Clutch Changes
Last night we rolled in some changes to the way Clutch is calculated. For the most part, the results remained the same, but there are a few key differences which you should be aware of.
Clutch is no longer calculated using OPS Wins. Instead it’s calculated using WPA/LI, which is WPA divided by LI on a play-by-play basis. Take Alex Rodriguez for example who has a WPA of 1.03 and a pLI of 1.51. His WPA divided by pLI on a full season basis would be .68. Unfortunately, that is weighting each play incorrectly and does not correctly neutralize WPA. When you correctly weight each plate appearance by LI, his WPA/LI is 1.16. To get the new Clutch score, you merely do: WPA – WPA/LI.
Tangotiger goes into the details of what’s going on here in comment #5: What Exactly Clutch Measures.
Print This Post
Even further explanation is in posts 11 and 12 of that same link:
http://www.insidethebook.com/ee/index.php/site/comments/wpa_to_arod_i_love_you/#11
I encourage all to read it. It’s an important concept, and big kudos to David for implementing it.
Thanks for the update, David. I think this adds a lot of value to the data already presented.
Unrelated question, but is there any way to separate batter WPA from baserunner WPA?
Example: In the Bottom of the 7th last night, with the Orioles losing 1-0 and a man on first, Kevin Millar hit a double. Sounds like a pretty valuable play, but the baserunner was thrown out at home, so Millar was credited with -.036 WPA.
It seems like correcting this might require a complete overhaul of the system, so is it worth looking at, or is this type of event too rare to be significant?
It wouldn’t require a complete overhaul of the system, but it would require some slightly more granular play-by-play data. This is something I’ll look into as the season continues, but over the course of the season, these types of plays probably don’t have a huge impact.
The alternative is to give the batter an average based on this chart:
http://www.tangotiger.net/destmob.html
Say you have a guy on 1B, 0 outs, and the batter singles. You would compute the WE based on the first table. So, give the batter credit for 68.2% of the win probability for 1B, 2B, 0 outs, and 25.8% for 1b, 3b, 0 outs, etc. So, the difference between the starting WE and this intermediate WE you give to the batter, and the difference between this intermediate WE and the actual ending WE goes to the runner.
It’s alot more work of course, and you won’t get the payoff. Maybe a 0.5 win difference either way. And, you’d want to figure this for each home park.
I’m following along with the Yankees-A’s game right now. Doug Mientkiewicz was thrown out at home on a wild pitch with runners on first and third and two outs. It looks like Bobby Abreu was credited with the negative WPA on the play (since he was the previous batter). This doesn’t seem logical to me. There has to be a way to credit hitting and baserunning WPA separately.
The data used in the live data is not “official” and is often corrected during the nightly loads around 3am EST. It’s a matter of how the play is broken up, and at the moment we have very little control over that. Over the course of the season, It shouldn’t make a huge difference (see Tangotiger’s previous comment), but we’re looking into getting more granular play-by-play data.
I was just wondering if there’s any separation between the pitcher and fielder as far as WPA goes.
For example if the batter grounds out, the batters WPA goes down and the pitchers goes up. Shouldnt the fielder receive some credit? Also if there’s an error should the deduction of WPA go to the fielder and not the pitcher?
Anthony: Doug Mienkiewicz in the “official” data was given separate credit for being tagged out at home. I wasn’t sure if that would be the case, but it looks like it worked out ok.
John: On FanGraphs the pitcher gets credit for everything the defense does. I’ll refer you over to this discussion: Fielding and WPA
Yeah I guess it would be kinda hard to separate the two unless you’ve watched all the games.
____________________________________________________________________
Lemme see if i have this right.
OPS Wins would be what a player does in a neutral context
BRAA factors in runners on base
WPA then factors in runners on base as well as inning/score
Wouldnt clutch be as simple as WPA-OPS Wins? Im probably missing something since I just started reading all of this.
Interesting. Thanks for clearing that up, David.