The Inconsistent One
The Dodgers spent a good bit of money on Hiroki Kuroda to bolster their rotation this winter, and overall, he’s pitched pretty well. His 3.64 FIP over 156 innings has resulted in a 1.36 WPA/LI, meaning he’s been nearly a win and a half better than an average pitcher. That’s pretty valuable, and the Dodgers couldn’t have hoped for much more when they signed him.
However, on a day to day basis, it’s been nearly impossible to predict what Kuroda will give you. He’s taken inconsistency to a whole new level. Take a look at this histogram of his starts by game score, showing the frequency of each type of start he’s had this year (using intervals of five).
Kuroda’s average game score for the season is 52 (a bit above average, as GS is set to be average at 50), but as you can see, the distribution is nothing like a bell curve. It’s more of a shipping barge. Or maybe a Dodger Dog with some weird toppings. The reason for that, of course, is that Kuroda has been sometimes miserable (three starts below 20) and sometimes awesome (one start at 90 and one at 91), and then just about every interval in between. He’s been horrible. He’s been bad. He’s been mediocre. He’s been okay. He’s been solid. He’s been good. He’s been excellent. He’s been great.
What he has not been is consistent. In total, the package has been quite valuable, but in isolation, you really have no idea what you’re going to get from Kuroda when he takes the mound. You want him on your team, but if the Dodgers do end up making the playoffs, I’m sure a few fans in LA are going to be holding their breaths when he takes the mound. He hasn’t exactly inspired a confidant expectation of results so far.
Print This Post
Dave, what’s typical in terms of game score variance? Perhaps more than any other statistical claim, the one regarding consistency seems to be thrown about most liberally with very little context. Yes Kuroda appears to have a good deal of variance. But compared to what and by what measure? What’s the distribution of game score consistency measures look like?
So does this mean Joe Morgan has a new arch-nemesis?
It would take a big query on a database like the one Sean Forman is running at B-R to answer the question exactly, but I’m fairly certain that Kuroda’s variance is much higher than usual. I looked at several pitchers (the original idea for this post was to compare Oliver Perez, Kyle Lohse, and Kuroda, all of whom have been up and down this year) and Kuroda was by far the most inconsistent. The standard deviation of his game scores was 22, compared to 15 for Lohse and 16 for Perez. I’d guess (and this is just a guess) that the average stdev would be in that 14-17 range, with Kuroda’s 22 being a pretty big outlier.
If the SD was 15, Kuroda would be at -1.5 SD, not exactly a big outlier. We’d expect something like 1 in 15 or 1 in 20 SP to be that inconsistent, right? (don’t have the tables in front of me) Though I suppose given the number of regular SP, that’s something like 6 to 8 total in MLB…
To compare, some of the better (more consistent?) pitchers out there:
Halladay = StDev 13.8
Lee = 16.9
Webb = 18.2
Harden = 14.3
Sabathia = 22.4 (two terrible starts early)
I’m curious, how would you guys recommend going about calculating a variance for hitters? What unit would you use — PA/game/series/week/month? Are you aware of any serious consistency analyses already completed?
Well, if we were looking for hitter variance in a similar timeframe to a pitcher’s per-start variance, we’d want to look at about 3% of the batters season totals, or about 18 plate appearances . That’s about four games worth, so a series is a bit too short and a week is a bit too long.
I think The Book had a chapter on consistency, but I don’t have my copy in front of me right now. I’ll check when I get home.
This is a prime example of “who needs pitching” and why hitting is (mostly) where it’s all at.