Question about park factors
by nny on Dec 23, 2009 12:39 PM EST
14 comments
From what I've gathered, generally speaking people use weighted 3 year park factors.
My question is, why?
For one, we have a control: The park itself (Unless the team makes changes to dimensions/field/ect).
There are two variables I can think of that then come into play: Weather and actual baseball results.
Since the control is what we are measuring, wouldn't that mean that it'd be smarter to use the biggest sample size possible, since the other two variables can skew the results? I mean, there's a reason home/away splits for a player aren't generally taken into factor, because there's huge random variance in it. I mean yes, the park will effect the baseball results, but there's so much variance in baseball results and that's why we always use the biggest sample size possible. And weather is going to be weather, again sample size ftw.
And yet people will say things like "Stadium Field Park played like a pitchers park this year, even though it was a hitters park last year." The park itself didn't change though. The results inside the park changed, sure, but the park stayed the same, the results changed because of other variables. So what am I missing? Why use weighted 3 year park factors over the biggest sample size possible?
And just as a lesser question: What's the reason behind getting K% by K/AB instead of K/PA? Fangraphs does K/AB but I don't really understand why they do that instead of K/PA (And other places, like statcorner, do K/PA). Does it even matter?
0 recs |
14 comments
|
Email
Print
Comments
The park is not the only thing that can change.
Remember, we typically express park factors in terms of the rate of run scoring compared to the rest of the league. Even if the park in question doesn’t change, the other parks in the league can and do change.
In other words, if we change Coors Field we change ALL parks in the league, albeit subtly. Over time, as we change more and more parks we change the league average more and more.
Is there a way we could adjust for this explicitly? Sure, I s’pose. But it’s a giant hassle and most people it seems have decided that using three years is “good enough” that it’s not worth the additional effort.
And – personally I would do nothing “per at bat,” but that’s just me. I can’t speak for what Fangraphs does or why they do it.
Ah, thank you. That definitely makes sense. And it’s not even about changing parks (or the ball in the case of coors), but adding/subtracting parks. Only 5 times in the past 20 years (1990) has their been a season that a new park was not introduced.
I also agree about never doing “per AB”, which is why I found it interesting that Fangraphs does that.
More reading, if you're interested
Patriot has a very nice write-up on park factors:
http://gosu02.tripod.com/id103.html
He uses 5-year park factors as a compromise between the issues that Colin brings up and sample size. I use his in basically all of my stuff. More recent park factors are released on his blog:
http://walksaber.blogspot.com/
As for K% at FanGraphs, I had no idea that they do this. I wonder if they even realize that they’re doing this…? I remember some years ago, they weren’t subtracting home runs from their BABIP calculations either. Sometimes simple errors (which is my guess as to what it is) like that never get corrected because they don’t realize it’s happening. I’d drop them a line.
-j
I write at:
Beyond the Boxscore | Red Reporter | Basement-Dwellers.com | Twitter: @jinazreds
Not a mistake.
The K/AB isn’t a mistake, but it’s something I’ve considered changing. The original reason K% uses at-bats on FanGraphs is because it’s really meant to simulate the inverse of one of Shandler’s stats: contact%. The argument is that if you don’t swing the bat, you can’t strike out, so it was more meant to try and isolate strikeouts as a “skill”, where K/PA is going to be muddied by walks.
Anyway, that’s the reasoning. There was a long thread in the forums a couple years ago about it, but hasn’t been a whole lot of talk about it since.
I personally don’t remember not subtracting HR from BABIP, but maybe you’re right, must have been within the first couple months of the initial launch over 4 years ago.
When put that way...
…it sounds like the DIPS way of figuring $K, or K/(PA-BB). In actual practice there’s probably no difference whatsoever.
While you're here...
Are the minor league wOBA’s on fangraphs park adjusted?
A few weeks ago I calculated Matt Dominguez’s OPS+ for 2008 and 2009 to show that even though there was a massive OPS drop (.853 to .753), put into context he still performed basically the same way (132 OPS+ in 2008, 130 OPS+ in 2009).
But his RC+ on Fangraphs is 136 for 2008, 117 for 2009.
Is it a difference in RC+ and OPS+ or is it that it’s not park adjusted?
(And more thanks to those who have posted since my last response)
Can't speak for FG, but I'm pretty sure no wOBA valus are park adjusted.
wRC+ is different. I’m inclined to say that it isn’t, but I’ll let David answer if he is still here.
Marlin Maniac, a Florida Marlins blog
Come attend Intro to Sabermetrics 101!
Check me out at Beyond the Box Score as well.
*values
Marlin Maniac, a Florida Marlins blog
Come attend Intro to Sabermetrics 101!
Check me out at Beyond the Box Score as well.
No minor league park adjustment
There’s no minor league park adjustment for wRC+, only major league and none of the wOBA values on the site are park adjusted.
wRC+ is always league adjusted though, so the AA vs A+ baselines are going to be considerably different and will definitely have a big impact on wRC+
Unless there’s some massive park adjustment going on, I’m not sure how Matt would retain a 130 OPS+ with such a big dip in OPS. I don’t think the league averages are that different, but I don’t have them in front of me so I could be wrong.
Yes, there is a massive park adjustment going on
Greensboro is a massive hitters park, Jupiter is a massive pitchers park.
I originally got the 132/130 numbers I was using unweighted park factors from 2002-2008 (since I don’t have access to 2009). Just going from minor league splits park factors (Which is 3 year weighted, plus has 2009 [I hope he posted 2009 soon, I sent him an e-mail asking for them], so going to be better than what I had anyway):
Greensboro: .354/.464/.818
OPS+: 130.389
Jupiter: .340/.454/.794
OPS+: 130.659
So actually, going by minor league splits’ park factors, he actually did better in Jupiter than he did in Greensboro, even though he had a 100 OPS drop. Park factors make that 100 OPS drop to 24 points, and league factor reduces the rest.
This is something I really hate about a lot of prospect watches, they just look at OPS without really looking at context. I mean, you will have the people who say “he played in X park” but don’t quantify just what it means (i.e. last year a lot of people were down on Dominguez cuz they though his numbers were all Greensboro).
So consider this my formal request to park factor the minor leagues. I’ve been thinking of doing a massive minor league OPS+ or RC+ dump to actually be able to see how minor leaguers actually do [Which is why I made this thread, to know what exactly to do about park factors], but if you guys (fangraphs) just park factors yours that’d certainly make everything a lot easier.
Sorry for the late response
David, yes, I noticed the BABIP thing very early in FG’s launch, roughly the same time that I started my blog. It hasn’t been this way for a long time, and I generally trust FG’s numbers more than any other site—especially now that THT dropped it’s stats!
-j
I write at:
Beyond the Boxscore | Red Reporter | Basement-Dwellers.com | Twitter: @jinazreds
Sort of tacking on the BABIP point Justin made above. I have a feeling FG isn't adding SF to the denominator of BABIP.
I know that B-R does, and their numbers are consistently lower than those on FG. Just a thought.
Marlin Maniac, a Florida Marlins blog
Come attend Intro to Sabermetrics 101!
Check me out at Beyond the Box Score as well.
Even if a home park doesn’t change, if a road park does that alters the home/road ratio.
What I do is create a series of matched pairs of ballparks, for example how the Pirates and Phillies did in Three Rivers compared to how the same teams did in Veterans, for as many years as the parks coexisted. Repeat for each road ballpark in the lifetime of the home ballpark (I track ballpark versions when configurations are changed), then sum home and road and compare the results.
by Brian Cartwright on Jan 6, 2010 12:16 AM EST reply actions
Comments For This Post Are Closed
Secondary Sidebar
User Tools
FanPosts
Community blog posts and discussion.
Recent FanPosts
BtB on Facebook
Certain photos copyright © 2010 by Associated Press or Getty Images. Any commercial use or distribution without the express written consent of Associated Press and Getty Images is strictly prohibited.
