Winexp: The Season So Far in Graphs
by Apricot on Jun 13, 2006 6:52 PM PDT
Email
PrintHow time flies. Baby Winexp was born September 3rd and she could barely do anything at the time. She could read a Yahoo Sports game log and figure out where baserunners are. How she has grown, and now that she is nine months old, she has just gained the ability to pound out simple graphs comparing player contributions over time. It is damned fun to play with.
These pictures show for every date what a player's TOTAL contributions to the team have been this season, so their graph will dip after bad games and go up after good games. Staying level is not a terrible thing. If a whole team played at win expectancy contributed = 0.000, they would play .500 ball, which is not something the A's can look down on this year.
(What is win expectancy contributed? Also, search the archives for diaries on Winexp.)
So when the year started, we were hoping Chavy, Bradley and Thomas would rake. Here's what's happened so far.
Indeed, Chavy finally had a stunner of an April (to be exact, the last half of April) everyone wanted. Unforunately, ever since then Chavez has actually been a slight drag on the team. Milton hasn't made much of a mark so far, and The Big H steadily sucked until the end of May when he went crazy and made up for all his suckitude by entering the positive side of the chart.
We also were counting on Swisher, DJ and Bunnies Crosby to mature.
Well this year, of course Nick Swisher's world-beating April jumps out at the eye. Along with Chavez, he was the reason the A's didn't collapse in April. Indeed, Baby W estimates his contribution at his peak at ~2.2 WXC, which is worth about 4.5 games above .500. Unfortunately, since the start of M*y, he's levelled off, but at least he hasn't gone in the tank. Bunnies has had a so-so year, and DJ saved himself a trip back to the farm with his recent streak. Indeed you can see from the graph he's been a consistent drag on the team except for one great little burst at the end of April. Let's hope Good DJ stays.
In contrast, here's how our starting pitching has been this year (Harden omitted to avoid uncontrollable sobbing from me).
Cupcakes started fast, then drifted below par until the end of May when, threatened with demotion, he made up for most of the harm he'd done by nearing the 0 mark. Loaiza we all know about, stinking it up hideously and then possibly beginning a comeback recently. While Cupcakes and Loaiza treaded water in May, Zito and Haren turned on the afterburners to keep us afloat while the hitters were stinking it up.
Finally, whatever happened to our bullpen? We were counting on Street with Kiko, Duchscherer and Kennedy setting up.
Well, Duke and the terrifying but effective Joe Kennedy flatlined on the graph due to injuries, and Calero has ups and downs but ends up treading water in value. In reality, it's Street's performance that has most closely tracked the team's play. A so-so April went out of control with a game Huston singlehandedly lost, followed by a so-so May that went down the tubes with Huston's bad stretch, followed by handsome Good Huston returning in June just in time to beat the world (and most satisfyingly the Yankees). Huston's contributions to date total about .4 WXC or 1 game over .500, which is exactly where our puzzling, maimed, exasperating, thrilling team is today.
--
By now, Baby Winexp should be called Toddler Winexp, but that's an awful name and I just like the ring of Baby W.
This has been a rollercoaster time in the Apricot clan, with illnesses, a funeral, weddings left and right. Indeed, Baby W's recent growth spurt was the result of Papa Apricot's ongoing fight with strep throat. (Welcome to amoxicillin, mf!) But the most important event of June has been baby Hopey's first birthday.
Somehow in one measly year, she grew from
to
.
I realized pretty soon she'll be asking me awkward questions like why can't she go on dates yet and how exactly do you calculate a Win Share. So I better go and start prepping.
42 comments
|
0 recs |
Do you like this story?
Comments
Very pretty pics
yeah
DJ has already overtaken 2006 mvp crosby?
Awwww!
great
she's very sweet
by californiagirl on Jun 13, 2006 7:16 PM PDT reply actions
awwwww!!!!
Apricot (or Sal, or whoever) , I was just thinking
I just can't think of the right formula right now. Volatility gives the importance of the play in numerical form, so it probably has to do with that, or something.
Some help?
Variance
So generally speaking a relief pitcher who was lights-out would be expected to have lower variance than one who was always living on the edge. A pitcher like Chad Bradford - good at getting double-play grounders, but also gave up a lot of singles - might also have higher variance. I'm not sure how much data you'd have to look at for the differences in this to be meaningful.
So you could do this. But really, if you want to measure something like that, I'd suggest just looking at something simpler like WHIP. Winexp is fun, but it's not the be all and end all of baseball statistics.
Yeah, that's what I'm thinking.
I could just look at WHIP, but WXC is something that takes game situation into account, which is something that no other stat does.
I'll give it more thought when I come back home.
curious questions about winexp
how much win xp is a slam while down 3 with 2 outs in the bottom of the ninth worth? is that the event with most winxp? how much winxp was Pujols blast off Lidge last year in the playoffs?
what event is worth the most winxp for a pitcher and how much is it worth? I would imagine a triple play up 1 with the bases loaded none out in the bottom of the ninth (or, is it with men on second and third?)
are there any events worth exactly zero winxp?
good questions
Dramatic Grand Slams
I happen to know a grand slam in the bot of the 9th, 2 out, down 3, is worth about +.910. This happens to be (within the precision limits of this whole operation) the same as hitting a 2-run HR down 1, bot 9, 2 out. Intriguing, huh? But obvious when you think about it. In both situations, the only run that counts is the one on 1st base. They are equally hard to get home. It is harder to come back from 3 down than 1 down, but if the bases are loaded, then other teammates already contributed some to make the grand slam possible, so the credit gets shared a little.
Pujols v Lidge
Lidge relieved Wheeler,
--. Top 9.0. R:-2. B:. P: Lidge. Rodriguez struck out catcher to first, (0.024, -0.032)
--. Top 9.1. R:-2. B:. P: Lidge. Mabry struck out (0.008, -0.016)
--. Top 9.2. R:-2. B:. P: Lidge. Eckstein singled to left (0.029, +0.021)
--. Top 9.2. R:-2. B:1. P: Lidge. Eckstein to second because of defensive indifference, Eckstein to second (0.014, -0.015)
--. Top 9.2. R:-2. B:2. P: Lidge. Edmonds walked (0.048, +0.034)
--. Top 9.2. R:-2. B:12. P: Lidge. Pujols homered, Eckstein scored, Edmonds scored (0.827, +0.779)
--. Top 9.2. R:+1. B:. P: Lidge. Sanders struck out (0.810, -0.017)
zero winexp
walk
...
though, actually, if you could isolate it to the events where that actually happened, I wouldn't be surprised if the walk had a negative winexp (since a decent number of them would have been in attempts to pitch around a superior hitter)
zero winexp
by matthias on Jun 14, 2006 2:53 PM PDT up reply actions
yes.
I was thinking about less obvious ones like:
Bot of 1st. 0-0. Runner on 1st, no outs. WX= ~63%
Bunt, batter out, runner makes it to third:
Bot of 1st, 0-0, Runner on 3rd, 1 out. WX= ~63.9%
which is pretty close to zero impact within the limits of precision to this thing.
What are the limits? Well, if you check, e.g. walkoffbalk.com, for this last situation, the WX was
63.9% in 1979-1990
65.1% in 1991-1998
61.7% in 2000-2004
That tells me that functionally, one shouldn't take fluctations of 2-3% too seriously. Typical plays cost a team .050 or 5% winning chances, so as a personal rule of thumb, I avoid any analysis that relies on precision past 5%.
In fact, I've been thinking about beginning to report winexp in a different way since non-geeks can't figure out what it means, and geeks make too much of the precision (those number sure do look cool though). My new notion was to convert it to a kind of "win share", so win expectancy contributed would be reported in units of "Games Above .500." So for instance, if a team played a bunch of games and as a whole had a (old-style) win expectancy contributed of about +1.000, that would mean they were 2 games over .500.
So in this new style, instead of saying Chavy has a WXC of +0.987, I'd multiply by 2 and round to report that Chavy has contributed +1.9 games about .500. Then if you add up all the A's contributions, you will get something very close to their actual record of games above .500.
Maybe next week.
As it turns out ...
interesting
I think it's interesting how Calero and Street's years have been so similar--check out between 5/1 and 5/22, especially. Even before and after that range, they were still much more alike than different. Is it because they were put in similar situations, do they have similar pitching styles, or is it just a fluke? My guess is the third choice, but still. It's one thing to arrive at the same total (not that they did, I'm speaking generally now), but another to get there by comparable paths.
Just wait till Hopey can teach Baby Winexp how to account for player performance on top of situation (i.e., in the exact same situation, whether or not the winexp is different for a Marco Scutaro-Johan Santana match-up or a Pujols-Lima match-up, and whether or not previous winexp success affects future winexp potential). Then you'll really feel old.
by spal on Jun 13, 2006 10:23 PM PDT reply actions
situational expectations
I think there's something so clean about the simple approach that I would hesitate to mess with it. When something is simple, it measures what it measures. When something is a godawful hack like Win Shares or VORP then it gets unclear how to interpret it, and you are relying on predictability and correlation studies to believe it corresponds to something.
maybe not situational...
I only start to wonder if it's too simple when something like (as I've heard it stated) the home team automatically has a 54% winexp to start the game. I wouldn't think, for example, that a team like the Royals would be expected to win if they're playing, say, the White Sox, even if it is in K.C. I would think that over the course of 100 games, all set in Kauffman, the current Chicago team would win more than 46 against the current Royals team.
by spal on Jun 14, 2006 9:11 AM PDT up reply actions
very interesting
By using historical (or historical model) data, win expectancy contributed measures in a straightforward way, how did a batter's and pitcher's performances compare to all other such performances in history? There is something elegant about the way that automaticaly takes into account all the weird machinations of the 9th inning because the historical data reflects that. We don't have to decide how much the odds are lower because Street dran Red Bull or it's a road game under lights. We don't decide anything about situational worth... we let history decide.
Another example...
The more I think about it the more it makes sense to just focus on the situation. Scutaro faced a situation, and in that situation the other team chose a certain way to play against him, and in that situation he succeeded. Opposing players vary widely in their skill level, but you're not really facing a player, you're facing a whole team - particularly in the high leverage situation that winexp weights most. Teams don't vary in skill as widely as players do.
Someone mentioned facing a pitcher vs. facing Pujols, but the truth is you only pitch to a pitcher when the opposing manager has decided that that's the best person to send up against you in that situation. So on the whole I think it makes sense to give the same credit.
by matthias on Jun 14, 2006 10:50 AM PDT up reply actions
well-stated
agree
by matthias on Jun 14, 2006 9:18 AM PDT up reply actions
what about Surly Teen WinExp?
heh
% winexp -a 20060615SEA@OAK
Sure. I'll get right on it.
% winexp -a 20060615SEA@OAK
Quit bugging me, I said I'll do it. I'm sooo excited too.
% mv winexp room/
You can't make me! You're not the boss of me!!
% sudo mv winexp room/
I hate you! you're not my real father!
etc.
hmm...
% winexp -a 20060613SEA@OAK
What?
% winexp -a 20060613SEA@OAK
Type louder sonny.
% winexp -a 20060613SEA@OAK
Oh, WINEXP.
% winexp -a 20060613SEA@OAK
What? Oh. The score was 2-0.
% winexp -a 20060613SEA@OAK
The score was 2-0. That's what counts. And steals and bunts. But nothing's more important than grit. ^C[Interrupt not possible]
Hey, listen and you might learn something you little punk. See, in my day, a player played for the love of the game, ^C[interrupt not possible]
NOT endorsements and women. Quiet I say, I have the floor. Now, this was before there were computers and Enron and colored people messing up the sport. In fact, I met Three-Finger Mordeci myself as a wee lad in ...
etc.
Starters vs. Closers
The visiting team scores one run in the top of the first. The score remains 1-0, and in the top of the ninth the visiting team's closer comes in to pitch a scoreless inning for the save. Question: how many innings must the starter have pitched to have a winexp contribution as least as great as the closer's?
Now repeat for a 2-0 games, 3-0 and so on up to a large number. This will answer my next question: a starter needs five innings to get credited with a win. What is the size of lead where he needs exactly five innings to be worth the same as the closer?
by matthias on Jun 14, 2006 7:06 AM PDT reply actions
I meant "bottom of the ninth"
by matthias on Jun 14, 2006 7:07 AM PDT up reply actions
intriguing
Analyzing the closer is easy: saving a 1-run game in the bottom of the 9th is worth about +.190.
To answer your particular questions it's not just a matter of comparing the winexp of a 1-0 lead in the 1st with a 1-0 lead in the 5th. Each inning one side is scoreless changes the amount of winexp up for grabs. So you'd have to sum up the changes for each inning. But approximately speaking, pitching a inning of scoreless ball in a tie game is worth about +.050 in the early innings.
So roughly speaking, holding a 1-0 lead for the first 4 innings is probably worth about +.200, approximately the value of a 1-run, 1-inning save.
ps. One interesting recent example was last night's game where Blanton left a 2-run game in the top of the ninth, no outs, runners 2nd & 3rd for Street.
Total Contributions:
2006-06-13 OAK Huston_Street* +0.292 4
2006-06-13 OAK Joe_Blanton* +0.267 29
Love the A's blanket
by Plunkage @ Athletics Nation on Jun 14, 2006 7:13 AM PDT reply actions
here's an old thread
http://athleticsnation.com/comments/2005/9/5/18329/72554/17#17
The AN store has some babywear I think too...
Hopey's adorable :)
Comments For This Post Are Closed
Secondary Sidebar
User Tools
FanPosts
Community blog posts and discussion.
Certain photos copyright © 2012 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.
