How do you calculate WAR for RP?
by nny on Aug 1, 2009 1:18 PM EDT
21 comments
I figured this would probably be the best place to ask.
The only place I've found online that even gets into is fangraphs. However, they very rarely ever actually gave formulas, it's mostly just words, which is very annoying. Would be so much easier and efficient if they just posted the formula then explained the variables.
But I digress.
At -> http://www.fangraphs.com/blogs/index.php/pitcher-win-values-explained-part-seven
He does give the basic formula for SP, which is (((18-AIP) * LRPG + AIP * FIP)/18)+2)*1.5
Where
AIP=Average IP per start
LRPG=Leage's average Run Per Game
and FIP=Fielding Independent Pitching.
This is the part that should be different for RP. Afterall, you can't have an average IP per start for RP lol. But what's the formula for this part for RP? The rest should all be the same, outside of the league average win percent added for RP, but that's addressed in an earlier posted (.470%).
0 recs |
21 comments
|
Email
Print
Comments
Formula:
X is (RS+RA)^.267, which should be some number similar a tenth or two below 2.0
RS is league-average runs scored per nine innings (not just earned runs).
RA is the reliever’s runs scored per nine innings (not just earned runs).
.470 is the rep level winning percentage for relievers. It’s more like .460 for AL relievers and .480 for NL relievers.
IP is easy.
The (LI + 1)/2 part is a shortcut that takes into account the leverage of the situations a pitcher pitched in, plus bullpen chaining. Bullpen chaining basically accounts for the fact that a if a closer is injured, he won’t be replaced by a scrub. Instead, the scrub goes at the end of the bullpen and everyone is bumped up a notch and the setup guy is the new closer. For more, read this: http://www.beyondtheboxscore.com/2009/4/29/856308/bullpen-chaining-and-reliever-war
I don’t think that’s how Fangraphs does it, but that’s how I’d do it and the results are probably pretty similar. For RA, you probably want to use tRA, FIP/.92 or something else that accounts for fielding and park.
Beyond the Boxscore Not a member? Sign up.
And crap, formula got deleted by autoformat. Grrr.
{ [ RS^X / (RS^X + RA^X) ] - .470 } * IP/9 * (1+LI)/2
Beyond the Boxscore Not a member? Sign up.
yeah using tra
thanks a bunch. I’d kiss you if that’s your kinda thing.
by nny on Aug 1, 2009 5:37 PM EDT up reply actions
RA * .92 = ERA, roughly
Or, 92% of runs are earned.
---
http://www.beyondtheboxscore.com
http://www.rightfieldbleachers.com
Yup, it puts FIP on the RA scale instead of the ERA scale.
Not perfect, but should be fine for most uses.
Beyond the Boxscore Not a member? Sign up.
since you're so helpful :D
We’re projecting future WARs, is it better to use the regressed tRA (aka tRA*) or normal tRA?
I’d assume normal tRA (and hoping because I’ve never worked with regression before lol), but better to ask more knowledgeable people :D
by nny on Aug 2, 2009 8:27 PM EDT up reply actions
At least tRA*.
More preferably you’re using three years of data. (Especially for relievers, who get such small sample sizes even in a full season; the most you’re likely to get out of a short reliever is, what, 80 IP?)
projecting minor leaguers
so depends what league they’re in. doing both SP and RP.
Would I regress to the league numbers, project the ML numbers based off the regression, then regress those to the MLB numbers?
Only regress MILB numbers? Only the MLB numbers?
You'd need to translate the minor league numbers.
Which is tricky – ideally you’d translate the individual components and then do the component ERA (which is what tRA is). MLEs are hard, very hard, to do right.
As far as what to regress to – it’s a tough question. That’s actually the subject of my THT article for this week, and the answer is not as clear-cut as I think you’d like.
You can never go wrong with regressing to the mean
F*** Billy Beane... actually, I kinda like Holliday
Sure you can!
You can do anything wrong.
Here’s what it comes down to – which mean are you regressing to?
I would say the league mean
But I guess you think you should regress to the population mean. However, it’s hard to know, except in obvious cases, which population a player falls into. That’s why in most cases, it’s seems okay to regress to the league mean.
F*** Billy Beane... actually, I kinda like Holliday
Which league, though?
We’re talking about projecting a minor leaguer as a major leaguer – so do you regress them to the league mean for the league they are currently in or the league you’re projecting them to?
Well it depends on how many plate appearances they rack up
There is obviously no cutoff, but after say, 600 plate appearances, you could say that this guy is a “major leaguer” and you would regress to the major league mean.
F*** Billy Beane... actually, I kinda like Holliday
How timely...
Just started writing about reliever WAR and what makes a HOF reliever. Glad to see LI is part of it. Was hoping so.
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.
