Looking for More Cy Young Separators

As I wrote yesterday, I have an NL Cy Young vote this year. It is a remarkably tough year to pick a winner, as there are three pitchers having award-worthy seasons; you can make a really good argument for any of the three. The reality is that those of us with votes are going to have to split the slimmest of hairs in order to sort out the top three spots on the ballot, and yesterday, I tackled the question of catcher influence.

In response, someone left this comment.

This is a great suggestion; when the overall influence of combined season numbers are this close, looking at the individual distribution absolutely could be a source of differentiation. And since I know at least one other puzzled Cy Young voter read yesterday’s piece, I figured it was worth exploring this idea, plus a couple others by the responses to the piece.

The easiest way to measure consistency within performance is to measure the standard deviation of a pitcher’s game score from their game logs at Baseball Reference. Game Score isn’t a perfect measure of pitcher performance — most notably, it makes zero attempt to account for the impact of defense — but it does give us a single measure of a pitcher’s results and puts everyone on the same scale, so for comparison purposes, it works okay for this idea.

To compare the game-to-game consistency of the three pitchers in the race, I took the game logs for each pitcher, dumped them into excel, and then ran the =StDev function on their game score columns. The results.

By average game score, there’s basically nothing to separate these three, as we’d expect from looking at their overall lines. However, Greinke has definitely been the most consistent of the three, with less variance between his outings. He’s been the epitome of a consistent ace, rolling out something like a 7 IP/1 ER performance every five days all season long.

However, before we can conclude that’s an edge for Greinke, we should note that a lower standard deviation isn’t definitively better. Matt Hunter explored this idea in some depth a few years ago, and his findings actually showed that more consistent pitchers were slightly less valuable than inconsistent pitchers. And part of the reason Greinke has posted a lower standard deviation is that he simply hasn’t had as many overwhelming dominating outings as either Kershaw or Arrieta.

To visualize this, Here are the distributions for each of the three pitchers, graphed.

You can see that Greinke’s starts are almost entirely bunched up in that 6-8 inning, 0-3 runs allowed window, with just two exceptions, where he gave up five runs once and six runs once. Both Kershaw and Arrieta have more bubbles to the right side, as they’ve had more games with higher runs allowed totals, but then, both have also thrown complete game shutouts, and complete games where they’ve allowed just one run. So Greinke’s consistency comes with a trade-off, as he has had fewer overall dominating starts, and has required his bullpen to finish out every game he’s started, which isn’t true of either Kershaw or Arrieta.

I hadn’t noticed Greinke’s lack of complete games until I started looking at the game logs, and in doing so, I also noticed that the Dodgers have had a much quicker hook with Greinke than they have had with Kershaw, or that the Cubs have had with Arrieta. For comparison, here are the total number of batters each pitcher has faced while going a fourth time through the batting order this season:

Kershaw: 40
Arrieta: 26
Greinke: 14

With only a few exceptions, the Dodgers have almost always removed Greinke from the game after he’s faced 27 batters, so his exposure to hitters in situations where they’d be most likely to perform well against him has been significantly limited compared to his peers. And in the 14 at-bats where Greinke has faced hitters for a fourth time within the same game, they’ve hit .308/.357/.538 against him, which is likely what has prompted Mattingly to use a quicker hook. Kershaw (.184/.225/.316) and Arrieta (.115/.115/.154) have been excellent even when facing hitters a fourth time, which is why they’ve thrown complete games, and Greinke hasn’t.

And this is one of the areas where Greinke’s low ERA is probably a bit deceptive, because despite the fact that he’s struggled a bit when the Dodgers have pushed him to try and give the bullpen a night off, the only one of the five baserunners he allowed in those 14 plate appearances scored when they hit a home run off Greinke. The other four? They were all stranded by Dodger relievers taking over after he allowed a rally to begin.

In fact, as another commenter noted yesterday, Greinke and Arrieta have both received much more help from their bullpens than Kershaw has. Number of inherited runners, and how many have scored, for each pitcher:

Greinke: 7 inherited runners, 1 scored
Arrieta: 8 inherited runners, 2 scored
Kershaw: 9 inherited runners, 6 scored

On average, 29% of inherited runners in the National League have scored this year, so relievers have shaved one run off Greinke’s total, while costing Kershaw an extra 3.5 runs compared to what he would have allowed had he gotten average runner stranding after he left the game. When a race is this tight, and much of Greinke’s case rests on his advantage in runs allowed, this actually looks like a pretty interesting fact to consider. And this is yet another reason why simply looking at ERA and calling it a day is not good enough, especially when trying to decide on the winner of a postseason award.

In the end, I don’t know that any of these small things really helps us decide who the winner should be either. Greinke has been more consistent, but there’s not actually a lot of clear evidence that consistency is a huge benefit, especially because part of the reason for that consistency has been a reduced workload late in games. Both Kershaw and Arrieta have been tasked with pitching more regularly with the times-through-the-order penalty working against them, and have succeeded in spite of being put in a more difficult position to succeed, while Greinke has been more often rescued by his relievers when he started to get into trouble in late-game situations.

When you look at a measure like RE24, which counts expected runs against a pitcher for leaving baserunners on for another pitcher to inherit, Greinke’s still in first place, but by a smaller margin over Arrieta than if you look at overall runs allowed, and Kershaw is closer to Greinke than ERA would suggest as well.

I don’t know that I’m any closer to figuring out who to vote for, but walking through these processes has definitely helped clarify some things. If you value consistency, bump Greinke up a notch or so. If you value pitching deep into games, bump him down relative to the other two. And if you’re going to just quote ERA and call it a day, at least remember to adjust for the fact that Kershaw has been screwed over by Dodger relievers. These are the hairs that are worth splitting in a race this close.