Yesterday, we talked about the four man rotation experiment that the Rockies are trying out, and while I expressed some skepticism that it is going to work in their specific situation, I do applaud the effort to try something new. What we know about the relative of performance of starters and relievers suggests that teams could theoretically get better run prevention by getting more innings to their relievers — or at least pitchers working in a role that looks something like a reliever.
So, is there a way to create a pitching staff where this effect is taken advantage of more thoroughly? In looking at the requirements that a pitching staff faces over the course of a season, I think there just might be.
A major league pitching staff is asked to throw about 1,450 innings per season. Last year, it took an average major league team 6,175 batters to get through those innings during the course of the year — or, essentially, 38 batters per game. Because each plate appearance takes about 3.8 pitches, each team threw about 145 pitches each time they took the field, broken down by about 97 for the starters and 48 for the bullpen. An average relief performance involves 17 pitches thrown, so in a standard MLB game, it takes three relievers to finish off the day’s work after the starter leaves. The average MLB game right now involves four pitchers, with one doing two-thirds of the work and the other three splitting the remaining one-third.
What if we kept that four-pitcher framework, but redistributed the amount of pitches they threw within a single game? If we made it completely equal — with each pitcher essentially doing one-fourth of the day’s work — we’d need them each to throw 36 pitches. Given that it takes about 3.8 pitches per batter faced, that’s essentially nine batters per pitcher — or one time through the line-up for each pitcher.
Based on what we know about a pitcher’s performance against the same batter multiple times within a single game, that could be exactly the kind of break down we should be looking for. I showed this data in yesterday’s post too, but it’s worth repeating here:
First PA vs SP: .247/.310/.393
Second PA vs SP: .260/.321/.417
Third PA vs SP: .271/.332/.444
The first time through the order, opposing batters put up a .704 OPS against starting pitchers, but that jumps to .738 the second time through and .776 the third time through. Most of the gap in starter/reliever performance comes from the fact that starting pitchers have to face hitters several times per game. Even with the fact that starters have to pace themselves for longer outings and have to face line-ups built to counteract their own platoon splits, the gap between OPS versus starters (.704) on the first match-up and OPS versus relievers on the first match-up (.691) is only 13 points. This data suggest that a significant part of the advantage of being a reliever is the one-and-done nature of the match-ups.
So asking four pitchers to go through the line-up one time each would get you 36 batters faced — or 95% of the way toward a complete game — based on normal standards under the current system. To make sure those last several batters are covered — and to give yourself a little more flexibility for extra-inning affairs — let’s be safe and say that each pitcher will be asked to face 10 batters per game, which translates to about 38 pitches apiece.
With 12 roster spots allocated for pitchers on most teams these days, that would leave you with three groups of four pitchers. Essentially, each pitcher would be given two days off between outings. Over the course of the season, each pitcher would have 54 expected outings and would have to throw about 2,000 pitches, or about one-third less than a standard starting pitcher.
From an overall workload standpoint, this is something like the role Alfredo Aceves filled for the Boston Red Sox last year. He pitched in 55 games (four of those being starts), faced 474 batters and threw 1,757 pitches. They didn’t deploy him in 35-pitch outings every three days, but over the season, his workload approximated what we’d be asking from every pitcher on our staff under this proposal. It would certainly require an adjustment, but it isn’t so strenuous that it couldn’t be done.
If we accept that it’s an acceptable work/rest balance that wouldn’t shred arms and ask pitchers to do things of which they weren’t capable, then the question is simply whether we could expect this system to be an improvement over the current setup.
The times-through-the-order bonus would essentially transfer 15 plate appearances from a second or third look to a first look, so dealing with league averages again, you’d be swapping out 15 plate appearances with an expected .753 OPS for 15 PAs with a (roughly) expected .704 OPS. That’s 15 plate appearances in each game, or 2,430 plate appearances over an entire season. That’s a significant difference, and a potentially large advantage in favor of the face-everyone-once strategy.
The cost of this system is the lack of ability to mix-and-match based on platoon situations and based on leverage. By rigidly asking each pitcher to face nine to 10 batters per outing — regardless of who those batters were or what the score was — you couldn’t reallocate the more important innings to the better pitchers. You’d also have to do away with the notions of getting selective platoon advantages through frequent pitching changes.
The leverage thing is a real negative, as this rigid structure would mean that you’d be forced into using the same pitchers regardless of the score, so you couldn’t ensure that your best pitchers were regularly on the mound in the situations where preventing a run was more likely to impact the outcome of the game. Pitchers would have their set days of pitching and rest, and whether it was 1-0 or 15-0, you’d have to stick to the plan. However, the hope would be that this system would allow each pitcher to perform well enough that you wouldn’t hate having any of them on the mound in a close game. Still, there’s no question that the lack of ability to give better pitchers the more important innings is a drawback to this kind of structure.
The platoon advantage isn’t quite as big of a loss, I don’t think, and not because platoon advantages don’t matter, but because this system would essentially trade off when those platoon match-ups happened. Ideally, you’d want two right-handed pitchers and two left-handed pitchers with your four daily pitchers, which would make setting the line-up for the opposing manager a bit of a headache. If you’re starting a right-handed pitcher but everyone knows he’s only going to face nine batters, do you bother stacking the line-up with lefties? Probably not. At that point, knowing that the late game chess match has essentially been canceled, he might as well just start his nine best players regardless of what side of the plate they bat from. And while that means that you’re not getting those late game left-on-left match-ups, you’re also going to be facing fewer opposite-handed hitters earlier in the game. You’d almost certainly want to mix up the order in which the pitchers appeared so that a manager couldn’t save a few pinch-hitters knowing that LHP Pitcher X always works the last nine batters, but overall, I wouldn’t expect that you’d see a drastic change in the amount of times the team faced a platoon disadvantage.
The other significant negative associated with this strategy is the need for a 13th pitcher. As we’ve outlined, this strategy gets you through nine or maybe 10 innings, but has no real plan for anything in innings 11 and beyond. In reality, there will be games that go beyond 10 batters, and unless you’re just willing to blow up the last pitcher’s arm by forcing him to keep pitching until the game ends, you need another guy on the roster who can take the mound in extra innings. However, a 13 man pitching staff also creates problems for position players, as three reserves just isn’t enough.
So, my solution? Any team trying this should acquire Micah Owings, or someone of that ilk. The two-way position player could be essentially used as an emergency pitcher for extra innings games, throwing enough on the side to be ready in case they are called upon but also serving as a pinch-hitter/defensive replacement and general backup at the positions they can play in the field. The Brewers used Brooks Kieschnick in this way in 2003 and 2004, when he combined for 144 plate appearances and 97 innings pitched. There are a good number of players who both hit and pitch in college, so finding one could fill this 25th man supersub role shouldn’t be impossible — though you’re accepting a lower quality performer on both sides in order to get the flexibility. In reality, you’d probably lose more of your extra inning games than you would under a standard system, since innings 11-plus would essentially be taken up by a part-time pitcher who is probably not up to MLB standards as a normal pitcher.
Would those trade-offs negate the gains from allowing pitchers to only face a batter once through the order, amplify their stuff in shorter outings, and see a reduction in total pitches thrown by the best pitchers? For a team setup to function under the current system, perhaps. If you already have Cliff Lee, Roy Halladay, and Cole Hamels, this system doesn’t make any sense. You don’t really want those guys pitching less than they do now, since they can perform at an elite level for longer stints.
If a team were going to try this plan, though, not having to pay premium dollars for high-end starting pitching would actually be part of the appeal. By utilizing some decent-but-not-great starters in this way, the hope would be that a team could get higher-level production without having to pay the current market rates for starters who can hold batters to a .700 OPS over six-plus innings. If you built a pitching staff with lower-cost arms who carried equal amounts the workload, you could then allocate more of your overall budget to the position players, and hopefully build out an offense and defense that were better than you could afford if you were also trying to buy market value starting pitchers.
Of course, a team would have to essentially create their own financial compensation structure, since wins and saves would no longer be targeted rewards that a pitcher could strive for and use in arbitration hearings. And this would have to be something that the organization bought into across the board, as you’d need guys on this kind of schedule in the minors in order to come up and fill in when a pitcher hit the disabled list. It would essentially require a wholesale rejection of the current five-man-rotation system, and a long-term plan that required committing to this kind of experiment for more than a few weeks. The media scrutiny would be extremely high, especially whenever a pitcher was removed while pitching well and replaced by a guy who instantly imploded.
It’s not a plan without costs, but it is also a plan that I think could allow a team to improve their run prevention simply through better controlling usage patterns, and would come with the side benefit of allowing them to spend more money on position players rather than pursuing highly coveted pitchers (who would have no interest in pitching for your team anyway). Is anyone crazy enough to try it? Probably not, especially given how well pitchers are doing under the current system. We might need to see a shift back toward expanding offensive prowess before anyone tries something this different, as the “don’t fix what isn’t broken” mindset has a powerful hold. However, the Rockies abandonment of the five-man rotation shows that MLB teams are thinking about new ways in which to use their pitchers, and perhaps we’ll see a team try some variation of this before we die.
Dave is the Managing Editor of FanGraphs.