Anatomy of a Collapse (October 15, 2003)
Here is how "Win Probability Added" sees it from the perspective of the Cubs.
Marlins 8th, Cubs leading 3 - 0
Chance of Cubs winning: .936
Mike Mordecai: Strike looking, Ball, Ball, Mordecai flied out to left.
Chance of Cubs winning: .956
Prior + Alou = +.020 wins (.956-.936=.020)
Juan Pierre: Ball, Ball, Strike looking, Foul, Foul, Pierre doubled to left.
Chance of Cubs winning: .924
Prior + OF = -.032
Luis Castillo: Strike looking, Ball, Strike looking, Ball, Ball, Foul, Foul, Foul,
Chance of Cubs winning, if Fan wasn't there: .955
Fan + Alou + Prior = 0
Fan = -.031
Alou + Prior = +.031
Castillo walked, Pierre to third on wild pitch.
Chance of Cubs winning: .873
Prior = -.051
Ivan Rodriguez: Foul, Strike swinging, Rodriguez singled to left, Pierre scored, Castillo to second.
Chance of Cubs winning: .792
Prior + OF = -.081
Miguel M. Cabrera: Cabrera safe at first on shortstop Gonzalez's fielding error, Castillo to third, Rodriguez to second.
Chance of Cubs winning: .684
Chance of Cubs winning, if SS makes play at 2B: .868
THIS PLAY WAS HUGE!!!!
SS + Prior = -.108
SS error = -.184 !!!!!!
SS clean + Prior = +.076
Derrek Lee: Lee doubled to left, Castillo and Rodriguez scored, Cabrera to third.
Chance of Cubs winning: .321
Prior + OF = -.363
THIS WAS THE BIGGEST PLAY!!!
Kyle Farnsworth relieved Mark Prior.
Mike Lowell: Intentional ball, Intentional ball, Intentional ball, Lowell intentionally walked.
Chance of Cubs winning: .314
Dusty = -.007
Jeff Conine: Conine hit sacrifice fly to right, Cabrera scored, Lee to third, Lowell to second
Chance of Cubs winning: .237 (Florida takes the lead here)
Farns + OF = -.077
Todd Hollandsworth hit for Chad Fox.
Todd Hollandsworth: Intentional ball, Intentional ball, Intentional ball, Hollandsworth intentionally walked.
Chance of Cubs winning: .227
Dusty = -.010
Mike Mordecai: Ball, Strike looking, Ball, Mordecai doubled to deep left center, Lee, Lowell and Hollandsworth scored.
Chance of Cubs winning: .033
Farns + OF = -.194
THIS WAS THE NAIL!!!
Mike Remlinger relieved Kyle Farnsworth.
Juan Pierre: Pierre singled to right, Mordecai scored.
Chance of Cubs winning: .017
Remlinger + Fielders = -.016
Luis Castillo: Foul, Ball, Ball, Strike swinging, Castillo popped out to second.
Chance of Cubs winning: .018
Remlinger = +.001
The tally:
Prior + SS = +.076
Prior + Alou = +.051
Remlinger = +.001
Remlinger + Fielders = -.016
Dusty = -.017
Fan = -.031
Prior = -.051
Gonzalez = -.184
Farns + OF = -.271
Prior + OF = -.476
Manager = -.017
Fan = -.031
Pitchers = -.368
Fielders = -.502
TOTAL: -.918 (.018 - .936 = .918)
Note: Win Probability changes are based on assuming that both teams are equal in overall talent at any point in time, and that the playing context is neutral between them.
--posted by TangoTiger at 03:39 PM EDT
Posted 3:41 p.m.,
October 15, 2003
(#1) -
tangotiger
Note2: I'm also assuming that the pitchers/fielders are equally responsible on balls in play, which again, may not necessarily be a good assumption when looking at these actual plays. Some balls were extremely hard hit, and most of the blame on those BIP should go to the pitcher.
Posted 4:10 p.m.,
October 15, 2003
(#2) -
fracas
Fascinating, Tango! And unlike some of your work, this didn't go over my head.... *:^)
Posted 4:14 p.m.,
October 15, 2003
(#3) -
Jon
(homepage)
Some of it still went over my head, but I thought it was really cool.
Posted 4:17 p.m.,
October 15, 2003
(#4) -
Dan Werr(e-mail)
Incredible stuff, Tango. I'm duly impressed. Thanks.
Posted 4:18 p.m.,
October 15, 2003
(#5) -
Scoriano
Awesome. The comments were spot on.
Posted 4:28 p.m.,
October 15, 2003
(#6) -
DK
Agreed with all above. This is cool.
one note: where is the SS in the final tally? Shouldn't there be a -.184 for the SS (or -.108 for the SS and Prior) since he booted the ball and didn't make the play.
Posted 4:29 p.m.,
October 15, 2003
(#7) -
Craig B
One play in particular that drew criticism...
On the Conine SF, Sammy Sosa tried to throw home for some reason (he had zero chance to get a tag at home unless Cabrera falls down) instead of throwing the ball into 2nd to keep Lowell at 1st.
If he does that, the WE is .243 instead of .237, so -.006 of the Conine play should go to Sosa alone.
Posted 4:29 p.m.,
October 15, 2003
(#8) -
bob mong
Great stuff, Tango! Breaking down a little further:
The tally:
Prior + SS = +.076
Prior + Alou = +.051
Remlinger = +.001
Remlinger + Fielders = -.016
Dusty = -.017
Fan = -.031
Prior = -.051
Gonzalez = -.184
Farns + OF = -.271
Prior + OF = -.476
Implies:
Prior: -0.226 (25%)
Alex Gonzalez (SS): -0.147 (16%)
Farnsworth: -0.1355 (15%)
Lofton: -0.1255 (14%)
Sosa: -0.1255 (14%)
Alou: -0.1 (11%)
Fan: -0.031 (3%)
Dusty: -0.017 (2%)
Remlinger: -0.007 (1%)
All other fielders (1B, 2B, 3B, C): -0.001 each (0% each)
Prior For Goat!
Posted 4:40 p.m.,
October 15, 2003
(#9) -
Randy Jones
Implies:
Prior: -0.226 (25%)
Alex Gonzalez (SS): -0.147 (16%)
Farnsworth: -0.1355 (15%)
Lofton: -0.1255 (14%)
Sosa: -0.1255 (14%)
Alou: -0.1 (11%)
Fan: -0.031 (3%)
Dusty: -0.017 (2%)
Remlinger: -0.007 (1%)
All other fielders (1B, 2B, 3B, C): -0.001 each (0% each)
Prior For Goat!
given that it was Dusty's call to leave Prior in, I would vote Dusty for goat...
Posted 4:42 p.m.,
October 15, 2003
(#10) -
Randy Jones
also wanted to add that this was interesting stuff...too bad the mainstream press will never look at it (and would find some way to misunderstand and misinterpret it if they did) and will continue to harp on the fan play.
Posted 4:42 p.m.,
October 15, 2003
(#11) -
tangotiger
DK: yes, I threw that in with "Fielders".
Craig: Good point. If I were to watch the inning over, we can do additional micro-level things as you are doing it. The key is for everything to add up.
As well, there were many hard hit balls in that inning, making the OF almost inconsequential in those cases, and Prior/Farnsworth should get more blame.
***
Looking at the Jeter play at the end, you can almost credit that play as if it was 80% hit, 20% out (and credit the pitcher as such). That Jeter makes the play gets him more credit. At the same time, Bernie looked like he might have been able to make it, so, technically, you could give Bernie some credit for being in the position to make that play (say if Jeter wasn't there, maybe Bernie would have made that play 85% of the time). You can really get into the nitty-gritty with all this. However, if you look at it on a seasonal basis, this kind of approach will lead to very few surprises (the breaks even out, more or less). The impact is really felt at the game or inning level.
Posted 4:43 p.m.,
October 15, 2003
(#12) -
bob mong
given that it was Dusty's call to leave Prior in, I would vote Dusty for goat...
I dunno - you can't blame the manager for everything bad that happens on the field...but I think you have a point in this case. Giving half of Prior's WE to Dusty puts Prior at -0.113 (13%) and puts Dusty at -0.13 (14%) - still below Gonzalez and Farnsworth.
Gonzalez For Goat?
Posted 4:48 p.m.,
October 15, 2003
(#13) -
tangotiger
By the way, when I did WPA for 1999-2002, one of the pitchers that was in the leaders was Mike Remlinger. I haven't looked at his performance this year, but how has he pitched? Is there a reason that Dusty did not go to him earlier? Was this his pattern all year?
Posted 5:02 p.m.,
October 15, 2003
(#14) -
jwb
Hmm. . . Getting close to loss shares here.
Posted 5:08 p.m.,
October 15, 2003
(#15) -
Alan Jordan
Damn!
Good work Tango. That's a great use for WPA that I wouldn't have thought of. Of course the fan will always be blamed because its easier to lock on to an anaecdote than to weigh facts.
By the way what is the fan's wins above replacement fan?
Posted 5:09 p.m.,
October 15, 2003
(#16) -
Randy Jones
Bob, you are probably right, I just wanted to point out that I didn't think Prior deserved quite so much blame for what happened.
Posted 5:17 p.m.,
October 15, 2003
(#17) -
Sam Hutcheson
I dunno - you can't blame the manager for everything bad that happens on the field...but I think you have a point in this case. Giving half of Prior's WE to Dusty puts Prior at -0.113 (13%) and puts Dusty at -0.13 (14%) - still below Gonzalez and Farnsworth.
Gonzalez For Goat?
Uh, yeah? The guy that booted the routine grounder that would have severely limited if not completely ended the inning? Yeah, that guy for goat.
It's good when analysis supports conventional wisdom. (Note, conventional wisdom does not equate with "what the Cubs fans and media are yabbering ignorantly about.)
Make the routine defensive play and the Cubs are waiting for an AL opponent.
And by the way, great work T.
Posted 5:17 p.m.,
October 15, 2003
(#18) -
Craig B
Hmm. I've gotta think that the fan's WARP is about -.031 as well.
(Tongue into cheek)
The replacement level fan is the guy who doesn't go to the game, I would think. You could consider that fan as pocketing the $30 for the game ticket and otherwise having no effect.
Average MLB payroll is about $55 million on revenues about twice that, and the average 81-81 team is about 30-32 wins above "replacement team" with a payroll of about $8 million for 25 guys + the DL. Call it 32 wins and $48 million, or about $1.5 million per marginal win. Now the fan's contribution is $30, of which half ($15) goes to player salaries. So the Wins Above Average of the replacement fan, courtesy of the $15 he takes out of the team's budget, is about $15/$1.5 million or -.00001 wins above average.
This fan was at the game, but was also -.031 wins above average for the play at the railing. So his Wins Above Replacement Fan was -.03099, I think. :)
Posted 5:23 p.m.,
October 15, 2003
(#19) -
steve d
I don't think that you can place that much blame on Dusty for not taking out Prior. Even if he did remove Prior earlier, he still probably would have put in Farnsworth who wasn't exactly mowing down hitters. The way I see it, I would defer some of Prior's value to Farnsy, not Dusty.
Oh and great work tango!
Posted 5:33 p.m.,
October 15, 2003
(#20) -
Matt
Great, great stuff Tango, many thanks.
Keep up the good work.
Two comments (1) This was not just about the Cubs blowing it, it was also about the Marlins getting one big hit after another to win the game. They aren't getting nearly enough credit. What happened to the old cliche of "good teams pounce on your mistakes"? The Marlins certainly did.
(2) I don't like the fan getting a negative score that assumes a sure out on this play. Alou very well might have caught the ball, but there is also at least some possibility that he wouldn't have. Again this doesn't make for good media coverage to admit the possibility that the play might not have been made anyway.
Posted 5:42 p.m.,
October 15, 2003
(#21) -
tangotiger
Thanks for the kind words, guys!
Matt: hmmm... I agree, it would not be a 100% sure out. Let's say that Alou would have made that play 80% of the time. So, the fan's WPA should be -.025 and not -.031. Alou+Prior get +.025.
Posted 7:09 p.m.,
October 15, 2003
(#22) -
David Brazeal
This is meaningless, Tango. Because it was the Cubs, and it was an LCS elimination game, they had a 0% probability of winning last night's game when it started, and had a 0% chance of winning the game at every point along the spectrum. Your silly "statistics" just can't factor in curses, can they? Admit your limitations!
Just kidding -- this was great. Very interesting analysis, and just enough went over my head to allow you to retain your Stephen Hawking-esque mystique.
Posted 8:09 p.m.,
October 15, 2003
(#23) -
FJM
Fascinating! Naturally, I have a few questions.
1)Did you consider the possibility that the Run Expectancy (and hence the Win Probability) had already changed during Castillo's at bat by virtue of the count going to 3-2? In the NL last season the overall OBP was .332. But it jumped to .467 after the count reached 3-2, a 41% increase. So instead of a 1-in-3 chance of getting on base the odds had already changed to nearly 50/50. That made the fan interference somewhat more costly than if it had occurred on the first pitch.
2)How does this one compare to other notable post-season collapses (2002 W.S. Game 6, the Kim blowups in 2001, Billy Buckner & Bucky Dent)? Was this the worst ever?
Posted 8:30 p.m.,
October 15, 2003
(#24) -
Jeffrey Maier & Don Denkinger
How does this one compare to other notable post-season collapses?
We may be ahead of Bartman in the non-player category.
Posted 10:05 p.m.,
October 15, 2003
(#25) -
Alan Jordan
Craig B -
This fan was at the game, but was also -.031 wins above average for the play at the railing. So his Wins Above Replacement Fan was -.03099, I think. :)
Impressive, but can you calculate how much the Cubs should pay him not to come to a game? :)
Posted 8:39 a.m.,
October 16, 2003
(#26) -
David Smyth
What about the probability of the *larger* collapse? What are the odds of the Cubs losing the last 3 games, with 2 of them at home, and the advantage in the starting pitchers?
Posted 9:18 a.m.,
October 16, 2003
(#27) -
Tangotiger
I'll guess it's around .4^3 = 6%
Posted 12:21 p.m.,
October 16, 2003
(#28) -
tangotiger
FJM: that's a good point, and I'm glad you brought it up. On a 3-2 count, it is definitely a hitter's count. In a random inning (say the 4th), if the team normally has a .500 chance of winning on an 0-0 count, it would have a .507 chance of winning on a 3-2 count. I will take a guess that if the chance of Cubs winning at the start of the Castillo PA was .924, then it was probably around .922 or .921 on a 3-2 count. So, instead of the fan costing .031 wins, it might have been .033 or .034. That's assuming a sure out, which it certainly wasn't. Giving Alou an 80% chance of coming out with that ball, that reduces the fan to .026 loss. At this point in the game, the count didn't come into play much.
On a side note, since the count was 3-2 before the fan got in the way, and it remained 3-2, there was no change in Win Expectancy for having the pitcher throw an extra pitch. It was essentially a "let".
Posted 3:04 p.m.,
October 16, 2003
(#29) -
DK
To continue that... does the batters chances improve the longer the at bat is? For instance, Castillo's at bat was 9 pitches long (the 8th was the infamous foul). Conventional Wisdom says that with each pitch the batter improves his chances of getting on base. I suppose its because the pitcher is laboring or he has shown all he's got. Or it may be because the pitcher will eventually, given enough time, throw 4 balls.
I guess what I'm getting at is that it may not have been a 'let' since Castillos chances improved the longer he was fouling off pitches. I assume its a miniscule improvement, say .001
Posted 3:30 p.m.,
October 16, 2003
(#30) -
tangotiger
I think Keith Woolner did some study along those lines. I don't remember the results of that. My guess is that if you are at 3-2, and you continually foul-off the pitch, you may get a miniscule advantage as a hitter, but I'd be surprised if the numbers would show anything that would be statistically significant.
Posted 3:58 p.m.,
October 16, 2003
(#31) -
DK(e-mail)
thanks tango. I really wouldn't expect it would change the WE of the game that much. Just something I thought of. I'll root around to see if I can find Wollner's study.
Posted 4:13 p.m.,
October 16, 2003
(#32) -
tangotiger
(homepage)
This is the Woolner study.
Aggregating the "at least 10 pitches" here's the breakdown:
5,935 PA / .242 / .430 / .406
A batter on a full count does better than that. I really don't think you see a tiring effect on that particular batter, but perhaps on the next batter or for the rest of the inning, maybe you might.
Perhaps consecutive 25 pitch innings might have an effect on the 3rd inning too.
Posted 6:34 p.m.,
October 16, 2003
(#33) -
FJM
I remember Woolner's study. I (and many others) wrote to him at the time that it was flawed in that it mixed two different effects into one: the number of pitches thrown and the pitch count. I believe he published a followup study in which he separated the two. Sorry I don't have the link.
According to Woolner, the weighted average OBP for pitches 12 and up is .461. That number will only go down as you bring in pitch 11, pitch 10, etc. (Pitch 11 and above is .449, for example.) Yet as I stated the NL OBP on a 3-2 count was .467 last year. So either the two data sets are very different or there are a lot of long at bats where the count never reaches 3-2.
Posted 11:10 a.m.,
October 17, 2003
(#34) -
Mike Emeigh(e-mail)
So either the two data sets are very different or there are a lot of long at bats where the count never reaches 3-2.
It's actually the other way around - there are a lot of 3-2 counts that never reach 10+ pitches. Most 3-2 counts are resolved within a pitch or two.
-- MWE
Posted 11:35 a.m.,
October 17, 2003
(#35) -
tangotiger
FJM: I agree, the Woolner study does not show what we want, and it was pretty clear of that the first time reading his article. He did followup on that in a "mailbag" issue I think. There was a similar issue with reaching base on error, from what I remember.
The "problems" in those articles, from the perspective of some of the readers, was they had a question to answer, and the data was not properly set up for them to do so. However, from the perspective of Keith, he was simply presenting data from a different perspective, and trying to infer things from that data.
This is the classic process that I describe: ask your question first, then find the data to answer that. Doing the reverse, like "here are the results of 10-pitch-plus outings...what does that tell you" won't answer every question you may have.
Posted 12:03 p.m.,
October 22, 2003
(#36) -
tangotiger
On BP (from probably Woolner), they are reporting
"Visitors have three-run lead with one out in the 8th inning, 1972-2003: 6281-445 W-L (.934 winning percentage) "
In my math model, I have it at .929. Again, all those things that differ from the overall average in this situation (8th inning, 3 run lead) doesn't amount to much, if anything.
Posted 11:34 a.m.,
October 24, 2003
(#37) -
Walter(e-mail)
Games are played on the field and not on a computer for a reason. The problem with the analysis is that it does not consider the physchological and emotional impart of Steve Bartman's interference. It also doesn't consider the fact that if Alou catches the ball, gold glove shortstop Alex Gonzalez is playing at regular depth and not in double play position. Basically, the analyis is useless.
Posted 12:14 p.m.,
October 24, 2003
(#38) -
tangotiger
The analysis is useful to the point that you accept the underlying assumptions of "win probability added". That assumption is that you only care about the current inning/score/base/out/pitch state, and not how you got there.
Therefore, the traumatic experience that Mark Prior had with the 4 fans reaching over, with one of them batting it away is irrelevant to him walking Castillo, under these assumptions.
If you mean to say that the assumptions are not acceptable, then I don't have a problem with that.