-
PAxFIP
Posted by Troy Patterson on Friday, August 21, 2009
Sabermetric StatisticsI gave this an initial try and wanted to see how the numbers looked. I am going to do a bit more work, but here is the cells with FIP, xFIP and PAxFIP(Park Adjusted expected Fielding Independent Pitching). The name "Park Adjusted" is a bit simplified since they aren't all home games, but the home park plays the largest factor. I still need to make adjustments for the Mets and Yankees new home parks, but here is a look at all the qualified started this year.
PAxFIP=(((FB*teamHR/FB)+(BB*3)-(SO*2))/IP)+3.2
teamHR/FB is currently being calculated by a 4 year running weighted average of a teams HR/FB. Not sure if that is enough data or to much, but I will play with it over the next few months.
Updated 9/28/09 -
I will keep a link to this in the top of the site for future use. It will not be updated daily though.
Thanks to FanGraphs for the HR/FB and FB% data.
6 comments:
-
Thanks Aaron.
The ultimate goal will be to prove its actually better than FIP or xFIP and then make a better way to show it. I don't have access to this data right now to set it up for day to day updates, but I would like to get a system to make it a quicker turn around.
The good news is since they are pitchers I won't be missing 5-7 days of data before I update each time.
-
FIP is nice but not quite useful. i know zimmermann has a better FIP compared to his ERA but the defense behind him sucks and his ERA wont be corrected until the defense is... who cares until someone is traded
-
Thanks MDS for commenting. I agree and the first point of xFIP is to get rid of the HR/FB problem. I attempted to get rid of the problem of team differences to HR/FB.
My next goal will be defense/BABIP effects. This is much harder as defense is very unpredictable from day to day. Over the season the Nats had a very bad defense, but adding Morgan to the outfield will help some. Incorporating these changes will be tough.
-
Thanks for doing this analysis. It's interesting to see how controlling for park effects gives you a result that is sometimes between FIP and xFIP (what you would expect if HR/FB is determined largely by park effects) and sometimes above or below both FIP and xFIP. I'm looking forward to seeing what you find when you try to compare the predictive power of PAxFIP to xFIP.
-
Wow, look at the gap for the Big Unit. Almost a point and a half between his FIP and his PAxFIP.
That is a lot of information. I've been looking for a convenient source of xFIP and the PAxFIP seems like it would probably be even better. I don't know if it's possible but it might be easier to read if the cells were color shaded. Say, green for anything under 4.00, yellow for 4.00 to 4.25 and so on. That would give a quick visual way to judge a guy and see and if there are big discrepancies between his numbers.