Why isn't every starting pitcher turning into Alex Wood?

If there's a poster child for the short-starts strategy, it's Alex Wood, the Los Angeles Dodgers' surprise star.

At 14-2, he has the National League's fourth-best ERA among starters (minimum 100 innings), its third-best WHIP, its fifth-best FIP. He also has never thrown more than 100 pitches in a start, and in 13 of his 21 starts, he has been pulled before his 90th pitch, even in five cases while throwing a shutout.

Those two sentences share a paragraph because they're presumed to have some causal relationship. Because Wood doesn't work deep into games, he's not seeing his numbers ruined by extra pitches thrown exhaustedly at the end of starts. (Whether it's fatigue or familiarity or a combination of the two, starters are considerably worse the third time they face a batter in a game than they were the first two.) And because he knows he's not going to be asked to throw 115 pitches, he can channel more of his energy into the 90 pitches he actually throws. Finally, from the Dodgers' perspective, they get to replace Wood with relievers, who as a group have lower ERAs than starters, since they get to come in fresh and face hitters who haven't seen them before. It's brilliant!

Except for one problem: There are only so many good relievers on a team, and the more innings you ask of them the shallower the pool gets. Conveniently for the narrative purposes of this article, this has been a problem for the Dodgers in Wood's starts: His relievers have had a 4.30 ERA after he has left the game. That's worse than the 3.16 ERA they have overall, and it might even be worse than Wood would be if he stayed in for an extra 20 pitches per game. On the other hand, maybe it's worth replacing a slightly better pitcher (Tired Alex Wood) with a slightly worse one (Fifth Reliever On The Depth Chart) if it's part of a larger role shift (90 pitches instead of 115) that helps turn a pretty good pitcher (Alex Wood) into an ace (New Alex Wood). That's the calculus. It's complicated and involves a lot of unknowns.

But here's the weird thing. Yes, Wood's starts are emblematic of a strategy that the league as a whole has embraced. Yes, this strategy has greatly affected the division of labor among big league pitchers, particularly as the trajectory of these changes has steepened in the past three years. But no, the consequences we laid out for the Dodgers in Woods' 21 starts have not been universal. And that's what I can't figure out.

So let's go through those three facts.

1. On Saturday, another Dodgers starting pitcher, Brock Stewart, was lifted from a start after four innings against the Padres. He'd faced exactly 18 batters. It was the 94th start this season in which a pitcher faced that many, and exactly that many, batters -- a new record, topping last year's mark of 93.

Now, in a sport with a 2,430-game season, 94 isn't that many, but it's still worth noticing. (We're now up to 97 such starts and counting.) As recently as 2014, there were only 65 starts of exactly 18 batters, and if you were to graph out how many outings lasted 12 batters, 13 batters, 14 batters and so on, you wouldn't see much of a spike at 18. In fact, here's that graph, which has three things we should notice:

The first is the aforementioned lack of a spike for 18-batter games in 2014 (the red line). In 2017 there is a spike, followed by a flat line at 19, suggesting managers might be actively targeting "18" as a place to pull pitchers -- as a way to avoid the third-time-through-the-order penalty.

The second thing to notice is that there is also a spike at 27 in the blue (2017) line. In 2014, 27-batter starts were the most common type of start, but they were just the peak in a fairly smooth slope up/slope down progression. But in 2017, managers disrupt the slope to pull pitchers after batter 27, again suggesting they are targeting a times-through-the-order moment rather than simply a pitch count or innings count.

The third thing to notice is, just since 2014, starts have gotten much shorter. Every length of start up to 25 batters is more common now than it was then; every length of start from 26 on is less common. More than half of all starts went 26 or more batters in 2014; only 37 percent do now.

2. And so if starting pitchers are throwing fewer innings, guess who must be throwing more?

In 2014, relievers as a group threw almost exactly half as many innings as starting pitchers did. That figure fluctuated a little year by year, but it had been close to steady for about two decades.

Then, in 2015, when the Rays got attention for embracing the 18-batter start, and when teams began to aggressively replace starting pitchers in games they were winning, the workload began to shift. In 2015, relievers pitched 54 percent as many innings as starters; in 2016, 58 percent. This year, it's a little over 60 percent. Relievers are on pace to throw almost 1,500 more innings this year than they did in 2014, blowing past the all-time record (which was set last year).

These innings aren't coming from the bullpen aces; those guys were already maxed out. Instead, they're coming from the bottom of the bullpen, or from pitchers who wouldn't have been in the majors at all under the (slightly) older model. To simplify things, those extra 1,500 relief innings mean every team is trying to get 50 innings out of a guy who "should" be in Double- or Triple-A, by 2014 standards.

So we have two variables: Starters, as a group, get to come out of games earlier, protecting them from seeing batters a third or fourth time and letting them put more energy into the pitches they do throw. And relievers, as a group, now include a bunch of scrubs stinking up the stats. The statistical gap between starters and relievers should be shrinking.

3. The gap between starters and relievers isn't shrinking!

Starters this year have allowed a .769 OPS, which is 6 percent higher than the .725 OPS relievers have allowed. Far from closing the gap, starters are actually falling slightly further behind:

Baseball, unfortunately, isn't a wind tunnel. A lot of things can affect the run-scoring environment, from weather to scouting technology to defensive strategies to the seams of the ball, so we can't say starters are getting worse or relievers are getting better. We can conclude only that relievers are getting better relative to starters. And so I'll say again: This is weird!

My hypothesis -- to be tested another day, perhaps -- is that it's the result of two decades without expansion. Strained depth traditionally affects pitchers more than hitters, and there are now simply too many qualified pitchers to fill 30 teams' bullpens. But there are other hypotheses. Maybe teams are now funneling better pitchers into relief, instead of giving every good pitcher every chance to become a starter. Maybe the "juiced" ball hurts starters disproportionately, because starters are temperamentally more likely to allow balls to be put in play.

Regardless of the reason, this weirdness matters, especially as October looms and managers make their in-game decisions with more urgency. It used to be managers wanted to get as many innings as they could out of their starters. Now, for all but the very best starters, managers want to get as many innings out of their relievers as they can -- the proverbial "shortening the game."

As long as good relief pitching is finite, this longing for more relief innings is thwarted. But the Dodgers have seven relievers (minimum 20 innings) with an ERA+ better than 130 this year. The Indians and Yankees have six each, and that doesn't even include Aroldis Chapman. At a certain point, the number of effective relievers teams have is going to be limited only by the size of the roster.