More on regression, xFIP, HR/FB, BABIP, and the like
There’s been some good discussion going on today that I thought I’d bring to your attention. Chris Liss mentioned my post from a couple weeks ago on Dan Haren, luck, and randomness in a column at RotoWire, and some interesting debate sprang up in the comments. As usual, I ended up writing a book for my comment, so I posted the majority of it over at the CardRunners site as well. There’s been some comments on both sites since then, so you might be interested in checking the discussion out.
Scott Pianowski’s comment about thinking of the top 15 pitchers is a flawed way of looking at things. Surely, ERA plays a large part in his opinion of who the best pitchers are. As a result, he will likely include many (and he said 11 of the 15) that will have suppressed HR/FB ratios that fool him into believing they are top 15 pitchers. Unfortunately, these pitchers may not have nearly enough innings to prove this is actually a skill and not just dumb luck.
Tommy Hanson is a perfect example of this. Career xFIP of 4.02, but just a 3.17 ERA, largely due to a career 5.6% HR/FB ratio. One could easily include him in a top 15 pitcher list, supported by his excellent ERA. However, he has only pitched 275.2 innings in the Majors, which is far from enough for us to have any idea whether he has some sort of HR/FB suppression skills.
So basically, you’re picking pitchers who have low ERAs, not realizing that a lot of the time the low ERA is due to a low HR/FB ratio, and then turning around and saying “ha, I told you most of these pitchers have low HR/FB ratios”.
So it’s the chicken or the egg question. Are these the top pitchers because they have HR/FB prevention skills or do you think they are the top pitchers because the low HR/FB ratios are fooling you into believing so?
Dear god, boys we need to support our own (Carty) as some of those comments are plain pathetic. Keep up the battle Derek, we got your back!
Haha, thanks tdotsports1.
I think that Chris’ confusion or doubt stems from not thinking about Bayesian statistics.
Exactly, Mike. I posted something similar in response to Scott over on the RotoWire thread.
Dan Haren is singlehandedly destroying my faith in FIP, xFIP, and SIERA.
I’ve kept him on my team all year long and he just continues to kick me in the nuts, like today with his 7 earned runs allowed. His K/BB is great, his strikeout rate is good, yet his BABIP continues to be off-the-charts high.
Meanwhile, I dropped Tim Hudson in early May because I thought his .240 BABIP was not sustainable and his K/BB was barely at 1.0. Hudson just keeps rolling along. His BABIP is now .235.
I’m beginning to think too much knowledge is a bad thing. I make moves based on underlying peripherals and with the thought of “regression to the mean” in mind, and I’m behind owners who pick up Carlos Silva and Livan Hernandez.
I think the problem is people misuse these peripheral stats.
You need to look at the actual player too. Some players are good at bettering the stats while others don’t live up to them. Matt Cain seems to better them while someone like David Bush is not. Bush had 2 season with a 1.14 WHIP and his ERA was like 4.4 and 4.2.
The other thing I think people assume is that if a guy is labeled unlucky by peripheral stats, he will be lucky for the rest of the year. And vice versa.
Lets assume a SP has a career BABIP of .300. Halfway through the season his BABIP is .240.
Does this mean he will post a BABIP of .360 the rest of the year to make that up? No, he has the same chance to posting a .240 BABIP the rest of the way as posting a .360.
I think stats come most in handy when selling high on a guy. If you can sell a guy with BABIP is .240 at or near full value, odds are good that this SP will not keep up and you will will better your team in the trade. ON the other hand, if someone is trying to sell you the guy with .240 BABIP (and you know he has knowlwgde about BABIP), you can point out that he has gotten really lucky and you might be able to get the guy below value because the other owner thinks major regression is ocming.