Baby Winexp and the Big Questions of Today

by Apricot on Aug 22, 2006 9:38 PM PDT

Okay, so I fell off the wagon and haven't posted any Baby W diaries in a few weeks. House remodeling plus travel plus insane energy baby = no time.

Anyway, without further ado, here are Baby W's comments on the pressing questions of the day.

(huh? what is this? and what are these pictures?)

Who is the A's MVP so far?

Baby W says that either Swisher, Haren or Frank Thomas would be reasonable choices, but she picks Swisher. No one else is close.

Name WXC Plays
---- --- -----
Nick_Swisher +2.503 525
Dan_Haren* +2.395 760
Frank_Thomas +2.032 413
J_Duchscherer* +1.358 181
Barry_Zito* +1.330 758
Kiko_Calero* +1.317 212
Milton_Bradley +1.127 265
Jay_Payton +0.961 441
Joe_Kennedy* +0.846 78
Eric_Chavez +0.674 455
Chad_Gaudin* +0.617 206
Dan_Johnson +0.502 275
Joe_Blanton* +0.375 687
Mark_Kotsay +0.352 478
Huston_Street* +0.322 248
Sco_Sauerbeck* +0.242 50
Bobby_Kielty +0.186 234
Rich_Harden* +0.124 143
Jay_Witasick* +0.001 60
Jimenez +0.000 2
Santi_Casilla* -0.020 11
Shane_Komine* -0.053 44
Steve_Karsay* -0.059 46
Ron_Flores* -0.061 21
Doug_Clark -0.072 7
Esteb_Loaiza** -0.086 6
Barry_Zito** -0.110 4
Joe_Blanton** -0.111 3
Dan_Haren** -0.166 7
Brad_Halsey* -0.191 374
Jason_Windsor* -0.255 39
Randy_Keisler* -0.269 42
Kirk_Saarloos* -0.319 453
Ron_J. Flores* -0.361 61
Matt_Roney* -0.390 19
Mike_Rouse -0.395 25
Marco_Scutaro -0.464 278
Bobby_Crosby -0.696 402
Antonio_Perez -0.853 97
Esteban_Loaiza* -1.115 471
Mark_Ellis -1.122 366
Adam_Melhuse -1.303 119
Jason_Kendall -1.778 468

pitcher pitching = *
pitcher hitting = **

Who Is The A's Least Valuable Player?

Least valuable players: Kendall beats out Ellis and Loaiza in a runaway race to the bottom. Yes, even counting his recent surge. While all three were equally terrible at the end of July, Loaiza and Ellis have had outstanding Augusts to drop out of the race, while Kendall only contributed .500 baseball.

Who are the MVPs?

Baby W says that it's Pujols and Ortiz and it's really really not close. And amazingly enough she really does agree it's a battle between Ortiz and Jeter in the AL. Remember, Baby W doesn't measure defense or grit.

STL Albert_Pujols 7.38 482
BOS David_Ortiz 7.12 555
PHI Ryan_Howard 5.23 530
NYY Derek_Jeter 5.23 565
CHW Jermaine_Dye 4.34 467
HOU Lance_Berkman 4.33 492
NYY Jason_Giambi 4.18 483
MIN Justin_Morneau 4.16 498
FLA Mig_M. Cabrera 3.91 524
BOS Manny_Ramirez 3.89 529
CLE Travis_Hafner 3.81 521
PHI Chase_Utley 3.74 574
NYM David_Wright 3.72 538
ARI Brandon_Webb* 3.63 737
ANA Vladi_Guerrero 3.58 541
KC Mark_Teahen 3.49 385
NYM Carlos_Beltran 3.49 501
MIN Joe_Nathan* 3.33 201
SF Barry_Bonds 3.29 383
BOS Jona_Papelbon* 3.18 245
BOS Kevin_Youkilis 3.16 561
CIN Adam_Dunn 3.02 543
TOR B.J. Ryan* 3.00 236
ATL Andruw_Jones 2.94 511
STL Chr_Carpenter* 2.92 686
CIN Ed_Encarnacion 2.84 332
ANA Fra_Rodriguez* 2.83 220
CHW Jim_Thome 2.82 495
SF Jason_Schmidt* 2.81 726
MIN Franc_Liriano* 2.69 478
CHC Carl_Zambrano* 2.69 766
CLE Grady_Sizemore 2.67 598
ATL Marcus_Giles 2.65 514
TOR Vernon_Wells 2.64 535
TOR Roy_Halladay* 2.62 745
COL Todd_Helton 2.61 488
NYM Duaner_Sanchez* 2.58 232
NYY Johnny_Damon 2.54 544
OAK Nick_Swisher 2.50 525

By the way, Edwin Encarnacion has taken over the position of Highest Ranked Player I've Never Heard Of. Amazingly, he isn't a setup guy, he's a genuine position player, a young 3B for the Reds, and someone who follows the NL can probably say why Baby W loves him so much.

Ethier or Bradley and Perez?

Baby W declares this... a draw. Amazing.

OAK Milton_Bradley +1.13 265
LA Andre_Ethier +0.22 344
OAK Antonio_Perez -0.85 97

And most of the difference between Bradley and Either comes from a spectacular event around AN Day. I wonder what that could have been...

Jeter or A-Rod?

In every previous year, Baby W has considered A-Rod to be way way way better than Jeter. So she spits on the idea that A-Rod is generally a choker. And I think any reasonable look at the numbers shows that A-Rod has historically not at all been a choker compared to Jeter. Quite the contrary. That said, much to my shock, this year, Baby W thinks A-Rod has been a dud, approximately contributing .500 baseball to the Yanks, while Jeter has been a star. Just for kicks, I include Ortiz and Damon for comparison.

How Are The Big Three Doing?

How are Huddy and Mulder compared to their replacements? Huddy and Cupcakes are contributing similarly, and while Haren has become a top-notch pitcher, Mulder fell off a cliff around the middle of May and hasn't recovered.

The Ex-A's We Like

Byrnes, Miggy have had pretty unamazing years. Ramon was killing the ball until he collapsed around mid-July. And I don't know what kind of Wheaties Dye started eating, but he is having a shockingly great year.

The Ex-A's We Curse

And we wrap with a look at Redman, Rhodes, the Incredible Foulke and Jason G. Gotta say, I still have negative feelings about this gang, though I forgave Jason G sometime around the depths of last year. The first three have done nothing to make us miss them, but Jason G has been, with Jeter, one of the best players in baseball. Let's hope it's due to great exercise and talent and not HGH.

That's all for now. I have a dream of putting Baby W on the web so I don't have to make these reports myself. It might happen, but the school year is starting...

Bonus Blurry Picture of Hopey and Maya at AN Day

Tweet Comment 47 comments | 0 recs |

Do you like this story?

Comments

Great work!

Thanks for all that but I have one small distraction.... I got confused when I read "Either or Bradley and Perez"?
I see now it's a misspelling on Ethier... not Either or !

And photos of babies at baseball games... so cute!

"Maybe if I lay here in the dirt, nobody will notice I've forgotten how to hit" ~Crosby in ANtics

by streetfan on Aug 22, 2006 9:49 PM PDT reply actions

corrected

Thanks!

by Apricot on Aug 22, 2006 10:04 PM PDT up reply actions

Awwww!

Hopey and Maya are adorable! BabyW hates Kendall, but as you said, she doesn't take heart and grit into account.

Stephen Colbert: "Kids who look up to athletes become athletes themselves."
The Word: Or bitter sports journalists.

by JLaff on Aug 22, 2006 9:51 PM PDT reply actions

Mr. Flores

That Ron Flores guy's Winexp surprised me almost as much as A-Rod's WE (I guess traditional stats don't tell the whole story). As for that Ron J. Flores guy, he just sucked.

On another note, is there a place where I can look up the number of inherited runners a bullpen pitcher allows?

"Loyal? I'm the most loyal player money can buy." - Don Sutton

by vignette17 on Aug 22, 2006 11:04 PM PDT reply actions

Cool!

Thanks for the WinExp updates, I love poring over these.

I knew Teahen was doing well this year, but 9th in the AL? Much better than co-Moneyballers Swisher and Blanton - Impressive.

by nickolai on Aug 22, 2006 11:07 PM PDT reply actions

Daaaad! Baby Winexp called me a moron!

I think I'll just keep watching the games... LOL

They're urban raccoons, and they're not afraid.

by Poppy on Aug 22, 2006 11:20 PM PDT reply actions

SPORK H8R!

"the reason why hitting the ball on the "sweet spot" of the bat feels so good is that the ball is contacting at the bat at a vibrational node." - salb918

by McFood on Aug 23, 2006 11:10 AM PDT up reply actions

Great work

Thanks for sharing

Mieses St?ck Baseball Fleisch

by elcroata on Aug 22, 2006 11:22 PM PDT reply actions

This is AWESOME!!!

Hey, Apricot...for me, will you do one last one?

Can you do a CAREER one for Jeter vs. A-rod, either entire careers, or just the years they played on the Yankees? A-rod is severely underperforming this year & I'm curious what career stats look like.

Great job; I love it!

"Why won't everyone do what I want???" ~SportySpice

by baseballgirl on Aug 22, 2006 11:47 PM PDT reply actions

you ask, I give

Here are 2004 and 2005 which are the easiest for me to do. As a bonus, I threw in Jason G.

2005 below.

2004 below.

by Apricot on Aug 23, 2006 8:57 AM PDT up reply actions

THANK YOU!!

You're the best!

Now, that was more in line with my expectations. A-rod IS having an off-year; that's too bad. Heh.

"Why won't everyone do what I want???" ~SportySpice

by baseballgirl on Aug 23, 2006 9:02 AM PDT up reply actions

Question...

If I'm reading this right, Loaiza hitting > Loaiza pitching?

by suprjo on Aug 22, 2006 11:48 PM PDT reply actions

No baseline.

Apricot, perhaps you can help brainstorm this issue. Milton Bradley's inability to contribute due to injury results in a flat line where he neither helps nor harms the team by the WinExp framework. This troubles me, since what this tells me is that the baseline WinExp is 0.00. The problem with that is that it does not correspond, at least in my mind, to any reasonable definition of baseline performance. Baby Dubbs suffers greatly from this lack of context. While the baseline could be anything from average to replacement level, do we know - or can we know - what WinExp a baseline performer would contribute? I suspect that the answer is not zero. This baseline would then help us contextualize performance by placing it above or below some expectation. My guess is that we can do this by averaging the leverage of all situations and then use a leverage -adjusted OBP/SLG to guess at what a baseline performer would contribute.

Lastly, because you would agree that WinExp is far from predictive, I hope that you agree that a trade evaluation based on the same is a fun toy but not something to be used as a ex post facto justification.

Great work, Apricot. If it's not too personal a request, might I suggest that you spend less time being a father, husband, and professional, and more time being an A's geek? Thanks.

Stat Wonk Futurist

by salb918 on Aug 23, 2006 7:39 AM PDT reply actions

poop

Yes, baseline is an issue mainly because there isn't a great way to understand winexp contributed as a rate. So I throw in the play totals to give some perspective. We should brainstorm more about this when I'm not running out the door.

Also, the trade evaluation definitely is for fun while all the players are active since winexp is probably not very predictive (though it does surprise me how consistent it is... I'll have to analyze how predictive it really is someday). When they're all retired, I think it's a little more meaningful, but even so it doesn't account for opportunity costs, short-term optimizations in context, etc. etc. Sadly, winexp does not explain everything.

by Apricot on Aug 23, 2006 9:01 AM PDT up reply actions

sal

i got a little confused with your post, but are you almost trying to find out what the VORP of the winexp is, or atleast, what the RP value would be, aka, "what WinExp a baseline performer would contribute?" and if so, why? what bothers you about inactivity contributing 0 to the players winexp?

"If people don't know who he is, they'd better turn on the television and check him out."

by jacobo2u on Aug 23, 2006 12:37 PM PDT up reply actions

You understand me correctly.

"what bothers you about inactivity contributing 0 to the players winexp?"

Because any statistic needs to be contextualized to a baseline. Knowing a player hits .370 is only important when taken in the context of what a typical batting average is. Inactive players contributing zero is a problem because I suspect that the average player contributes more or less than zero WinExp on a rate basis.

Stat Wonk Futurist

by salb918 on Aug 23, 2006 12:52 PM PDT up reply actions

but inactive players...

also have a batting average of .000...(using the three dots there is confusing, looks like decimal points...there, now we're good). I know you're talking about projecting what a replacement level WE would be, kind of, but you couldn't just assign that to inactive players...you would simply use that as a baseline that applies to the entire function...right?

I am no drunkard, I'm a hard-working man.

by Cutthemullet on Aug 23, 2006 1:30 PM PDT up reply actions

You could do either.

You could either assign baseline WinExp to inactive players a la VORP or you could view a player's WinExp only in the context of what a baseline performer would have done given the same opportunities (i.e., leverage).

I prefer the former option, because -- while an inactive player does not see a change in his rate or cumulative statistics -- his team must replace his production via a substitute. That is why the baseline framework is so valuable: inactive players get dinged for their inability to stay on the field.

The latter option would work as well, but would get cumbersome because the baseline would be different for every player. You have to adjust for leverage; otherwise, we'd be comparing Marco Scutaro with Jason Giambi without accounting for the fact that that Big G comes up with men on base all the time and Marco rarely does.

Stat Wonk Futurist

by salb918 on Aug 23, 2006 1:39 PM PDT up reply actions

correct me if I'm wrong

but the baseline for VORP is simply zero, right? Why not establish the same for WinExp...in fact, isn't that the default one being used here? But shouldn't it be adjusted from year to year, based on the overall relationship between hitting vs pitching in that year? Forgive me if my understanding of the concept is not very thorough, because I don't think it is.

Also, Apricot, could you post a league worst WinExp list? Something tells me Rondell White should be up there this year (he's on the DL now, but he was having a horrible year hitting high (cleanup, initially) in the order of what's become a good offense.)

I am no drunkard, I'm a hard-working man.

by Cutthemullet on Aug 23, 2006 2:06 PM PDT up reply actions

Yes.

correct me if I'm wrong but the baseline for VORP is simply zero, right?

Yes.

Why not establish the same for WinExp...in fact, isn't that the default one being used here?

The "zero" for VORP has a definition (replacement level). The zero for WinExp doesn't have a definition -- at least, I didn't recognize the defintion until andeux pointed it out below.

Stat Wonk Futurist

by salb918 on Aug 23, 2006 2:14 PM PDT up reply actions

Average

WinExp by its nature is already comparing players to average (or, to be a little more precise, to how other players in the same situation performed, on average). I agree that for purposes of assessing value, those stats that compare players to replacement level (like VORP) are more useful than those that compare players to average (like WinExp) or to an artificial Zero level (like win shares). After all, a player who contributes a net winexp of zero over 600 plate appearances or 200 IP is quite valuable (probably worth at least $4 million on the free agent market), while Keith Ginter, who has contributed a net winexp of zero by virtue of being not good enough to make the team, has been somewhat less valuable.

There's also the related issue of opportunity cost. How do you compare players like Harden or Bradley who has put up a high WinExp in a short period of time to others who put up a similar WinExp in a full season. On the one hand, the former players have a higher "WinExp rate" so can be said to be better in some sense, but on the other hand their absences due to injuries force the team to give playing time to other, presumably below-average, players. I don't think we can really expect something like WinExp to give a single answer that encompasses these different notions of value.

We hold these truths to be self-evident / We drink beer in bars and act irreverent - Lambchop

by andeux on Aug 23, 2006 2:06 PM PDT up reply actions

Rosetta Stone

WinExp by its nature is already comparing players to average (or, to be a little more precise, to how other players in the same situation performed, on average).

A-ha. I get it now; given the way that WinExp is derived from historical data this seems obvious and I should have thought of it before. The question is -- and I suspect the answer is yes -- does this translate through the play-by-play data so that WinExp contributions are automatically baseline adjusted, or does it still make sense to view a player's WinExp only in the context of the average leverage of his PAs?

Stat Wonk Futurist

by salb918 on Aug 23, 2006 2:12 PM PDT up reply actions

I can have these on my desk

or on the screen and it looks like I am "analyzing"....cool

the great playoff miss of 2004 followed by the good try of 2005...and in 2006 don't make me sick.

by ak_A on Aug 23, 2006 7:44 AM PDT reply actions

I love it

I can look at these all day and not worry about nosy people sneaking a glance at my screen.

by boilerdan on Aug 23, 2006 7:55 AM PDT up reply actions

least valuable player

"Least valuable players: Kendall beats out Ellis and Loaiza in a runaway race to the bottom."

well, kendall is -1.778 over 109 games, whereas melhuse is -1.303 over just 40 games.
perhaps if melhuse had played in 109 games, his WinExp contribution would be -3.55?

A's v Giants "is kind of like the difference between going to see the Ramones and going to see the Bee Gees. A's fans will go see the Ramones." -BB 07/27/05

by xbhaskarx on Aug 23, 2006 7:52 AM PDT reply actions

circuit breakers

It is very impressive to rack up a big negative winexp since if you really stink you don't get playing time.

As far as projectability... that's a fuzzy area that I'd like to look into someday. Also see Sal's comment above about baselines...

by Apricot on Aug 23, 2006 9:03 AM PDT up reply actions

well

projectivity may not be needed when you could have a winexp/PA or game, which would tell the story just as well.

"If people don't know who he is, they'd better turn on the television and check him out."

by jacobo2u on Aug 23, 2006 1:00 PM PDT up reply actions

ok, i might as well...

Kendall -1.778/465= -.00382winexp/PA
Melhuse -1.303/119= -.0109winexp/PA

So Melhuse is, acourding to winexp, 2.85 times worse than kendall. When you compare the twos OPS', I guess the only logical conclusion is that kendall has a LOT more grit!

"If people don't know who he is, they'd better turn on the television and check him out."

by jacobo2u on Aug 23, 2006 1:11 PM PDT up reply actions

I like that Apricot's thoughtful

analysis is being outrecommended by Cutthemullet's drunk diary.

Stat Wonk Futurist

by salb918 on Aug 23, 2006 9:15 AM PDT reply actions

There, I just recommended Apricot's.

Just because I'm a moron doesn't mean I can't tell that this diary's awesome... ;) (as opposed to the drunk diary, which is merely hilarious)

They're urban raccoons, and they're not afraid.

by Poppy on Aug 23, 2006 9:55 AM PDT up reply actions

Killjoy.

Stat Wonk Futurist

by salb918 on Aug 23, 2006 9:55 AM PDT up reply actions

That's my middle name.

No, wait... ForgetTheStats is my middle name. Uh... Killjoy is my Confirmation name.

They're urban raccoons, and they're not afraid.

by Poppy on Aug 23, 2006 10:00 AM PDT up reply actions

Is Lovejoy your antique seeking name?

"the reason why hitting the ball on the "sweet spot" of the bat feels so good is that the ball is contacting at the bat at a vibrational node." - salb918

by McFood on Aug 23, 2006 11:14 AM PDT up reply actions

I found this sequence interesting:

ANA Vladi_Guerrero 3.58 541
KC Mark_Teahen 3.49 385
NYM Carlos_Beltran 3.49 501

Didn't KC trade Beltran for Teahen, John Buck and Mike Wood? Looks like a pretty decent trade!

Day to day

by oblique on Aug 23, 2006 11:33 AM PDT reply actions

Great diary

This is fascinating, and I love the visuals. I have a couple comments/questions:

If I understand the methodology correctly, you're not adjusting for defensive position and are only looking at errors to evaluate defense. If so, you're really just looking at the most valuable batters and pitchers, not players.
Why do you think there's such a large difference between your metric and WPA? WPA as Haren and Zito as the two "most valuable" players on the team, with Haren well ahead of everyone else.

by Danny on Aug 23, 2006 3:01 PM PDT reply actions

great questions

Errors aren't factored in at all. I punt on defense because it's generated automatically from play-by-play and I don't have a reasonable way to assess defense from simple play data. If Retrosheet ever gets zones worked out for all games, there may be a fighting chance. There are some other ideas I've written about before by Gleeman that are reasonable, but I just haven't done anything about it.
This is an issue of religion and bugs.

First of all, the data I printed uses historical results to assess win prob for every situation. This means there is bound to be some error from the "true" win prob, but every method has a form of this issue, so part of it is religious preference.

Secondly, I use play-by-play from Sportsline.com which has bugs. It is certainly missing a few plays from various games.

Third, I did add a beta ability to run the numbers using Studes's spreadsheet which I think works 90% well. Using that, I get somewhat different numbers, (e.g. Zito goes from +1.33 to +1.23). Studes spreadsheet is what Lookout L and Fangraphs, in theory, use to calculate WPA.

Finally, where do you see Zito is most valuable? I find that pretty hard to believe, even given the bugs in Baby W. I'd like to compare with whatever source you're using.

by Apricot on Aug 23, 2006 4:46 PM PDT up reply actions

Fangraphs

Fangraphs has:
Haren* 2.66 Zito* 2.07 Swisher 1.81 Thomas 1.73 Calero 1.67 Duchscherer 1.63 Street 1.25 Bradley 1.17 Payton 1.14

I would have expected differences due to the "religion" issue to be at most .5 (if not much smaller).

We hold these truths to be self-evident / We drink beer in bars and act irreverent - Lambchop

by andeux on Aug 23, 2006 5:08 PM PDT up reply actions

okay, I looked into this

Fangraphs does not (as far as I can see) break plays down into a play-by-play, so it is hard to find discrepencies. I did a game by game compare and nothing jumped out.

Fangraphs gets +2.07.
Historical +1.33
Studes, adjusting for run environments of each park, +1.23
Studes, NOT adjusting for run environment, +1.78

If I were forced to guess, I'd guess they didn't adjust for park effects, gave Zito a break on a play or two that was called an error (it's hard to parse play-by-play for complex plays where there are hits and errors; Baby W punts and calls most of those plays errors).

In general, my automatic numbers are closer to theirs if I DON'T adjust for park effects. Therefore I like my numbers better, even if they are slightly wonky. :)

(The Studes spreadsheet lets you adjust the run environment for each game, which effects its model for win probability.)

by Apricot on Aug 23, 2006 5:59 PM PDT up reply actions