A PHP Error was encountered

Severity: 8192

Message: Non-static method Comment::form() should not be called statically, assuming $this from incompatible context

Filename: libraries/Template.php

Line Number: 3147

Bonds held back by own pitchers!

Bonds held back by own pitchers!

November 06, 2003

Sorry about the tabloid headline.  I guess I’m getting giddy with this whole blogging thing.  Originally, I had a very academic-sounding title for this research article, but the results so surprised me that I thought a different type of headline would be in order.

One of the most frequent criticisms I hear about Win Shares is that there are no “loss shares.”  Now, I have a feeling that everybody means something a little different when they mention loss shares, and I’m not quite ready to go there.  But if we talk about the simpler concept of negative Win Shares, well I can do that.  In fact, it turns out that the concept of negative Win Shares has an impact on Barry Bonds’ total Win Shares.

First, most Win Share followers know that Win Shares for individual players don’t go into the negative.  If a batter’s marginal runs are lower than the 52% threshold, then the batter is credited with zero Win Shares, even if the batter is way below the threshold.  The numbers don’t get negative.  This is something Bill James is rather vehement about.

As you can imagine, this is more of an issue in the National League, where pitchers bat for themselves.  So, Bill James added a wrinkle to Win Shares: Pitchers actually are credited (debited?) with negative pitching Claim Points when their batting Marginal Runs are below the threshold.  Note to self: Huh?

Let me try again.  When calculating pitching Win Shares, each pitcher’s relevant pitching Claim Points are calculated , and then each pitcher’s batting Claim Points are subtracted from the pitching total if that pitcher is below the batting Marginal Runs threshold.  If a pitcher is a good hitter, he gets positive batting Win Shares and all of his pitching Win Shares.  If he’s a poor hitter, he gets zero batting Win Shares, and his pitching Win Shares are reduced.

So, you see, Win Shares actually does include “loss shares.”  They’re just moved to a different bucket.  And they’re only applied to pitchers.  And this seems to have some unintended consequences on the Win Share/Marginal Run relationship we’ve been talking about.

Here’s a specific example.  Moises Alou had 60 Marginal Runs and 18 batting Win Shares for the Chicago Cubs.  Shawn Green had 65 Marginal Runs, but 16 batting Win Shares.  Alou got more credit for less Marginal Runs.

Why did this happen?  I used my new park factor adjustments for this calculation, so that’s not the reason.  Also, the Dodgers and Cubs had similar Pythagorean variances, so that’s not the reason either.

The reason lies with the batting performance of the respective team’s pitchers.  In all, Cub batters (mostly pitchers and Lenny Harris) had 20 negative Marginal Runs, while the Dodgers’ batters (pitchers and Daryle Ward) had 52 negative Marginal Runs.

When a team has more negative Marginal Runs, that team’s batters have to “carry” those negative Runs in their total to make sure the sum of all players Marginal Runs equals total Marginal Runs.  This reduces their batting Win Shares.  The fact that pitchers have negative batting claim points included in their pitching Win Shares doesn’t help the batters, because those negative points have been taken out of the batting bucket, and put in the pitching bucket.  So, in an effort to avoid negative Win Shares, James reduced the validity of the Marginal Runs/Win Shares relationship.

Am I making any sense here?

To correct for this, I took my park-adjusted Win Shares and gave negative batting Win Shares to all those batters with negative Marginal Runs, and negative pitching Win Shares to all those pitchers with negative Pitching Claim Points.  I also didn’t take the negative batting Win Shares out of pitching Win Shares.  And I posted the results here.

It’s a big page.  If you don’t want to see the whole thing, here’s a list of the leaders:

Player     Team     BatWS   FieldWS TotalWS FinalWS  POS    OrigWS  Var
A  Pujols  STL      40.28     2.21   42.49      42    OF       41    1
B Bonds     SF       39.1     3.15   42.25      42    OF       39    3
T Helton   COL       33.2     3.33   36.53      37    1B       34    3
G Sheffel  ATL      32.42     2.95   35.37      35    OF       35    0
J Lopez    ATL      25.16     5.35   30.51      31     C       30    1
J Thome    PHI      28.46     2.37   30.83      31    1B       30    1
B Abreu    PHI         26     2.85   28.85      29    OF       28    1
M Giles    ATL      21.96     6.61   28.57      29    2B       28    1
R Sexson   MIL      25.45     3.23   28.68      29    1B       26    3
C Jones    ATL      24.37     2.06   26.43      26    OF       26    0
D Lee      FLA      22.88     3.69   26.57      26    1B       25    1
E Renteri  STL      21.02        5   26.02      26    SS       25    1
L Berkman  HOU      22.55     3.61   26.17      26    OF       25    1
M Loretta   SD       20.5     5.08   25.59      26    2B       24    2 

See what I mean about Bonds’ pitchers holding him back?  Originally, Bonds was two Win Shares behind Pujols, but now they’re tied.  The reason: there were about 55 negative Marginal Runs on the Giants, but only around 20 on the Cardinals.  Cardinal pitchers were better hitters than the Giants’.  So the phenomenon I’m describing impacted Bonds more than Pujols.

Green and Alou?  They both wound up with 19 batting Win Shares.  Now, the only difference between them is the slightly better Pythagorean performance of the Cubs.

The variance by batter is driven by three factors:
1. The number of negative batting Marginal Runs on each team.
2. Each batter’s proportion of total team batting Win Shares.  Helton accounted for 27% of his team’s batting Win Shares.  Bonds and Pujols both accounted for about 25% of theirs.  So they get more absolute batting Win Shares allocated to them.
3. Basic rounding.

One other quirk: If you look at the entire list, you’ll see that Al Leiter’s total Win Shares actually increased under this new method.  One of the things I found is that certain hitters are so bad, they create negative runs.  In fact, Leiter created ten (!) negative runs in 2003, according to the Runs Created formula used in Win Shares.  He is one bad hitter.

Negative Runs Created are funky and create problems.  For instance, park factors don’t work well with negative numbers.  I don’t believe James intends for there to be negative Win Shares—it just happens due to some of the latest modifications to Runs Created, such as adjustments for clutch hitting.  So I set the minimum number of Runs Created at zero.

So even though Al Leiter has two negative batting Win Shares with this new method, he actually had four Win Shares taken off his pitching Shares with the old method.  And he receives a net gain of two Win Shares.

After performing this analysis, I do feel strongly that Win Shares should include negative totals.  If not, individual batters can see significant changes in their Win Share totals due to inappropriate contexts.

I know this would make Bill James spin in his swivel chair, but the relationship between Marginal Runs and Win Shares should overrule all other considerations in order for the system to remain valid.  In my opinion.  big surprise

As best I can tell, Loss Shares is unworkable.

I’ve taken quite a look, as did Bill James, through what are probably very similar methods. The issue is for hitters that hitting ability diverges from the mean too fast for Loss Shares to be of any real value. To keep the Loss Shares/Claim Points ratio more or less constant, I found I could set the threshold level for Outs/PA to no less than 93% of the league average, and even then I still had a fairly high correlation of LS/CP to actual team Loss Shares. No such problem exists for pitchers.

The best solution is to use an expected Win Share value, and base things as a ratio of that. You can subtract negative Win Shares from this (so you’re adding the positive total of this) to further refine things. You have to do a few tricks to get some things to work right, but it’s better than the alternative, wherein someone with a .390 OBP has no loss shares, nor does anyone else in the top 10 of MVP voting.

Posted by .(JavaScript must be enabled to view this email address)  on  11/10  at  05:48 PM

Hi Charlie.  Thanks a lot for dropping by.  I appreciate your comments.

I haven’t even tried to contemplate Loss Shares.  I don’t know how they would come into play, except I was vaguely thinking of taking your approach.  Maybe I’ll just skip it altogether!

Posted by studes  on  11/10  at  08:05 PM

Someone has looked at this briefly already.

Look under the drawbacks section of this web page


Posted by .(JavaScript must be enabled to view this email address)  on  11/11  at  04:22 PM

Tango has a number of very good observations about Win Shares.  Unfortunately, he’s sort of burnt out on the subject.

He seems to suggest a relatively simple approach to Loss Shares: Everything above 50% is a Win Share, and everything below 150% is a Loss Share.  The reverse would apply to hitting and fielding, I would guess.

Does this make sense as an approach?  I wonder if Charlie looked at it this way?

Posted by .(JavaScript must be enabled to view this email address)  on  11/11  at  05:25 PM

I agree with Charlie that the “clean” way is to have an “average WS” to tail along with the player.

This way, you don’t have any negative numbers, and it’s clear what’s going on.

The question you are answering is: “given this context, what would an average player have done?”

Of course, since the player you are transposing himself contributed to that context, you’ll have a little problem.

Posted by .(JavaScript must be enabled to view this email address)  on  11/13  at  08:32 AM

He seems to suggest a relatively simple approach to Loss Shares: Everything above 50% is a Win Share, and everything below 150% is a Loss Share. The reverse would apply to hitting and fielding, I would guess.

Does this make sense as an approach? I wonder if Charlie looked at it this way?

Much after the point, but yes, same results for hitters. Actually worse. Hitting ability decentralizes too quickly.

Posted by .(JavaScript must be enabled to view this email address)  on  12/11  at  06:15 PM
Page 1 of 1 pages
Commenting is not available in this channel entry.
<<Previous Article:  Historical Gold Glove/Win Share Comparison Next Article:  Clutch Hits - Baseball Primer (November 5, 2003)>>