Great stuff. However, using the Bill James study about power pitchers being more effective at night would have no scientific correlation to performance of non-power pitchers at night or in the day. I have always believed that home plate umpires have a much bigger strike zone in the day and tighter at night. Guess they can see the corners better in the sunlight. I love pin-point control guys throwing in a day game
I sorta want to give you a hard time because you piggybacked off of Joe Posnanski for a second straight post. But this blog and his blog are two sites I'm always checking for updates, so it's nice when they compliment each other. Plus, when I saw the day/night splits on Poz's Banny Log, my first thought was, "I wonder what Rany thinks about this."
May 12, 2008 at 2:53 PM
Anonymous said...
I wonder if it has anything to do with "get away days"? I haven't checked his other starts, whether they have been on get away days or not. Just a thought...
This raises an interesting issue. Are long term statistics less meaningful when they mainly reflect large, sharp changes in performance than when they are relatively steady? What I mean is this: consider two pitchers, A and B both with 4.00 ERAs over 200 innings. Now suppose player A has a 4.25 ERA for the first 60 innings, a 3.625 ERA for the next 80 innings, then back to 4.25 for the final 60 innings. Meanwhile, player B has a 4.25 ERA for the first 90 innings, a 1.75 ERA for 20 innings, then a 4.25 ERA for the remaining 90 innings.
Both players have posted identical numbers over an identical sample, but the shape of their performance is different. How, if at all, does what we should expect going forward depend on that shape? How does what we should expect from Bannister depend on the fact that his great daytime performances and poor nighttime performances have all come in a string this year, instead of being spread out over his career?
The problem with large sample sizes is that they are made up of smaller sample sizes. It isn't obvious to me why career stats that include this season are less convincing than those that do not.
I hate to speak for Rany, but I think the point is that data can be stratified to identify if short-term changes are reflective of the larger sample.
Basically, if these 8 starts are a real trend, it's a new one.
May 12, 2008 at 4:03 PM
Anonymous said...
Interesting comment posted about getaway days.
Certainly, a significantly larger percentage of day games are on getaway days than night games would be.
Perhaps umpires have larger strike zones, and players less disciplined approaches on getaway days? After all, that would only be human nature at work, even on a subconcious level, knowing one had a long trip ahead, wanting to get the game over more quickly.
Everyone's missing the boat. Banny's eyelids are obviously jammed during night games. Getaway days, sample sizes, Bill James and sunlight have nothing to do with Banny's nightmares.
Conclusion: He needs something with lace, something sexy to wear under the pajamas. That'll get his mind on the game and his eyelids will breathe freely again. Also, if Rany's piggy-backing, what's Poznonkey doing? A READER tipped him on this. Plus, what the heck are we doing? Answer: Piggy-backing on Rany everyday by commenting, so lay off.
My question goes to the validity of this stratification. What is the relationship between the time-ordering of points in the sample and the appropriateness of including them or leaving them out of the sample. I'm fairly sure we wouldn't remove 8 starts in the middle of last year from the sample, so why remove 8 starts at the beginning of this year?
The first thing I did when I read Posnanski's post was to calculate the ERAs during the day and at night, excluding this year, exactly as Rany did. It certainly seems natural to remove this years anomalous stats in pursuit of an overall mean. But I'm wondering exactly what circumstances make such stratification of the data valid for future predictions. After all, discounting all anomalies would yield boring and wrong results, so some procedure is needed to determine which anomalies to keep and which ones to throw out. I don't know if this has been done yet in sabermetrics.
May 12, 2008 at 10:36 PM
Anonymous said...
Nathan,
Joe's point was that this was a trend that was true of Bannister's whole career. Which is true, but it is only true of his whole career because it is true of the last eight starts.
I think if you could spot an anomaly in his record, and then you could find a stretch of six weeks within his career that explained it almost entirely, the question I would raise is, what is weird about that guy's career, but what was weird about those six weeks.
So it could go either way - maybe this becomes a persistent thing that affects him for the rest of his career, and this six weeks was the harbinger of it. Or it could just be a weird six weeks.
May 13, 2008 at 9:04 AM
Anonymous said...
I think it has something to do with Jessica Simpson.
May 13, 2008 at 3:53 PM
Posnanski wrote his usual wrap-up of Bannister's start yesterday, and updated his log to point out this rather unusual split in Banny's performance: Wow, as someone who claims to be one of the nation’s leading Brian Bannister scholars, I am embarrassed to say I missed this. But I did. It is brilliant reader PC who points this out: Brian Bannister by day (this year): – 4-0, 0.62 ERA, 29 ip, 12 hits, 3 runs, 2 earned runs, 0 homers, 7 walks, 18 K. Batting average against: .126; OPS: .320; OPS+ -10(!)*, Babip: .156. Brian Bannister by night (this year): – 0-4, 8.02 ERA, 21 1/2 ip, 33 hits, 19 runs, 19 earned, 5 homers, 5 walks, 11 K, Batting average against: .344; OPS .960; OPS+ 158(!), Babip: .350....Now, you can say: Well, sure, but that’s nothing, it’s a small sample size, it’s a fluke. Maybe. But as PC continues to point out … here are his career numbers: Brian Bannister by day (career) – 10-1, 2.65 ERA, 88 1/3 ip, 67 hits, 28 runs, 26 er, 4 homers, 23 walks, 38 Ks. Batting average against: .212; OPS: .576; Babip .226. Brian Bannister by night (career) – 8-13, 4.58 ERA, 165 ip; 168 hits, 88 runs, 84 er, 20 homers, 55 walks, 87 Ks. Batting average against: .261; OPS .763; Babip: .275. Again, we’re not looking at a big sample by any means. But I think it’s big enough to say: “Wow, that’s kinda weird.” For whatever reason, it seems like Brian’s stuff is just much harder to hit during the day.
Those are pretty dramatic splits, certainly. But I think it's dangerous to read too much into them. Yes, this year Bannister has been much, much, MUCH better in the daytime than under the lights. But you can't say it's not a fluke simply because his career numbers also reveal a split - because those career numbers include this season.
For his career, Bannister has a 4.58 ERA at night, a 2.65 ERA during the day. But if you strip out 2008, his ERA at night is 4.07; during the day it's 3.64. A difference, but a small difference. And it's a difference which overstates the mark, if anything. Here are Bannister's career totals at night and during the day, prior to 2008:
His hits per nine innings during the day (8.34) is a fraction better than his ratio at night (8.46). His strikeout-to-walk ratio is significantly better at night than during the day. His BABIP - this is approximate - is .256 during the day, .266 at night.
Prior to 2008, there was no conclusive evidence to speak of that Bannister was a better pitcher during the day than at night. Do eight starts change that perception? If his start in Arlington, with 30 mph winds gusting to right field, when Bannister said he felt like he was pitching on the moon, had happened to occur in the day, how much would that skew these numbers?
I do think that Bannister is probably a little more effective during the day, in part because Bill James ran a study about 20 years ago - I think it was the 1987 or 1988 Abstract - which showed that power pitchers appear to be signficantly more effective at night. Bannister is basically the opposite of a power pitcher, so it would stand to reason that he might be more effective during the day.
But I don't think the difference is great enough to influence how the Royals use him. For one thing, he's not the only Royals' starter to pitch better during the day. Gil Meche, in a career of over 1000 innings, has a daytime ERA (3.78) more than a point lower than his nighttime ERA (4.84). Now that is significant. I suspect that most pitchers pitch better during the day, because day games tend to be clustered in the colder months of the season, and colder weather tends to lower offense. (Just a theory I have. Which means I'm probably wrong.)
If the Royals have a doubleheader and choose to start Bannister in the day game and someone like Greinke (whose ERA is 15 points under the lights) in the nightcap, great. But let's not go overboard with moving starters around to take advantage of an effect that might not actually exist.
"Bannister in the Daytime."
10 Comments -
Great stuff. However, using the Bill James study about power pitchers being more effective at night would have no scientific correlation to performance of non-power pitchers at night or in the day. I have always believed that home plate umpires have a much bigger strike zone in the day and tighter at night. Guess they can see the corners better in the sunlight. I love pin-point control guys throwing in a day game
May 12, 2008 at 2:23 PM
I sorta want to give you a hard time because you piggybacked off of Joe Posnanski for a second straight post. But this blog and his blog are two sites I'm always checking for updates, so it's nice when they compliment each other. Plus, when I saw the day/night splits on Poz's Banny Log, my first thought was, "I wonder what Rany thinks about this."
May 12, 2008 at 2:53 PM
I wonder if it has anything to do with "get away days"? I haven't checked his other starts, whether they have been on get away days or not. Just a thought...
May 12, 2008 at 2:58 PM
This raises an interesting issue. Are long term statistics less meaningful when they mainly reflect large, sharp changes in performance than when they are relatively steady? What I mean is this: consider two pitchers, A and B both with 4.00 ERAs over 200 innings. Now suppose player A has a 4.25 ERA for the first 60 innings, a 3.625 ERA for the next 80 innings, then back to 4.25 for the final 60 innings. Meanwhile, player B has a 4.25 ERA for the first 90 innings, a 1.75 ERA for 20 innings, then a 4.25 ERA for the remaining 90 innings.
Both players have posted identical numbers over an identical sample, but the shape of their performance is different. How, if at all, does what we should expect going forward depend on that shape? How does what we should expect from Bannister depend on the fact that his great daytime performances and poor nighttime performances have all come in a string this year, instead of being spread out over his career?
The problem with large sample sizes is that they are made up of smaller sample sizes. It isn't obvious to me why career stats that include this season are less convincing than those that do not.
May 12, 2008 at 3:14 PM
I hate to speak for Rany, but I think the point is that data can be stratified to identify if short-term changes are reflective of the larger sample.
Basically, if these 8 starts are a real trend, it's a new one.
May 12, 2008 at 4:03 PM
Interesting comment posted about getaway days.
Certainly, a significantly larger percentage of day games are on getaway days than night games would be.
Perhaps umpires have larger strike zones, and players less disciplined approaches on getaway days? After all, that would only be human nature at work, even on a subconcious level, knowing one had a long trip ahead, wanting to get the game over more quickly.
May 12, 2008 at 5:42 PM
Everyone's missing the boat. Banny's eyelids are obviously jammed during night games. Getaway days, sample sizes, Bill James and sunlight have nothing to do with Banny's nightmares.
Conclusion:
He needs something with lace, something sexy to wear under the pajamas. That'll get his mind on the game and his eyelids will breathe freely again. Also, if Rany's piggy-backing, what's Poznonkey doing? A READER tipped him on this. Plus, what the heck are we doing? Answer: Piggy-backing on Rany everyday by commenting, so lay off.
May 12, 2008 at 9:06 PM
Bart,
My question goes to the validity of this stratification. What is the relationship between the time-ordering of points in the sample and the appropriateness of including them or leaving them out of the sample. I'm fairly sure we wouldn't remove 8 starts in the middle of last year from the sample, so why remove 8 starts at the beginning of this year?
The first thing I did when I read Posnanski's post was to calculate the ERAs during the day and at night, excluding this year, exactly as Rany did. It certainly seems natural to remove this years anomalous stats in pursuit of an overall mean. But I'm wondering exactly what circumstances make such stratification of the data valid for future predictions. After all, discounting all anomalies would yield boring and wrong results, so some procedure is needed to determine which anomalies to keep and which ones to throw out. I don't know if this has been done yet in sabermetrics.
May 12, 2008 at 10:36 PM
Nathan,
Joe's point was that this was a trend that was true of Bannister's whole career. Which is true, but it is only true of his whole career because it is true of the last eight starts.
I think if you could spot an anomaly in his record, and then you could find a stretch of six weeks within his career that explained it almost entirely, the question I would raise is, what is weird about that guy's career, but what was weird about those six weeks.
So it could go either way - maybe this becomes a persistent thing that affects him for the rest of his career, and this six weeks was the harbinger of it. Or it could just be a weird six weeks.
May 13, 2008 at 9:04 AM
I think it has something to do with Jessica Simpson.
May 13, 2008 at 3:53 PM