Coming out of the All-Star Game, Royals manager Ned Yost set his rotation to go Greinke -> Chen -> Bannister -> Davies -> Lerew. Had he said nothing about it, I think there would be little to question about his decision. If given the chance to set the rotation, I would probably have come up with the exact same one. Even if you want to argue another way to set up the pitchers, I doubt you would do so vociferously. The only absolute is that Greinke should be the #1 starter. Other than that: meh, no big deal. What did catch my eye was Ned Yost’s reasoning behind having Bannister in the 3rd spot. From Ned Yost, via The Kansas City Star:
“The spot that has the most day games is the three spot,” Yost said. “Four of the nine starts are day games, and Banny excels in day games, so why not (put him there)?”
Why not, indeed? It isn’t as if Bannister really belongs as the #1 starter and even if the 3rd spot didn’t get the most day games, he probably belongs there anyway. So I can’t really argue with his decision, however I think his reasoning is a little suspect.
On the surface, Ned Yost is correct. Brian Bannister has been better in day starts than in night starts. Prior to his start yesterday, he had a 2.37 ERA in day starts and a 7.45 ERA in night starts, which is a pretty extreme split for a pitcher to have. He had pitched 38 innings in the day and 64 at night, smallish sample sizes to be sure. So lets take a deeper look:
Bannister’s ERA by year and Career in Day and Night
|year||Day ERA||Night ERA|
Over Bannister’s career, the difference between the daytime and nighttime version is roughly the difference between 2008 Gil Meche and 2009 Gil Meche, which is significant. The interesting thing about the ERA stats is that other than his brief 2006 year, he has put up a better daytime ERA than nighttime ERA every single year. In three of those years, he was a lot better in the day than the night.
So in general, what Yost said and the basis of his assumption is true. Bannister has over the course of this season and his career pitched better in the day than in the night. Again, since he was probably going to put Bannister as the third starter regardless, it doesn’t matter in the end but I think this is an excellent time to try and sort out why statistical analysis is important in baseball.
Yost made a rotation decision upon some statistical knowledge he had. The facts he was using were correct, but was his use of them correct? That, is the crux of the issue. Don’t get me wrong, I don’t expect Ned Yost to do deep statistical studies on issues. I am sure he does not have the time to do that, which is why I can’t argue with his decision on the matter. He saw a statistical anomaly, figured he would run some more tests of it by having Bannister throw more times in the day to see if it is true. Nothing really lost by putting him at that slot in the rotation anyway.
But in the grand scheme of things, what Yost is attempting to do is make future predictions upon a statistical oddity. In order to truly do that, we need to go much, much deeper than ERA and then there is absolutely no guarantee that we would actually find out if he is truly better in the day anyway. Because in order to make the most accurate future predictions of players, we need to not only find statistical evidence, but we need to be able to determine a reason for the evidence. To me, that above all else is the disconnect between the statistically inclined baseball fans and the anti-statistics crowd. In most cases, we use statistics to locate something odd and then try and find out why it is happening. The former is easy, the latter ranges from difficult to impossible.
The first part of that, is to look at even more statistics to help uncover more evidence regarding the particular oddity. So let’s look at a few more stats for Brian Bannister. Day and Night Strikeouts per nine innings, strikeout to walk ratio and Batting Average on Balls In Play.
So Bannister strikes out more guys at night, but he also walks more guys. On top of that the batting average of balls put in play is higher too. So we have a little more evidence that he does somethings well in the day, like limiting hits and walking fewer guys, but he strikes out fewer guys in the day time. That is a very odd thing, because if there is one single thing a pitcher can do to lower his ERA, striking out more guys is that thing. So now, we have some evidence that Bannister is better during the day, but it is getting a little shakier. Lets move onto the next part, the proof.
So in order to reliably predict that Bannister is going to do better in the future when he pitches in the day, we need to connect the dots. We need to come up with some reason that the day is different for Bannister than the night. Here is where it gets tricky/impossible. I don’t have access to Brian Bannister 24/7 let alone for 10 minutes, so I can’t confirm or deny that he does something different to prepare for day games that makes him particularly better suited to pitch those games. It’s absolutely possible that he does, but he hasn’t indicated that he does and from most things I’ve heard him say he doesn’t do anything differently. Since the main difference between the day and night in most games is that for one, the sun is out and the other it is dark (duh, I know), what could possibly change from one to the other that would give Bannister the ability to walk fewer guys in the day than the night? Honestly, I can’t come up with any and I don’t really think anyone else could either.
So in the end, we have some statistical evidence that leads us to think something works for Bannister in the day that doesn’t work for him in the night. We have no real way of proving whether or not it is merely a statistical anomaly or if it is something which can predict future success for Bannister in his day starts. It’s almost a dead end of sorts. The only way we can find out much more is to keep running him out in the day and see if his stats don’t even out over time. Because the biggest fly in the ointment for any statistical analysis is that randomness is by nature random. It provides us with things which seem like they are correlated but have no actual correlation. Sorting out randomness from actual reasons is what separates a true statistical breakdown with merely showing someone some numbers. Its the fomer that can get hired into Major League Front Offices sometimes.
In the end, we may never know the truth about Bannister’s day/night splits. However, personally it’s these kinds of things that add an extra dimension to watching a game and makes it more enjoyable. For a major league team, I believe that proper statistical analysis can be the difference between being a .500 team and a playoff contender. I don’t know much about the Royals statisticians on staff, other than Dayton Moore says he has some. If they are any good, they would probably realize rather quickly that attempting to determine whether they should give Bannister more daytime starts is kind of a waste of time. There is no great statistical reason in my mind to choose one or the other, so just go with your gut, see if we can’t ride the wave of good daytime starts for Bannister and hope for the best. Which is exactly what Ned Yost did.
Yesterday was a hot and sunny daytime start. No better time to break out the Day-Banns and pile up some more statistical evidence one way or the other. Bannister went out and threw 6.1 IP, 5ER, 2k, 6BB and 2 HR. It’s only one game, so we can’t just claim that he isn’t a better daytime pitcher yet. However, he does have a 5.29 career ERA against the Athletics. Here’s hoping he gets more starts against the Detroit Tigers in the 2nd half because he has a 2.16 ERA against that team.