THE BOOK--Playing The Percentages In Baseball

You can tell by that quote that James thinks that he is still a pioneer in cutting-edge analysis. It’s time for him to get a bit more humble in his attitude in this area, and start being who he is--a great early, ahead-of-his-time, analyst who is now just one of many quality analysts in the field. He should concentrate more on his skill of writing, which is still at the top of the list. Instead of trying to create the best new stats, he should simply use the best existing stats in writing about baseball from a saber viewpoint.

Exactly.  His strength isn’t fine-tuning metrics, or even developing them anymore, though he was the biggest factor in making that widespread - it’s the philosophizing, which he does better than anyone.

He is without question a fantastic writer and an excellent thinker.

In Moneyball, Lewis said that James preferred leaving an honest mess rather than a tidy lie.  I believe in that, and live that myself sometimes.

But, that means that when he leaves an honest mess, and we clean that up, then he’s got to live with our cleanup.  BaseRuns for example, is the cleanup of the mess that is Runs Created.  Bill James should shout that from the rooftops.  He doesn’t, and I don’t know why.

James always asked the best questions.  He didn’t always provide the best answers to those questions.

James, comparing his upcoming book to his sublime Abstracts:

Similarities. . .

1.  Mix of writing with player and team data charts.

2.  Team by team approach.

3.  Spring annual looking back at previous season and ahead to the next.

Differences:

1.  Better design work by the publishers.

2.  Less detailed research.  This always bothered me about the Abstracts, that I always felt that I had to report all of the details of my research, as an obligation to readers, but I always knew that 99% of the people reading didn’t care about the details and anyway couldn’t understand what I was talking about..  The existence of the online enables us to move that heavy lifting stuff over the online, and keep the articles more straightforward, presenting fewer barriers to the readers.  I hope.

Bill James:

I am still working on Win Shares and Loss Shares.  I have done a lot of work on them, and I plan to start sharing that work with you on this site within a few weeks. 

There are some pretty significant flaws in the system, and I don’t have time to back up, un-do the mistakes and start over, so I’m just going to have to share them with you and expose the flaws, and then. . ..hopefully. . .we’ll work on them together.

That’s quite an admission. 

For those who are new around here, I took a critical look at Win Shares, exposing some of its obvious flaws.  (I didn’t even get into the fielding aspect of it.) It culminated with a debate with Rob Wood, where each of us decided to take one side, and give the pros/cons of it:
http://tangotiger.net/#Winshares

I will be very interested to see what Bill does now.  As you may know from another thread in this blog (and makes quite clear in the PDF above), that the whole idea of splitting absolute numbers like Bill doesn’t make much sense to begin with (start at post 22 if you don’t want to read the whole thing):
http://www.insidethebook.com/ee/index.php/site/comments/spread_in_talent/

Read the PDF and thread before commenting on my claim here.

***

Anyway, good for Bill.  I’d like to see if he thinks that Runs Created also has a significant flaw, and if he will finally adopt BaseRuns.  David Smyth is a genius for doing the (now) obvious of separating the HR.

***

Classic Bill James:

There are media guys who try to represent “booing” as saying “We hate you” or “We don’t respect you” or something.  That’s not necessarily what it means.  The crowd has a very limited number of ways to express themselves.  When somebody fouls up, what are they supposed to do, organize a collective frown?  Clap backwards?  They’re paying a lot of money to be there, and you can’t tie their hands.

I doubt very much that James will switch from using absolute numbers in the split. But yes, it will be interesting…

They’re trickling in.  Here are the WS/LS of the players Bill James just noted in his Trammell article:

378-237 Yount
325-231 Ozzie
295-127 Larkin
282-176 Trammell
257-273 Vizquel
173-157 George Bell

Do these make sense?  Let’s take Larkin, who is noted as 295 WS and 127 LS.  That gives him +84 win shares above .500 (211-211 is average).  That’s 28 wins above average.

I think Larkin is probably something like +40 or +50 wins if I were to do it the Linear Weights way.

Ozzie is 47 win shares above average, or +16 wins above average. That’s terribly low.

Yount works out to +23 wins, which is also fairly low.

Based on this, I’m not optimistic that James has figured out how to do Loss Shares…

If however, the WS/LS are more aking to WA/LA (win advancement / loss advancement), and therefore you need both the wins above .500 AND the losses below .500, then you double all those numbers, and James is fine with Loss Shares.

But, it’ll break down at the team level if that’s what he’s really doing.

Somewhere, Loss Shares will make no sense.  As soon as he rolls out his method, I’ll tell you where.

Against average or against replacement?

Larkin is +19 batting wins.  Give him half a win for every year at short and he’s +25.  Without checking my own numbers I’m pretty sure he’s under 100 runs career totalzone, so +30-35 seems right. A little low but not by that much.

Ozzie is -14 batting (B-ref), +8 for SS, +5 for basestealing, and totalzone puts him in the 150-200 range defensively.  Looks like a bingo, if I may borrow the term.

Yount: +21 batting, moved from one premium position to another, and probably average defensively over his whole career.  +8 for position and a positive baserunner.

WS has him a little low, but not horribly off.

Vizquel -20 batting, close to break even stealing, +8 for SS.  If he’s a +90 run defender that would be a perfect match for win-loss shares.

Ok, let’s just keep it per offense, since James breaks it down for us.

Larkin: 223-110 for batting WS-LS, making him +19 batting wins.

Hmmm…

Yount: 293-175, making him +20 batting wins.

Those match the Palmer Batting Wins at b-r.com.

I’ll take it back then.  It looks like he’s at least gotten the hitting part done right.

The question is if he’s got a good balance on the fielding side.

And of course, if he did the pitching right.

***

Rally: I’ve found your fielding numbers, while generally lining up with my list, fairly low overall. That is, the spread isn’t as great as mine.

When you’ve compared to UZR, do you have the same standard deviation (spread)?

I’ll have to check.  Probably not, but I do have some all-time great seasons in the +30-+35 range.

I guess this is a pretty good thread as to guage where the mainstream fan is at, on a sabermetric scale:
http://freakonomics.blogs.nytimes.com/2008/03/25/bring-your-questions-for-sabermetrician-bill-james/

Rally, let’s continue the fielding discussion in the TotalZone thread here:
http://www.insidethebook.com/ee/index.php/site/comments/totalzone_a_new_fielding_measure/

James actually has the year-by-year for a few players… and guess what: NEGATIVE LOSS SHARES!

I can’t believe it.  In 1985, for Ozzie’s fielding, he has him with 10 WS and negative 1 LS.  That’s +1.8 wins above average.

I can’t believe it.  When Win Shares came out, I said:
http://www.tangotiger.net/winsloss.html

So, Pedro is worth about +8 wins over a .500 pitcher. Or +24 win shares over a .500 pitcher.

How many win shares does a .500 pitcher have, given about 24 decisions? Well, if you’ve got 81x3 win shares to distribute, and the pitchers get 35% of that, then all pitchers should get about 85 win shares. Therefore, if you split that up by innings (or decisions, or whatever reasonable thing you want to use), it’s fair to say that a .500 pitcher should have about 12 or 13 win shares. This means that Pedro must have 12+24 = 36 win shares.

This .500 pitcher should also have 12 or 13 loss shares. Pedro therefore would have 12-24 = -12 loss shares.

I will be super excited if James has fallen in line with this idea, and finally accepted coming over to the dark side of negative loss shares.

I remember in an online chat when WS first came out, I asked Bill about negative loss shares, and he responded something like: “If you are getting negative loss shares, then you are doing something wrong.”

I’m glad he seems to have finally come around. Once I see Pedro’s and Bonds’s seasonal lines, I can tell you if he actually did.

What this means, by the way, is that James has now accepted Linear Weights. 

This is the sabermetric equivalent of Nixon going to China.

Cool. Maybe you can put down the axe now.

Yes, finally!

Now, it’s on to set Baseball Prospectus straight…

The only problem I have with this is that some people will take it literally.  If you use his model, than negative losses (and negative wins) are inevitable, and that’s fine, as long as you recognize the limitations of the model.

Some of the less sophisticated folks out there already seem to operate under the assumption that Win Shares represents an “absolute zero” baseline.  The existence of negative wins and losses shows that it does not (and I’m not saying that BJ claims that it does, only that there is a subset of his audience that holds that perspective).  I’m not looking forward to hearing about how “Ozzie Smith was such a good fielder, not only did he add X wins, he also subtracted Y losses!”

What’s wrong with that?

That’s kind of how I explained the concept when I did Basketball win shares a few years ago.  Some players are just so bad that they negate positive wins that their teammates contributed.

I don’t think it’s misleading, and if it adds a little variety to the commentary, I see that as a good thing.

I did the same thing with hockey about 15-20 years ago.  BAck then, we had Gretzky, and he had such a high “win share” and such large “negative loss shares” that it just seemed so absurd to try to explain.  That’s why I abandoned it back then.

I can’t wait to see the negative loss shares for Barry Bonds and Pedro. They will be unreal.

(You can already figure out what they will be.  If Bonds was around +11 wins above average, and if he has 54 win shares, then his loss shares will be negative 12.)

But, you’re right.  That’s how I explain Gretzky, that he’s so good, that he cancels out the loss shares of his teammates.

Again, I don’t have a problem with it terms of the Win Shares system.  It is necessary in within that system.

As long as people recognize that WS is just a model, and understand what negative numbers mean in that system, it’s fine.  It’s just that there will be those out there trying to figure out what W% 20 and (-4) makes, and how to compare that to the W% from 25 and (-1).

I’ll have to look into that.  If memory serves, he told me that was why he didn’t like my approach to WSAB.  Same issue—which we all discussed over at Baseball Graphs several years ago.

Crazy.

Studes, I’ve got to approach it as if he was pretty stubborn, was very anti-LWTS, and eventually relented to our onslaught of logic.

BTW, I can’t look at the article—their website is funky.  But it’s worth noting (if James didn’t) that Trammell’s Win Shares declined from 318 to 282.  So he changed the whole system.

Here’s the actual quote for post 16, from an ESPN chat of 6 years ago (man, time flies):

Tom: Isn’t it possible that the reason that you didn’t show Loss Shares is that you ended up getting “negative” values for Bonds and Pedro, et al?

Bill James: If you ended up with negative values for Bonds or Pedro, you’d be doing it wrong. But I didn’t know how to do it right, so I didn’t do it.

***

Studes, I remember you said that James was mortified that you did WSAB.

Like I said, James has come around it seems.  I said a long time ago that it would take James to denounce his own work, for others to move on.  Basically, his fans place a great deal of trust in him and his process, and if he gets it wrong, it takes the rest of us 20 years to undo what he can undo in 20 minutes.

Anyway, this is a good day.

Once he adopts BaseRuns over Runs Created as his base, his transformation will be complete.

Studes, you are right that his WS are now that low. 

I haven’t had a problem logging in or navigating, but the web experience there is, as I’ve said, rather poor, straight out of 1998.  No good blogging or discussion forum software.  WordPress, vBulletin, phpBB, ExperssionEngine, and countless other, are used everywhere.  I can’t understand why they didn’t do this.  And the formatting of his articles is like a bad MS Word job.  Try cutting/pasting his tables.  It’s impossible.  Clearly, the website is basically a bunch of static, non-linking pages, as if you scanned a book and put it online.  It really bothers me on a technical level, because it’s just so poorly executed.

I know, I know, I should keep quiet, because some form of Bill James is better than none.

You’re right Tango—the site is cludgy.  And the wild thing is that it’s easy these days to implement a non-cludgy site.  Oh well.

Or is that “kludgy?”

I agree, if I can do it, it’s easy. There are many people with much better web skills than I out there.

If you’ve got a bunch of player pages, you should easily be able to jump from players to teams and to other players.

Kinda like this site of fictional players:
http://home.comcast.net/~briankaat/player_1776.html

Studes, you might remember Joe Thomas from a THT article.

I certainly remember the article, Rally.  A classic.

Studes, I remember you said that James was mortified that you did WSAB.

He was definitely mortified when we rolled out WSAA, but I don’t think he understood that it was meant to be a starting point--not an end point.  Not sure I ever got a good read on his feelings for WSAB.  But I do have a vague recollection that he really didn’t like the idea of negative win shares or loss shares.

My take is that any system with a floor above zero should allow negative totals, cause you’ll have problems with your results otherwise.

James just showed Derek Jeter’s year-by-year, and in 1999 he shows Jeter, on offense, with 26 win shares and negative 1 loss shares, which he shows as:
1998 19-6
1999 26+1
2000 18-6

As you can see, since “19-6” would be the standard way to write a won-loss record, he flips the sign with the “negative 1 loss share”.  He also says:

The “26+1” for Jeter’s offense in 1999 means that the positive impact of his offense in that season was greater than the area of responsibility assigned to him as an offensive player. . .in essence, 26 wins in 25 decisions, thus creating a negative loss.  Since a loss is itself a negative, that creates a negative-negative, and I write that as a +1.

In the Win Shares debate I had with Rob Wood:
http://tangotiger.net/winshares.pdf
(I highly suggest you guys read this, as it basically lays out all the problems with Win Shares at the time.)

page 27:

If the average hitter has 9 game slices, that would mean that the average hitter has 4.5 wins contributed.

But does this mean that a hitter has to top off at 9 wins created, or +4.5 wins above average? I think we can both agree that this is not true. That Bonds is actually over 10 wins above average. So, Bonds would have say 15 wins created, but with only 8 or 9 game slices.

To me, this is acceptable, because of the way runs and wins are created. They are not created in a vacuum. They are based on interdependence, and therefore we should NOT expect things to add up nice and neat. We should expect things to add up, but not all nice and neat.. The result of this is that some players will have more wins than game slices. And since the difference of game slices and wins is losses, Bonds has negative losses.

I believe that Bill James has gone through this whole exercise. That he couldn’t fathom negative losses or having more wins than games, and therefore said “I couldn’t get it to work”. He can’t get it to work because of his insistence that wins be bounded by 0 on one side and games on the other.

However, wins is better conceptually as wins above average + share of average wins So, if Bonds is +10 wins above average, and if his PA slices gives him 4.5 wins, then Bonds in this case is 14.5 wins in 9 games.

page 33:

This goes to the heart of the matter, and the reason that I think Win Shares fails. The process to separate a player’s *marginal* contributions in his attempt in trying to help his team towards winning into an *absolute* win contribution value is bound to fail at some point. And that point is that you get absolute win contributions that are greater than absolute game slices. If this can be accepted, then I can accept the premise of Win Shares.

Yr.  WS LS James Palmer
1995 01 01 0.0 -.2
1996 13 11 0.3 0.2
1997 17 11 1.0 0.4
1998 19 06 2.2 2.2
1999 26 -1 4.5 4.9
2000 18 06 2.0 2.5
2001 18 07 1.8 1.9
2002 18 09 1.5 1.1
2003 15 05 1.7 1.8
2004 17 11 1.0 1.3
2005 19 08 1.8 2.5
2006 22 03 3.2 3.0
2007 16 10 1.0 2.0

The first column is the year

The second column is Jeter’s offense win shares.
The third column is his offense loss shares.

The fourth column is James’ implied wins above average.  Take the year 2000, where he has 18 WS and 6 LS.  With 24 game shares, the average would be 12-12.  That makes Jeter +6 win shares.  With 3 win shares per wins, that makes him +2.0 wins.  In shorthand, that’s: (WS-LS)/6

The fifth column is Linear Weights, as reported by Baseball-Reference.com, using Palmer’s version.

As you can see, a fairly strong match.

How does James come up with game shares (on offense)?  Here’s how I had proposed it: Jeter in 2006 had 715 PA, while the Yanks had 6455 PA.  Jeter’s share of that is 11.1%. 

If offense gets 48% of all game shares, then 162 * 3 * .48 = 233 game shares to offense.  And 11.1% of that is 26 game shares.

And in 2006, Jeter had 22 WS and 3 LS, for a total of 25 game shares.  Bingo.

Basically, what James has done is *started* with Palmer’s Linear Weights, converted that to wins above average (WAA).  And then figured out the won/loss record for the average player.  And then added WAA to the wins and subtracted WAA from the losses.  And to finish it off, he multiplies by 3.

Every time I write this stuff, I’m thinking of the scene when Anakin finally gets his mask put on him at the end of the movie, and is told that he killed Natalie Portman.

Transformation: complete.

The student has become the master.