Last year, Russell A. Carleton wrote several pieces dubbed the Fake Manager series, for which he took control of the 2005 Cubs and tried to act as a real manager would. It’s a great series and left me wondering about one thing: tandem starters. Could they work?
Carleton has written a few more pieces on the subject (you can find those here, here, and here). In these articles, he posits a model in which paired pitchers (tandems) throw 50 pitches each, in theory working as a single starter. The catch is that these starters wouldn’t pitch on five days’ rest but on a two- or three-day regimen.
It makes sense—you want to avoid the times-through-the-order penalty as much as possible and keep the other team/manager on their toes. But, like he wrote in the first article, there’s no clear linearity between the amount of work a pitcher could handle and days needed to rest.
So, why revisit the subject? Well, using BP’s new descriptive pitching metrics, it might be time to take another look at the notion. By focusing on pitchers’ Stamina Score—which measures the workload a pitcher could handle on a six-day rolling average—we want to find which types of pitchers would be ideal candidates for such an experiment.
A Rotation Spot Made For Two
I looked at starters and relievers between 2008 and 2017—the years for which we currently have Stamina Scores—binned them by stamina score, and checked the number of batters faced, innings pitched, and our usual pitching metrics (ERA, DRA, and FIP).
Before I show you the results, I want to lay some ground rules for the data set I used:
- Starting pitchers are pitchers who threw at least 80 percent of their innings in starts.
- Relievers are those who threw at least 80 percent of their innings in relief appearances.
- For starters, there’s a minimum of 60 innings to avoid spot starters and those who suffered an injury.
- For relievers, the minimum number of innings is based on the median: 32.
This is how the bins look for starters:
|STM range||# Pitchers||Avg IP||Avg BF||ERA||DRA||FIP|
As to be expected, as the Stamina Score (STM) increases, so does the number of innings pitched and batters faced. No surprise there, as this is what Stamina Score measures. But look at ERA, DRA, and FIP. As stamina increases, these get worse, seeming to be inversely correlated.
We can also add that, for starters, our innings-eaters, work horses, or aces appear to have a Stamina Score higher than 70. They’re able to face multiple batters—or throw multiple innings—without suffering much lateral effects on their next outings. In other words, they’re consistent.
How do relievers measure up?
|STM range||# Pitchers||Avg Age||Avg IP||Avg BF||ERA||DRA||FIP|
Contrary to starters, relievers are a more closely-knit group—63 percent of relievers are in the (40, 50] bin. It makes sense. Most relievers are dishing out 20-25 pitches at maximum effort and it’s taxing on them to do it again on continuous days. We don’t want that. We want to focus on relief pitchers who can handle the strain of three innings every two or three days. So, we’ll set a minimum stamina score of 50 to try to squeeze as much as possible from our relievers.
Now that we’ve established who our targets are, here comes the interesting part: building a group of tandem starters.
With the parameters we’ve set for starters and relievers, our current data set contains 1,319 candidates. If we focus solely on 2017 pitchers, then our data set comes down to 350 potential pitchers, grouped the following way:
|STM range||# Pitchers||Avg Age||Avg IP||Avg BF||ERA||DRA||FIP|
By mixing and matching these 350 players, we could get anywhere between 64 to 342 potential innings from a tandem starter over the course of a season. However, the idea is to try to get as many innings as possible without burning out our pitchers. After much thought, our new pitching corps would have to look something like this:
- Four standard starters; 6-7 IP/outing
- Four traditional relievers; 2/3-1 1/3 IP/outing
- One “closer”; 1 IP/outing
- Four tandem starters; 3-3 2/3 IP/outing
Yes, we would want four tandem starters because we also need to account for the fact that they may be needed for some mop-up duty every now and then—long relief outings when one of the traditional starters is knocked out early. This also helps save the arms of the four traditional relievers.
The pitching cycle would probably look like this:
SP1 > SP2 > Tandem Pair > SP3 > SP4 > Tandem Pair > SP1 …
Just looking at 2017, this is how many pitchers each team had that would qualify for tandem starting:
Given their pitching depth, it’s no surprise that the Dodgers topped the list. But the Angels? Well, they had three starters who could’ve been used in tandem and 13 relievers who could’ve probably been pushed for one or two more outs.
As is, every team would be capable—to some extent—of deploying this model and test if it works.
Tandems vs. Regulars
Let’s assume teams are open to the idea. Who gets replaced in the rotation? No, the ace (or first pitcher in the rotation) is not the answer. Innings-eaters aren’t viable candidates, either. The only option that makes sense is the fifth starter (and maybe some fourth starters).
|Starter Spot||Number pitchers||Avg Stm Score||Average GS||Average IP||Median ERA||Median DRA||Median FIP||IP per Start|
Some notes about this table.
- The starting spot is based on the number of innings per player per team. In other words, if Pitcher 1 on Team A threw 160 innings, and Pitcher 2 threw 157 innings, then Pitcher 1 would be SP1 and Pitcher 2 would SP2, and so on.
- The minimum innings is 120, which was set to avoid spot starters or single-game call-ups.
- I chose to use median ERA, DRA, and FIP as it shows the midpoint for all pitchers in the bucket rather than doing a mean of means.
- There’s a noticeable drop in SP5 because fifth starters are usually interchanged with spot starters and rookies who may have an innings limit.
Looking back at the tandem starter bins, we could envision replacing the fifth starter within a duo. In fact, running a random simulation pairing of 260 pitchers, we get the following results:
|# Tandem Pairs||Avg Stm Score||Average IP||Median ERA||Median DRA||Median FIP|
The average Stamina Score is 51.1. If a fifth starter has an average score of 67.0 and delivers 5 2/3 IP, then by having two pitchers average 51.1, we could get at least six innings. If that’s the case, then we’ve got a good duo.
We should also point out that tandem pitchers are throwing eight fewer innings than a regular starter, which is not bad. But it does mean that another pitcher is going to have to cover those innings, which may mean extra work for the bullpen or the closer getting some four- or five-out saves. Nevertheless, bear in mind that in our pitching corps we have four tandem starters, so we shouldn’t dwell that much on this.
What is remarkable, however, is that ERA sees tandem starters as better than a fifth starter. In fact, the only pitcher who looks better by this metric is the rotation’s ace. FIP, however, has them a bit worse than the rotation, and DRA’s outlook is pessimistic to say the least.
All in all, tandem starters don’t seem like a bad idea. We’re just replacing a fifth starter—and perhaps a fourth one—with two extra bodies.
It Takes Two Roster Spots To Tango
Tandem starters would require using an “extra” roster spot that could go to some other, more valuable player. It’s a roster spot that instead of going to a bullpen ace—though not the closer—goes to some guy who couldn’t make it as a starter but could give you three innings every three days.
There are also many other potential issues with such a model. Injuries, spot starts, extra innings—even playoff usage, if it comes to that—are not being factored into my calculations, and any of these could probably deviate the use of the tandem starter, let alone the rotation.
Also, Stamina Scores are measured taking five days’ of rest into account. We don’t know how scores would reflect players—or even how fatigue would look—unless teams start to experiment with a tandem starter.
Finally, one last issue that Carleton mentioned was that teams would have to groom players in their farm system to work in such a situation, something that would be more difficult for pitchers acquired outside the organization. Pitchers tend to have either long outings (starters) or really short ones (relievers). The use of a tandem model would require new preparation methods to acclimate themselves to such work conditions—something that may not be so simple. Then again, if a team decides to experiment, rest assured that they will bear this in mind.
In his Fake Manager series, Carleton managed to take the Cubs to the 2005 playoffs using the tandem starter model. Unfortunately, he did not make it out of the first round of the playoffs and the tandem model died there. Unless another brave soul takes up the mantle—or, even better, a real team does it—we may never know if it would work in the long run.
Thanks to Russell A. Carleton for listening to my crazy ideas and giving me feedback.