|Millwood Should Regress in Second Half|
Written by Justin Zeth (Contact & Archive) on July 05, 2009
Evan Grant ripped Rob Neyer a new one over Neyer's heretical assertion that Rangers pitcher Kevin Millwood's 2.80 ERA is more the result of exceptionally good luck than any particular skill. Short version of Neyer's point: Millwood's ERA is much lower than his K/BB/HR rates would suggest, because an unusually low percentage of guys he's put on base have scored, and decades' worth of statistical evidence suggest pitchers can't really control what percentage of their baserunners score. Some guys have a higher baseline for that than others (because some guys stink at pitching out of the stretch and some guys are better pitchers than others in the first place) but 1. Kevin Millwood isn't one of them, and 2. Even so, that percentage vacillates from year to year and month to month in an apparently random manner.
That really pisses off a lot of people around baseball, because, you see, driving in runners is the manliest thing you can do in baseball. And for a pitcher, preventing runners from getting driven in is how you prove how awesome you are. And further, it makes perfect sense on an intuitive level. I know I'm impressed when somebody hits a double to drive in the runners on third and first, and that obviously has a huge impact on the game. I'm very impressed when there are runners on second and third and one out, and the pitcher gets a popup and a harmless groundball (in that order) to get out of it, and again, obviously that's crucial. The statistical analysts -- or stats nerds, if you will -- will of course agree that those are very important things, because after all they're not total idiots. It's just that those are not predictable or repeatable things for any one particular hitter or pitcher, beyond the normal percentages suggested by his underlying stats (AVG/SLG for hitters, K/BB/groundball rate for pitchers).
That was explanation, just in case you weren't familiar with the issue -- explanation of the basic tenets of the ongoing philosophical war in baseball. The traditionalists* just get driven nuts by the stats guys saying RBI aren't that big a deal... well, RBI are a big deal, but 1. No bigger a deal than runs scored, which the traditionalists dramatically underrate, and 2. RBIs are not as big a deal as AVG/OBP/SLG (especially the latter two), because RBI is basically AVG/OBP/SLG times runners on base (oversimplifying, but you get the idea). That makes AVG/OBP/SLG much more useful in evaluating hitters, since it's not subject to what his teammates do, as RBI are (you can't drive in a runner when he ain't there, Junior**).
* When did 'traditionalist' become pejorative?
** Unless you're using ghost runners, which unfortunately aren't provided for by the MLB rulebook. I know I personally drove in approximately 36,480 ghost runners during my illustrious career. Eat your heart out, Jim Rice.
And for the pitchers, same thing. Most people around baseball (including the writers, who after all are the link between the players/managers and the public) are not about to sit and listen to somebody tell them that, yes, preventing runners from scoring is good and important, but it's not, you know... important, because once again it's just a function of more foundational stats. (Groundball pitchers generally allow fewer runners to score than flyball pitchers, for obvious reasons.) Pitchers can and do outperform that all the time, and underperform it, for a month, a half, a year, two years.
I just want to cherry pick a few quotes from the linked article to move this along...
Yep, he's prohibited too many runners who have reached base from scoring, thus underscoring why Millwood has been so lucky and not good. And here I am thinking that not allowing runners to score is actually an attribute for a pitcher
That basically sums it up. Evan Grant is completely (and I assume willfully) missing Rob Neyer's point. I don't think Neyer ever argued that not allowing runners to score is a (good, presumably) attribute for a pitcher. Neyer's saying -- mostly correctly -- that it's not really a pitching attribute/skill unto itself; rather, it's a secondary function of two primary functions (pitching skill, as measured by K/BB and ground ball/HR ratios, and runners already on base).
Put a different way: .333 isn't far off from the league average OBP (a little higher, I think). Take a die and roll it repeatedly, and see how often you get a one or a two, which we'll use to represent a guy reaching base (three through six means he made an out). Sometimes you'll get five and six ones or twos in a row; that's a big inning. Sometimes you'll go long stretches without ever getting three of them in five rolls; that's an inning where the pitcher let a runner or two reach base but didn't allow any runs. You can roll the die 500 times and you expect to roll a one or a two 167 times. Sometimes you'll actually roll them 210 times. Sometimes you'll only roll them 140 times. More often you'll get fairly close to the 167 baseline.
Following me? But here's the thing -- that's just step one. That just measures how many runners reached base versus expected. A pitcher can be 'lucky' (I don't really know that 'lucky' is the correct term, though Neyer uses it all the time) simply by allowing fewer hits than his baseline ratios establish he should. But what we're talking about -- the percentage of guys already on base that end up scoring -- is the second level. To figure that you have to see how many of those 153 ones and twos out of 500 rolls were clustered together. And that will vary up and down even more than your baserunner rolls did.
Well, that's basically what Rob Neyer and many others like him are asserting -- we can look at a pitcher's stats and establish how many guys that have already reached base we expect will score a run, and anything above ('lucky') or below ('unlucky') the expectation is the result of randomness. That does not make it unimportant; just unpredictable and therefore not useful for measuring our future expectations. (That's my assertion, not necessarily Neyer's or anyone else's.) It's wonderful that Millwood's ERA is so low, but based on his statistical record, you should not expect it to stay that low. In fact, his ERA, currently 2.80, is very likely to finish the year at least above 3.20, assuming Millwood pitches the whole year. I would lay five to one on it, and I strongly suspect Evan Grant would not take me up on the proposition, because Kevin Millwood is not fundamentally, has never in his life been fundamentally, a 2.80 ERA pitcher. We all basically understand that. Tom Glavine in his prime was fundamentally a 2.80 ERA pitcher.
Right now, after (a league leading, I should add) 119 innings pitched, his ERA happens to be 2.80, and it happens to be because an exceptionally low percentage of his baserunners have scored. We expect that to normalize. Neyer is right about that. Once in a while it won't immediately normalize -- he'll maintain that low percentage all year, and in fact Millwood has done it before; he led the AL with a 2.86 ERA in 2005, and he wasn't that good a pitcher then, either -- but if it doesn't this year, it will next year.
One more thing before I finally let the four of you that are still reading go. There is a line of thought in some corners of the statistical community that goes like this: Kevin Millwood is not really a 2.80 ERA pitcher. We all know that. His fundamental stats are really not very good: 5.6 K/9, 2.9 BB/9, 1.1 HR/9. He's fundamentally an average pitcher. Therefore he has no business being on the All-Star team, which should be reserved for pitchers who have posted the best basic ratios regardless of ERA, which is largely subject to the whims of team defense and luck (which is true).
I, for one, think that's bunk. ERA is not an especially useful stat for predictive purposes; what ERA does, pretty well, is record what actually happened. What has actually happened in the first half of the 2009 season is Kevin Millwood has allowed very few runs to score. A pitcher's job is to not allow runs to score; Millwood has in that sense been one of the best pitchers in the game for these three months. He's not likely to sustain it in the second half -- in fact, he's extremely unlikely to -- but that in no way erases what he's already done.