect a starter’€™s health. In fact, it might lead a manager to assume that since he’€™s more rested, a pitcher can throw more pitches in a start, and that’€™s known to be dangerous.

The Clayton Kershaw problem

The biggest objection to a six-man rotation basically comes down to "€œIf you have Clayton Kershaw, why would you want him to start less often?"€ If a team happened to have five Kershaw clones and then picked up Pedro Martinez fresh out of the TARDIS from 1999 (The TARDIS is Dodger blue and the Dodgers do have a lot of money), then it might work to have a six-man rotation. But the fact is that most team’€™s fifth starter is a guy who makes you hold your breath when he takes the mound on his appointed day. The sixth starter is the guy who couldn’€™t beat him for the job in spring training. That’s the guy who will be taking starts away from Kershaw.

There were 185 pitchers who threw more than 50 innings as starters in 2014. If we take the top 30 and call that group the "€œaces€," the pitcher in the middle at #15 (Johnny Cueto) had a WAR (according to Fangraphs) of 4.1. That’€™s your average "€œNo. 1"€ starter. Going to the next group of 30, we see that #45 on the list is John Lackey (2.8 WAR). So, an "average" team might have a starting rotation that looks like this:

Average rotation

Pitcher WAR value
#15 Johnny Cueto 4.1
#45 John Lackey 2.8
#75 Jake Peavy 1.9
#105 Doug Fister 1.3
#135 Kevin Correia 0.7
#165 Justin Masterson 0.3

Your mileage may vary pending on the actual construction of the team, but it’€™s easy to see that letting Justin Masterson make 25 starts or so would rob five starts from Cueto, Lackey, Peavy, Fister and Correia. (Now imagine taking five starts away from Felix Hernandez!)  A starter might make 34 starts over the course of a season under the five-man system, so taking away five of those would rob him of 15-20 percent of his value. Taking away 15 percent of the productivity of these top five "average" starters would rob the team of 1.6 wins and replace them with a guy who is likely near or at replacement level. Ouch.

Playing roulette

Now, just because a strategy has a cost doesn’€™t mean the cost isn’€™t worth it. Letr€’s accept that going to a six-man rotation loses us roughly 1.5 to 2 wins of value because the good starters pitch less often. But what if it gains us something important? For example, let’€™s say that halfway through the season, Cueto suffers a season-ending injury. That’s roughly 2 wins lost. If we knew for a fact that using a five-man rotation would get Cueto hurt but using a six-man rotation would keep him safe, then the six-man rotation is worth it. Given the rash of Tommy John surgeries, maybe itr€™s not such a bad idea. But of course, we never know for certain what’€™s going to happen. We have to guess.

So let’€™s play roulette. Suppose that by switching to a six-man rotation, you could prevent all pitcher injuries, but you would be accepting the 1.5-2 win penalty that comes with letting a replacement-level pitcher start instead of Cueto. Your other option is that you can go with a five-man rotation and snatch that 1.5 wins back. You have a 50 percent chance of Cueto and Lackey getting hurt and wiping out 3 or 4 wins worth of value and a 50-percent chance that they remain healthy. If the chances are equal and there’€™s so much more to lose, then be conservative, right?

It’€™s actually a bad idea to be conservative in situations like this, at least for most teams. Take a look at the Baseball Prospectus’€™ projected standings and notice how many teams are clumped around 80-84 wins. That’€™s not enough to make it to the playoffs, but one or two of them will catch some luck during the season. That might come in the form of a few nice bounces or a breakout season that no one saw coming. Maybe it’€™s just good health in general, but it will push them past what we expected of them by 4 or 5 wins and maybe into a wild-card spot. Why cripple yourself at the outset (remember, the six-man rotation comes with a penalty) and need an extra couple or wins of luck to make the playoffs?

If you use a five-man rotation and get unlucky and Cueto and Lackey get hurt, then you weren’€™t in the playoffs anyway, so it doesn’€™t matter if the net loss is 5 wins or 20. If you use the five-man rotation and get lucky and they stay healthy, you’€™ve reduced the amount of dumb luck you’€™ll need to fake your way to October. For most teams that profile as middle-of-the-pack, there’€™s a lot to be gained from a few extra wins. There’€™s not much to lose from even a bunch of extra losses.

So, the six-man rotation will never work?

Not so fast. Maybe we’€™re being a little too doctrinaire about the six-man rotation. Normally, the six-man rotation assumes that the pitchers will pitch in the order A-B-C-D-E-F. What if we came up with a model that sidestepped a few of the common objections to the six-man rotation? Let’€™s assume that our team has six starters, but three of them (A, B, and C) are rather good, and the coaching staff would rather have those three guys go as often as they can. What if we made it so that those three guys went every fifth day (rather than every fifth game), and resurrected an old role to fit a new need?

The other three gents (D, E, and F) would be swingmen. They would start on the other two (and if an off-day happens, just one) beats, and because there are three of them for two jobs, it’€™s likely that one would be available in the bullpen, at least on some nights. Such a pitcher would probably need to be placed in the game at the beginning of an inning and probably given more warmup notice. Starters usually have a longer warmup routine, but a manager could at least alter his tactical plan for the game to allow for this.

Pitchers D, E, and F stay a little fresher over the course of the season and the setup gives the manager some flexibility. If one member of the rotation needs an extra day because of a specific issue, the replacement is already there. You’€™re really only "stealing"€ starts from D and E, and finding a guy who justifies stealing starts from D and E is a lot easier than finding a guy who justifies stealing them from your ace. Plus, if there is a game coming up against a team that is susceptible to lefties, a manager could simply play around with who pitches when to gain the platoon advantage. The combinations aren’t infinite, but there’€™s some flexibility and fungibility among these swingmen. Maybe that only affects 2-3 games a year, but in a static A-B-C-D-E, a manager is at the mercy of the calendar as to who pitches in that game.

Not only that, but if a team plays its cards right, it might target a very specific type of starter to fill those D, E, and F roles. Some time ago, when I was investigating whether a tandem starter model might work in MLB, I found that there are some pitchers who were good for the first 50 pitches of the day, but not as much for the second 50. They’€™re generally serviceable fourth and fifth starters, but they had this strange pattern that emerged. If their only task was to get through the lineup once, they would be amazing. I used an ERA estimator called xFIP (you can read about it here), for all starters in 2014 and looked at their xFIP over the first 50 pitches of their outings and the second half-century. I found a list of 17 pitchers who were roughly half a run better in ERA at the beginning of their starts than at the end. Sometimes more.

Some of these names (Chris Sale, Felix Hernandez) aren’€™t available because while they were a little better at the beginning of starts than at the end, they were pretty good at the end too. But then, there are names like Jon Niese, Marco Estrada, Brad Hand and Mike Bolsinger. They’€™re the sorts of pitchers who aren’€™t free, but they can be had if you are looking for them and they all better than a league average reliever in their first 50 pitches. Because the three swingmen really only have to cover one or two starts in a five-day period, there’€™s a good chance that one of them could be used out of the ‘€˜pen, at least on most days.

Maybe it would just be to mop up some garbage innings and save the "€œreal"€ relievers, but they would get a chance to throw a few 40- to 50-pitch relief outings, something that they seem rather well-suited to do. That compensates for the fact that the sixth starter’€™s roster spot would probably have to be "borrowed"€ from the bullpen. If each throws this sort of relief outing 10 times each and goes 2-3 innings at a stretch, then the team gets 75 or so bullpen innings that are better than a league-average reliever. That’€™s worth something. Plus, the team gets a couple of serviceable back-end starters and some of the flexibility perks that come with a six-man rotation. Not a bad deal.

Ah, but would anyone go for it?

In theory, this six-man rotation would give the manager the flexibility of six starters, and on a few days a week, a team would have a better-than-average multi-inning reliever in its bullpen. It wouldn’€™t have to take playing time away from its good starters but could give them extra rest if needed. Filling slots A, B, and C (or some variant of this … a four-starter, two-swingman rotation could work as well) would be easy enough.

But would anyone go for the swingman job? Even if it were logistically possible, as C.J. pointed out, pitchers are creatures of habit, and this would be something very different from the usual "You’€™re going out there every fifth day."€ Maybe a team would have trouble staffing the roster with these guys. Who would want to accept a demotion to something that sounds like a "€œsorta" starter. What would a free agent tell his mother when she asks what he’€™d be doing for the new team?

I think here we run into a language problem. Language is a powerful thing and there’€™s no name for this sort of job, and so people don’€™t even think about it. We have a word for the guy who goes out every fifth day and pitches 6-7 innings ("€œstarter"€). We have a word for the guy who pitches an inning or two and rarely, if ever, starts ("€œreliever"€). What do you call a guy who does a little bit of both, and does so purposefully? Right now, the only name for that is "œswingman," but that usually is a nice term for "€œthe guy who made the team as a minor-league invite and who we mostly send out there as a sacrificial lamb when we’€™ve run out of other warm bodies and who will probably be sent to AAA at the next convenient opportunity."€

I wonder if teams would do this if they could say to a pitcher, "€œWe want you to be a gribble for us,"€ and everyone knew what that meant.

Sometimes it’€™s worth thinking outside of the box and challenging the assumption that A-B-C-D-E is the best way to construct a team and a roster. Sometimes, we lack the ability to even conceive of such things for want of a good word to talk about what we mean. A team that embraced this six-man rotation model and who could convince three guys to take on this new/old gribble/swingman role could probably find guys whom the league only valued as back-end starters and turn a bit of straw into gold as a result. Maybe they’d even find pitchers on the cheap and could free up money in the payroll to shore up some other area. Of course, it wouldn’€™t actually be this easy to execute, but I’€™d say that in the season of crazy ideas that is spring training, it’s at least worth a second thought.

Maybe six is the new five, just not in the way that everyone assumed it would be.