WS/162 in New Bill James HBA (2001, 2003)

[Paul Wendt]

This goes back to "BBF Progressive HoF Election: 1955", page 3, #60-63, and for me back to some now anonymous participants in the Hall of Merit discussions (2002-2007). Thanks to Domenic and BlueBlood.

Bottom line:
Win Shares per 162 games played or "Per 162" as published in both editions of the New BJHBA (2001, 2003?) is full of arithmetic errors. Don't rely on WS/162 in either edition. Under the label "Per 162" it's the final column in each of the big tables that rank players by fielding position.

I don't know whether James wrote the prose between the tables by reference to WS/162 correctly calculated or by reference to the tables.


Details:
2009-07-25, Domenic quoted SavoyBG and replied.
http://www.baseball-fever.com/showth...21#post1573621
>>
> WIN SHARES per 162
> Doerr - 22.61
> Gordon - 21.07

The WS/162 calculations from the Bill James New Historical Baseball Abstract are incorrect in this case...

Joe Gordon had 242 Win Shares in 1566 games... that equates to 25.03 WS/162. Bobby Doerr had 281 Win Shares in 1865 games... that equates to 24.4 WS/162.

Gordon also bests Doerr in WARP3 per 162 (7.0 to 6.3), three best seasons by Win Shares (85 to 81), five best seasons by Win Shares (134 to 127), and OPS+ (121 to 114). Both receive A's in DWS.
__________________
This is an actual case of the truth.
<<

He's right. That is an actual case of the truth.
--and true of the not-really-revised second edition.

Domenic #63
>>
I don't actually have the Abstract in front of me - I found their Win Share numbers by searching around B-F, incidentally. Now I'm curious to see if there are further errors of this nature, or if this is some sort of isolated incident.
<<

Paul Wendt #67
>>
From Hall of Merit discussion threads I recall vaguely that there are many errors in the "per 162" columns of tables.

I suppose that data management was co-author Jim Henzler's department. Did Bill James routinely, sometimes, or never work with the miscalculated rates? Somewhere the text may be revealing.

There is a second edition with some corrections. Maybe this one? I have asked one reader.
<<

BlueBlood owns the second edition (2003?). I directed him to Domenic's note (quoted here, bold), confirmed that it's true regarding the first edition (2001), and asked about the second.

BlueBlood reports that these particular arithmetic errors are reproduced in the second edition, which merely has a "postscript at the end of the book which included James' revised rankings. As far as I can tell, he didn't actually update anything else in the book."

That's amazing given the harsh reviews of the first edition as a hasty production.

[Domenic]

I feel famous.

This is such simple math, particularly in comparison to the other calculations required by Win Shares and the like. I'm actually curious to see if James uses some sort of pro-rated games played or something of the sort - it's doubtful, and I'm not usually one to give the benefit of the doubt following repeated errors . . . but I'm curious nonetheless.

[Paul Wendt]

That is your picture to the left of your username, right?

BlueBlood replied only this morning, perhaps using Veterans' Day to catch up on internet this and that? I admit that I had to use the forum search feature to find the old thread, and I remembered "Savoy" rather than "Domenic", but I won't misremember at again!

[Domenic]

If by me, you mean Albert Camus, then yes.

I don't post too frequently in the Hall of Fame sub-forum, with the exception of ballots in those projects that I vote it . . . so, it's understandable that you'd forget my name.

[Ubiquitous]

The numbers/rankings in the book are largely gobblygook. I believe in the revised edition he does write that he made some glaring spreadsheet errors. If you follow James' formula you will discover that for the most part the rankings don't make sense. Even his examples of how the formula works don't make sense.

[Paul Wendt]

Are you able to be specific?

[Ubiquitous]

Go to google books or open up a copy of the book and look at the follow along examples he gives. Pick a spot in any of the rankings and do the formula for ten or so guys in that ranking. You'll quickly learn that it is all gobblygook. The formula means nothing.

[Paul Wendt]

Are you able to give a page number for one recommended example?

[Ubiquitous]

Go to google books, look up the abstract, turn to the chapter in which he shows you the formula, follow along. You'll see the numbers he quotes and the actual numbers are not the same. Look at the rankings for the first 20 or so first basemen. Do them yourself. You'll see that the rankings don't match up with with what you did. Go to the secondbasemen, do the top three, you'll see that Bill James' statements in that section are false in regards to the rankings. Go to the CF section, do the top three, again you'll see the formula you use gives a different answer than in the book. The formula is meaningless for the most part in creating Bill's rankings. They don't mean anything.

[BigRon]

Go to google books, look up the abstract, turn to the chapter in which he shows you the formula, follow along. You'll see the numbers he quotes and the actual numbers are not the same. Look at the rankings for the first 20 or so first basemen. Do them yourself. You'll see that the rankings don't match up with with what you did. Go to the secondbasemen, do the top three, you'll see that Bill James' statements in that section are false in regards to the rankings. Go to the CF section, do the top three, again you'll see the formula you use gives a different answer than in the book. The formula is meaningless for the most part in creating Bill's rankings. They don't mean anything.

I'm not at home, so don't have access to James' Handbook or WS data right now. So, I can't comment on whether I agree that his WS/162 calculations are in error.

THat said, I'm confused by your statement. James does not use WS/162 to rank his players. As I recall, he uses some combination of WS/162, top 5 WS seasons. top 3 consecutive WS seasons, timeline modification, and "soft" evidence- or something like that. I don't know what weight he puts to each of these elements. But, HE DEFINITELY DOES NOT USE WS/162 as the sole criteria for ranking players.

[Ubiquitous]

I'm not saying he only uses WS/162.

[AstrosFan]

James said he uses a subjective element in the rankings in the NHBA, which accounts for why the formula doesn't match up with the rankings. When you say the formula means nothing, are you arguing that he used the subjective element just to fit the rankings to his preconceived ideas, rather than having the formula as the foundation of the rankings, then adjusting based on his ideas of weighting factors that the numbers cannot capture?