Saturday, February 3, 2018

Should the Astros consider Starting Pitching tandems 40% of the time??

The Background to the Question:
Heading into the 2018 baseball seasons, a number of pitching-related facts are apparent:
  • The Astros are the defending World Series Champions (I never get tired of saying that!)
  • There is a perception that the Astros struggled to consistently fill their rotation throughout 2017, and as a result, the bullpen felt gassed at the end of the year*
  • Their bullpen - at least in terms of their full-time relief pitching - struggled in the playoffs**
  • The Astros' most frequent starters' in 2017 are not what one would define as "workhorses"***
  • The Astros appear to have an excess of starting pitchers next year, including presumed top-5 of Justin Verlander, Dallas Keuchel, Gerrit Cole, Lance McCullers and Charlie Morton.  Other available starters with significant major-league starting experience include Brad Peacock, Collin McHugh and Frances Martes.  This doesn't include other potential starters with upper-level minor-league success, such as David Paulino, Forrest Whitley and Rogelio Armenteros.
* Astros starters' IP were 10th in baseball with 899.2.  But, the Astros managed only 1 complete game, and handsomely lead the league in "cheap wins" (wins in non-quality starts) with 23.  The league average was 14.  The 899.2IP is more a measure of the effectiveness of the pitching staff, rather than the durability.
** FIP's 2017 regular season versus 2017 post season for Astros relievers >5IP in postseason: Ken Giles 2.39 / 7.59; Will Harris 3.33 / 5.66; Chris Devenski 3.49 / 5.53; Joe Musgrove 4.38 / 8.56
*** Mike Fiers was the only Astro to start more than 25 games, and he will be pitching for the Tigers in 2018.  Fiers was followed in number of games started by Charlie Morton (25), Dallas Keuchel (23), Lance McCullers (22) and Bradley J. Peacock (21).  Of those, only Keuchel has at least one 200IP season under his belt (2).

The Astros' rotation looks much more durable entering 2018, compared to 2017.  Recall that the rotation entering the 2017 season looked like Keuchel, McCullers, Morton, McHugh and Fiers.  Three seasons of 200IP or more* lie amongst that starting five.  Heading into 2018, the presumed or projected rotation of Verlander**, Keuchel, Cole, McCullers and Morton looks a lot more durable, with 14 seasons of 200IP or more on their resumes.***

* I know it is a dirty, imperfect cutoff, but I am going to use 200IP seasons as a measure of "durability"  in this article.  To pitch 200 innings, a starter would need to take the ball 32 times per year, and average six-and-one-third innings each start.  I think that is a good proxy measure of (i) the ability to take the ball repeatedly over the course of the year, and (ii) the ability to remain effective enough so that the starter can pitch deep-ish into most of their games.
** Verlander is an incredible workhorse.  10 seasons of 200IP since 2005, led baseball in IP three times in four years.  One of the under-reported positives of the Verlander trade - not only did they add a potential top-of-the-rotation starter, but they added one who doesn't seem to ever get injured.  His non-debut career-low for IP in a season is 133.1, which was recorded in 2015.  As long as Verlander is effective, he looks like a good bet to continue to rack up 30+ games starter per year.
*** It is not so much the raw number of 200IP seasons because 10 of them are attributable to Verlander, but more the fact that three different starters have multiple 200IP seasons in 2018, rather than one player in 2017.

So how might it look??
This idea is pretty simple.  In games started by Verlander, Keuchel and Cole, it should business-as-usual with regards to starting pitching and bullpen use.  This is partly because all of them are top-of-the-rotation types, and partly because they have a history of durability and effectiveness as outlined above.  For the sake of argument and to preserve the bullpen, lets not put them on back to back days.

To me, McCullers seems to be an injury risk going forward.  He is small, yet he pumps gas in the mid-90's much of the time.  He throws breaking balls around 50% of the time, which must put stress on his arm and shoulder.  He has missed time with injuries the last few seasons*.  I am not saying he can never throw a 200IP season, but (i) I think it is unlikely and (ii) it won't happen this year, because the Astros would never let it happen following a season in which he pitched 118-and-two-thirds.

* Although, arguably, in his 2015 year, his "injuries" were as much about keeping innings off his arm.  Much more concerning is McCullers' history since then, when he has had to rework his throwing motion to alleviate stress on his shoulder.

Morton is older, but is in a similar boat.  He was famously signed last offseason after a year in which he managed 17IP, because of a torn hamstring.  In addition to this, Morton also suddenly added velocity to his fastball late in his career, and I wonder how this might play with shoulders and elbows (although there is no indication that this is currently a problem).  Morton's career high in IP is 171.2, set way back in 2011.  Since then, his year-to-year inning totals are as follows: 50.1, 116, 157.1, 129, 17.1 and 146.2 (the last of which was with the Astros).

McCullers and Morton are probably the key players in this consideration.  Both are pitchers with incredible stuff, and are virtually unhittable when they are locating with both their hard and breaking stuff.  If the Astros think that this idea would take a significant number of innings from McCullers and Morton, then I doubt they would consider it.  However, given the high chance that the Astros make the playoffs next year, I would rather see both entering October with a solid year under their belts and feeling ready to pitch for another month, rather than setting career highs in IP during the regular season, and feeling gassed for the postseason.

So your potential rotation, with tandems, could well look something like this:
  1. Justin Verlander
  2. Lance McCullers / Brad Peacock
  3. Gerrit Cole
  4. Charlie Morton / Collin McHugh
  5. Dallas Keuchel
A couple of notes here.  Firstly, the order is not important, and has been created as much for demonstration purposes as anything else.  Note that the only back-to-back "normal" days for the 'pen would be on starts from Dallas Keuchel and Justin Verlander - both the Astros' most effective horses.  This could change, of course, as the season progresses. Secondly, the tandem pairs are also negotiable.  I don't have a strong opinion around who goes with who, but I think that McCullers and Morton are the best bet to 'lead' both pairs, and therefore probably should not be paired together, despite what happened in Game 7 of the ALCS.  Also, McCullers and Peacock together should bring back warm memories of Game 3 of the World Series, when Bradley J finished the game with three-and-two-thirds of scoreless, hitless baseball.

What are the advantages??
I think that the principle advantage of this is to spread the workload more evenly around your starting and relief pitching, with an eye toward October.  The argument for that is laid out earlier in this article, and is based on the premise that health, for starting pitchers, is almost stand-alone skill with a number of contributing variables (such as height, build, in-game pitch management, mechanics, conditioning and genetics).

But would this also address the "times-through-the-order" penalty??  The premise around the times-through-the-order penalty is that starting pitchers tend to struggle the more times through the order that they face an opposition batting lineup.  I guess there are a number of factors here, including the fact that they are likely to be tiring more the longer the game wears on, and the fact that the opposition batters have had more time to adjust to what they are throwing.

However, the times-through-the-order penalty also as a potential source of bias - if a starting pitcher is not effective, then they will not get to face the opposition 3 or 4 times through the order, and therefore their ineffective outings are masked by not contributing to statistics from late-game situations.  In addition to this, ineffective and/or short outings are likely to further bias the sample by artificially increasing the early game stats.*

* For example, if a Justin Verlander is dominant, and shuts the opponent down for the first 18 outs, then he probably gets to go through the order a third, and perhaps a fourth, time.  Let's say that the last six outs are also the fourth time through the order.  The only reason that he is even getting that opportunity is because he was effective the first three times through the order.

Lets then pretend that Justin Verlander then gets shelled in his next start, being knocked out in the 3rd inning after giving up 8 ER.  He wouldn't get an opportunity to pitch the third and fourth time through the order, and his stats for the first two times through would look pretty awful.  So the real effect of these two starts is that the late-game statistics would be largely dependant on early-game dominance, whereas early-game ineffectiveness protects the late-game statistics to an extent.  Therefore, the times-through-the-order penalty, on a global level, is most likely more uneven than one would think based on the statistics.  There are also issues around sample-size discrepancies.  One solution would be to look in detail at the more granular data, such as batter exit velocities and launch angles, for a better guide of the actual extent of the penalty.

Below are some measures of the times-through-the-order penalty for the seven potential Astros' starters for 2017, acknowledging the imperfections mentioned above.  The format is name: OPS's in order from first to fourth; wOBA's in the same order in italics (sorry about the morass of numbers):
  • Justin Verlander: .578, .739, .659, .753; .252, .314, .287, .340
  • Dallas Keuchel: .531, .678, .583, .1.411; .232, .298, .260, .594
  • Gerrit Cole*: .749, .627, .852, .929; .320, .271, .360, .392
  • Lance McCullers: .536, .664, 1.004, 2.000 (in 1 at-bat); .244, .292, .422, .877
  • Charlie Morton: .544, .789, .789, .000 (in 5 at-bats); .245, .339, .335, .000
  • Brad Peacock (as SP only): .420, .691, .995, .000 (in 1 at-bat); .196, .304, .420, .000
  • Collin McHugh: .726, .653, .949; .322, .281, .396
* obviously, Cole pitched in the NL last year, which may have led to a few removals because his place in the batting order was due up.

Ok, so the above table turns up a few clues.  Brad Peacock was actually the most effective of all the starting pitchers the first time through the other.  Verlander and Keuchel appeared to be the two most dominant pitchers on the third time through the order.  Gerrit Cole seemed to tire some.  Lance McCullers and Collin McHugh appeared to struggle the most with the third time through the order last year.  But broadly speaking, these statistics support the overall shape of the rotation and proposed candidates for tandem stating.

What would some other advantages be??  I can think of a few, and you may wish to add to them in the comments:
  • Keeping effective pitchers in a regular rotation
  • Distributing more innings to the non-starters in a manageable way (or, predictable innings for Peacock and McHugh)
  • "Resting" the bullpen for 40% of games
  • Having 7 pitchers on the ML roster who are stretched out and/or are used to multi-inning appearances, meaning that spot starters are easier to find without having to muck around with the 40-man roster
  • Having the opportunity to "flip" pitchers around - for example McHugh could start one of every three games, and Morton two, or they could take it on a start-to-start basis, in order to minimise injury and fatigue and keep the pitchers on regular rotation
  • Better pitching management in the event of double headers
  • Rest days can be used more effectively, as the tandems can be bumped up a day if the other starters would benefit from extra rest
  • Days of work would be more predictable, allowing relievers to rest, work on the side or work with long toss or weights to keep themselves fresh during days off
What do the innings numbers look like??
Ha-ha!! An exercise in futility.  People have been trying to project innings pitched since baseball was invented, I imagine.  But, assuming no injuries (another ha!) let's take a look at how this may all shake out...

The baseball season is 162 games long, 81 home, 81 away.  In all home games, the home side will have to pitch 9 frames.  In some of the away games, the away team will pitch either 8 innings (if they lose) or 9 innings (if they win).  Lets pretend the Astros a record of approximately .500 on the road.

Therefore: total number of IP = (81*9) + (41*8) + (40*9).  Or 1417 IP in total to distribute around the staff.

Lets give Verlander, Keuchel and Cole 200IP apiece, because we have been kind of obsessed with that figure to this point.  That leaves 817 innings to be pitched.

Lets then pretend that the tandems each make 33 starts, and throw 9IP each time (297 innings), like it is all drawn up.  That subtracts another 594 innings.

So a McCullers and Peacock tandem would throw 297 innings, and it would have to be divided up somehow.  If McCullers went five each time, he would throw 165IP on the year, whereas Peacock would throw 132IP.  For that particular tandem, I probably would weight it a little more toward 50-50, but one could squint, and see both pitchers being able to handle that workload.  Of course, there may be games where McCullers is efficient, and manages to go 6 instead of five.  Or there may be times when Peacock winds up getting stretched out for more than 4 innings at a time, either because he is the "starter" for that day, or because he enters the game earlier than anticipated for some reason.

That leaves 223 innings between 5 other 'pen pitchers, or about 45IP per year per pitcher, if evenly distributed.

This is a quick-and-dirty analysis, of course, which does not account for extra-inning games, or other multi-inning relievers, such as Chris Devenski.  But it provides a useful guide as to what sort of numbers we are looking at.  Personally, having a closer who has only thrown 50-odd innings over the season entering the postseason seems like a great idea.*

* Ken Giles' regular season IP since his debut in 2014: 45.2, 70, 65.2, 62.2 with 7.2 more innings of 11.74 ERA in the postseason.

But there would be disadvantages too, right??
I imagine there would be some disadvantages, too, including but not restricted to:
  • Ego of the starters.  There is a certain sense of achievement around being a starting pitcher, and this way four of the potential Astros starters would need to give that up
  • Ego of the bullpen.  This means that narrow leads in 40% of games would most likely not be given to the Nominal Closer to protect.  I don't really care about that, but some would, possibly including the Closer himself.  Fewer saves may mean less money in arbitration.  Giles did just win his arbitration case against the Astros, but that is largely inconsequential
  • Negative press.  This is a great idea, until Collin McHugh or Brad Peacock blows a save in the bottom of the ninth.  Then, because it is a "different" idea, the hacks will go nuts, writing it off a case of a smart team outsmarting itself.  (I am, of course, assuming that this is a "smart" idea).  This is how things work in Baseball, given the traditional nature of the the game and how it should be managed.  This leads on to...
  • Managing high leverage innings over the course of the season.  All of this becomes much easier if the Astros jump out to an early division lead, and cruise the rest of the season.  If that happens, they can afford to throw the odd high-leverage inning (9th inning, one run lead) at a Peacock or McHugh.  If they don't get out to an early lead, they may not have that luxury.
  • Getting shelled or long, extra-inning games.  This is not a great idea if a starter - especially one of the "horses", records a very short start.  Then it may throw the 'pen into disarray because you no longer carry a long-man, or force one of the next tandem starting pair into action a day early.  If this was an infrequent occurrence, it could be managed fairly easily, but a bad week or two would throw this idea into chaos.
  • Mopping up.  This plan essentially moves the mop-up guys into a starting role.  Who mops up in the event that inconsequential innings need to be absorbed??  Are you comfortable seeing Joe Smith or Ken Giles throwing in a 10-run game?
  • Long breaks.  Bullpen guys do best with regular work.  This idea is less flexible for the 'pen guys, because there would be days blocked out where they almost definitely would not get to appear.
  • Predictability.  Imagine being the other team.  "Right guys, we need to read up on Lance McCullers.  Then we know Brad Peacock will follow him.  We don't have to worry about any of the bullpen - we know we probably won't see them today".  So there's that... not sure how relevant that may be.
  • In game timing.  This may be less of a problem than I would think, as all of Peacock, McHugh, Morton and McCullers came out of the 'pen at various times during the postseason.  Peacock and McCullers also tandem-started in the minors.  I would think they would manage their throwing routine better as the season went on, but there may be a couple of hairy moments if the game threw up a couple of very quick innings in a row around the time the second pitcher was warming up.
  • Roster inflexibility - adding one more mop-up guy may restrict the size of the bench.  I will discuss this in more detail below to see how it may work.
I am sure that there are other disadvantages, too.  I would suggest that a key unmentioned disadvantage would be watching Nolan Ryan's head explode in disbelief as he watches Lance McCullers being pulled with a shutout going after 5.  Real men, after all, should be able to go 300IP per year, while striking out 450.

How would this look with the roster??
This is perhaps the main area of concern.  I don't think that this idea can be easily implemented without the result of reduced flexibility in other areas.  But the Astros may think it is worthwhile, if the advantages were to outweigh the disadvantages.

So lets fill out the roster.  In our hypothetical setup, we would need 7 starting or tandem pitchers: Verlander, Keuchel, McCullers, Cole, Peacock, Morton, McHugh.  

We also need 9 position players: McCann, Gurriel, Altuve, Correa, Bregman, Fisher, Springer and Reddick.  Gattis will DH for the purposes of this exercise.  That makes 16 out of the 25 man roster, leaving 9 spots for the bench and the remainder of the bullpen.

We need some reserves, and we already think that we can manage with a bench of four.  So lets go with Stassi, González and Marisnick for three of the spots.  Marisnick can cover the three outfield positions, González can cover all infield positions and LF as well as providing the sole lefty bat off the bench, and Stassi is the backup catcher.  The Astros have already suggested that this is how they are entering 2018 with regards to catching.  Lets throw a bat in there as well - say Tyler White, who seems to have the best combination of upside and proximity to the Majors.  But you could bang Tony Kemp or A.J. Reed in here if you want a lefty.  Or J.D. Davis if you want power.*

* J.D. Davis has the potential advantage of being a mop-up pitcher too.  He was considered as a legitimate draft choice as a pitcher, and is known to have a great arm.  Last year, he threw 1.2 innings, allowing one hit and one walk while striking out three.  I know the Astros love them some flexibility, and I doubt that this gets any real consideration at all, but if it is close between Davis and another player... could this be the tie-breaker??

So that leaves us with a 5 man "bullpen" (not including the two tandem starters).  I am going out on a limb here, and thinking that Ken Giles, Will Harris and Joe Smith get three of the spots.  Chris Devenski will get the fourth spot*, and I would recommend that he nearly only be used in multi-inning stints to keep his overall appearance numbers down.  That leaves one slot in an otherwise all-righty 'pen.  If the Astros weren't worried about handedness, then Hector Rondon would be the fifth guy (although I think he is not that low on the pecking order), with James Hoyt as the main backup.

* unless the Astros want to send him to AAA to stretch out as a starter.  If that was the case, then is Devenski palpably worse than Collin McHugh??  One has options, and the other doesn't - and I think could be considered because the Astros hate giving something away for nothing.

If the Astros wanted a lefty, then Tony Sipp, Buddy Boshers, or Anthony Gose would be the three options currently on the 40-man roster.

Tony Sipp, of course, is owed 6MM this coming year, and has managed to survive on the Astros' roster despite two poor years.*  Buddy Boshers is new to the Astros' organisation, but has an option, potentially making it less likely that he breaks camp with the team.**  

* Tony Sipp's ERA / FIP over the last four years, in chronological order: 2014 - 3.38 / 2.93; 2015 - 1.99 / 2.93; 2016 - 4.99 / 6.19; 2017 - 5.79 / 5.22.  He hasn't been good for two years.  I think he gets Spring Training to show what he has, and that is all.  But the Astros seem to have supported him through thick and thin, so perhaps my read of the situation is poor.
** However, Boshers has a career line of .243/.277/.344 against same-handed batters, including a .220/.258/.397 line in 2017.  That said, much of that is from his dominance in pitching-friendly Target Field, where he has a career .230/.269/.284 line.  His stats are considerably worse on the road.

Anthony Gose is, potentially, the most interesting option here if the Astros want to carry a lefty.  If they don't put him on the roster, then he needs to be offered back to the Rangers.  This is Gose's second stint with the Astros, after he was involved in the Roy Oswalt trade with Philadelphia.  The Astros quickly flipped him for "hitting machine" Brett Wallace*, and Gose became a Blue Jay instead.  However, last year, Gose started transitioning to a pitching role in 2017, a decision that was made easier by (i) his career .240/.309/.348 batting line and (ii) a 100-mph fastball from the left side.

* the Batguy sums up Wallace's tenure with the Astros in this excellent article

Gose also potentially also fits with the Astros is as a two-way player.  The Rangers agreed to give him a look as both an outfielder and a pitcher, but failed to add him to their 40-man.  The Astros swooped in and grabbed in him the Rule 5 draft, and (at least) get to look at him in the offseason and in Spring Training.  While the Astros have stated (in the previous link) that they want to look at him as a pitcher, there are a variety of ways in which Gose's flexibility could be used.  This includes as a pinch runner*, or as someone who could spend some time in the outfield while Joe Smith or Will Harris comes on to mow down the righties.  It would be shock if Anthony Gose made Jake Marisnick expendable as a reserve outfielder, so I think any position role will be limited.  Taking this idea to the next level, Anthony Gose and J.D.Davis could combine to mop up innings, with Marwin González shuffling between third base and left field depending on the handedness of the batter.

* Gose has a career 57 stolen bases, against 23 times caught stealing. 

Conclusion
The Astros have enviable depth, particularly in the starting pitching ranks.  Having tandems to pair with McCullers and Morton is a possible solution to the problem of the durability of those two particular starters, while trying to keep them fresh for another possible October run.  This would arguably make the roster slightly less flexible, but as discussed above, there are other ideas that increase the flexibility of their roster, and may make this manageable.  If the Astros want to carry Anthony Gose on their 25-man roster to start the year, then his flexibility may contribute as part of the solution.

At this point, this is an interesting hypothetical discussion point.  So feel free to comment.  Does this idea have legs??  Or is it too crazy, even for the Astros?

ADDIT:
This article is a suggestion about what the Astros do in 2018.  It does not mean that Lance McCullers should always be used in tandem, or that he (or any other pitcher mentioned) is incapable of throwing 200 frames in a year, then being effective in the postseason.  I meant to put that into the original post, but it slipped my mind.  Apologies.

10 comments:

ntxlfty said...

The union would never go for it. How would it affect stats, in effect, salaries?

Masked Marvel said...

Couple of points.

The union has it's hands full at the moment, with the "free agent crisis" happening. They have to also monitor how many seats on the team bus can be reserved for personal chefs and manicurists.

Secondly, the union can't really give any team advice about how they run their baseball teams. They can advocate on behalf of the players, but that would only involve helping the player voice their concerns.

Thirdly, rate stats - such as ERA, K/9 etc - would probably improve if these starters were able to tandem. It is the counting stats that would suffer, such as wins or innings pitched. No one knows what stats the arbitration panel take into account, but doubtless that it is a mix of both types of stats, and that someone's "special role" would be acknowledged.

Finally, if McCullers were to go 5.1IP, and Peacock inherited the lead and took over to finish the remaining 3.2 innings, McCullers would get the win, and Peacock the save.

Anonymous said...

I think you're understating the BP role in tandem starts. In majority of milb games we use a reliever to clean up the last inning of the starter and often still employs a closer if the second starter isn't efficient. Love this idea btw. I'd nix White, and keep Gose.

Tom E. Snyder said...

Absolutely! I've been pushing this since the playoffs!

Anonymous said...

Just make Morton the closer.

bradness said...

This tandem scheme has been employed in the Astros' lower minors at least as far back as 2013 or thereabouts. Keuchel, himself, was even engaged in it at some point.

Not sure if it ever rose to AA Corpus or AAA Fresno (or OKC, prior), and also not sure if it's in use currently anywhere in the system, but if nothing else, Astro brass would already have plenty of stats/opinions on the tandem bit, only because they've already done it in the organization.

Not sure, of course, how their ultimate opinions shook out on it all, but they're one of the few organizations to have tried it before.

Hope this hasn't ripped the clothes off your jaunt down the street, Emperor Masked Marvel......just keepin' it rill!

OneMinus said...

Great piece of writing, very interesting concept, please keep this up.

I think the biggest obstacle is the reduced flexibility. If we employ this sort of strategy, we choose to pitch 2 relievers only twice a week. But if one starter crumbles early in the game, the whole structure falls apart and we end up using multiple relievers for longer innings, in consecutive days. That's a tough to stomach on the backs of last years summer bullpen meltdown, where overworked relievers struggled to finish games. We are essentially stationing this to happen again if we reduce the relievers role down so narrowly.

Again, very interesting concept, but I just think the risks outweigh the rewards.

bradness said...

Well put, OneMinus! This is one of those things that looks incredible on paper, but once put on its wheels in "real life," so many unforeseen wrenches (injuries, egos, 2-inning starts, logistical paperwork nightmare, etc) get, magically, thrown into the engine that throws the whole paradigm into a tailspin!

The best outcome for this proposal? Grab a vintage Strat-O-Matic dice/computer card game, and go nuts with it! Then, and only then will this merry-go-round mound of rebound even come close to working....effectively and efficiently.

In the meantime, anyone got any dice to lend A.J.?

KBassStache said...

What this really points out is that th Astros are poised to take advantage of a team with the need for a starter come the end of spring training...

...but ignoring that, I love this idea and have been pitching some version of this for a couple of months now.

I think that in reality though, this would take an even more drastic reconsideration of how bullpens are used as the single inning relief appearance would need to start disappearing.

I think that while limiting the Morton and McCullers innings is great, there is probably also an inefficiency in one inning guys. I think you start having more go once through the lineup.

But yeah, it would be great to try this out once you have a u game lead in May.

extrainningsbaseball said...

Honestly was not expecting much, but there are some good ideas here explained quite well. The numbers back up your supposition and, while they would still need short relievers for piggyback games, this is not totally implausible. Keep up the outside-the-box thinking.