Bill James began a recent article like this: "It is entirely possible and entirely practical, in modern baseball, for a team to use a three-man starting rotation. I realize that this is probably not going to happen, but ... this is how it could work, and this is why it would work."
Check out the full article at Bill James Online, but to summarize: The workload of starters has been in decline for 40 years, resulting from the change from four-man rotations to five-man rotations in the 1970s, then to the imposition of pitch limits, then to even more limits on pitch counts. This was done to prevent injuries -- as James notes, "it may have made starting pitchers more effective, but that's an after-the-fact argument" -- but James' belief is these changes haven't reduced injuries.
Pitcher injuries have not decreased because we're pursuing a chimera, a shibboleth, a mirage. The way we use starting pitchers now is NOT, in fact, an effective way to keep them healthy and in rotation. That is what I believe.
But I am not arguing here for a return to pitchers pitching 300 innings in a season. I'm arguing for a change in pattern that would put the top limit at about 245 innings in a season -- actually not that far from where it is now.
But here's the thing: starting pitchers would be pitching a few more innings, but in a pattern that reduces the stress load associated with those pitches.
With limits of five innings or 80 pitches, a starting pitcher could have a theoretical workload of around 52 starts and 245 innings, with the emphasis being that the pitcher is throwing fewer pitches when fatigued, since he's capped at 80 pitches per start.
As James points out, he's not the first one to suggest the idea. Other ideas have included just using three pitchers for three innings at a time (while mixing in a couple ace relievers as needed). The argument against that idea is that starting pitchers would hate it: They would never get a win. With James' idea, while they would often be removed before completing five innings and becoming eligible for the win, they pitch five innings often enough and since the healthy ones would start 50 times a year, heck, they could win 25 games or 30 games, maybe more on a great team.
So could a three-man rotation work? Two main issues:
(1) Would it tax your bullpen too much?
(2) How would pitchers respond working on two days of rest after throwing 70 or 80 pitches?
James says an eight-man bullpen would be enough, meaning each reliever would average 88 innings, assuming the starters average about 4.5 innings per start. "That doesn’t strike me as an extraordinary number," he writes. Of course, only a couple relievers a year pitch that many innings now, and the mediocre relievers who pitch 50 or 60 innings a year might be much worse if they're pitching an extra 30 innings a year. A nine-man bullpen would have to average 78 innings per reliever. Keep in mind that you'd have some displaced starters in the bullpen, guys used to throwing more innings, so not every reliever is throwing that many innings. Keep in mind as well that starters moved to the bullpen do better, so a current mediocre starter might end up being a very good reliever (see Wade Davis).
The second issue is the great unknown, although it gets back to James' point about injuries: Guys used to start on three days' rest all the time while routinely throwing more than 100 pitches per game, so maybe they'd stay healthier throwing fewer pitches per game, even with less time off between starts. But is two days of rest enough recovery time? The average pitch count per game in 2015 was 93 pitches. As reader Guy123 posted in the reader comments on the article, "Stated another way, the notion here is that if you reduce starters' pitch count by 14%, you can increase their frequency of starts by about 65% -- with no loss of effectiveness. I'm sorry, but that just seems wildly improbable."
One thing Bill doesn't address: It's possible that pitching injuries have been reduced. For example, how many pitchers got hurt in the minors back in the day when 19-year-olds would pitch 200-plus innings a year in Class A? And considering pitchers throw harder now, maybe there would be more injuries if they carried the same workloads as in the 1970s.
One more note: Eighty pitches over five innings is 16 pitches per inning. In 2015, 133 pitchers pitched at least 100 innings as a starter; 66 of them averaged 16.0 or fewer pitches per inning. I think that's where the idea starts to get a little problematic; you'd absolutely need your starters to be efficient and reach that five-inning mark as often as possible.
Ultimately though, I think the idea could work. (Maybe for a team like the Diamondbacks.) You're giving fewer innings to your fourth and fifth starters, who might end up being more effective working fewer innings as relievers. Yes, pulling Clayton Kershaw with a 1-0 lead after five innings would be tough, but he'd be starting again in three days instead of five.
We're not going to see any team try it anytime soon. If anything, we're more likely to first see a team go with two or three traditional starters and "tandem" pitchers in the fourth and fifth spots going three innings apiece before we see a three-man rotation. The game usually evolves in small ways. Is it ready for a big change?