A top 42 list based on Win Shares
Gus Mahler
Bill James recently published two books, The New Bill James Historical
Baseball Abstract, and Win Shares. The first book attempted to rank
the 100 best players at each position, as well as rank the top 100 of
all-time. In order to do so, he developed the Win Shares method.
However because the Win Shares method is so complex, another book was
needed to explain the system.
The Win Shares system is James's attempt to rate players based on how
many wins they contributed to their team. The formula is very
complicated, but the result is that a season of 20-29 win shares is an
All-Star type season, a season of 30-39 win shares is an MVP type
season, a season of 40+ is exceedingly rare (27 times in history) and
50+ has only been achieved 9 times.
Rather than just rank the players by their total number of win shares,
the Historical Baseball Abstract describes a system which combines
peak value and career value. The components are:
1) total win shares (harmonic mean of career win shares/10 and 25)
2) average win shares in the top 3 seasons
3) average win shares of best 5 consecutive seasons
4) average win shares per 162 games (per 43 starts for pitchers); and
5) era adjustment (year of birth - 1800 divided by 10). This is to
control the skew towards players in the first half of the 20th
century.
There is also a subjective component which takes into account elusive
factors such as clutch performance and, e.g., rates Mays ahead of Cobb
for reasons which aren't explained and downgrades Cy Young and others
for pitching before 1893.
I took his list of the top 40 players in career Win Shares from the
Win Shares book, added in the two players in the top 30 from the
Historical Baseball Abstract that were left out (Grove and DiMaggio),
and calculated the total based on the above formula.
The reason I did this was to see where Bonds would end up. The
Historical Baseball Abstract had Bonds rated at #16 (including two
Negro league players ranked ahead of him), but didn't include the 2000
or 2001 seasons. I was curious as to how he would have been ranked had
the 2000 and 2001 seasons been taken into account. By my calculations,
he should have been at #10 at that point (not including the two Negro
league players). After the last two seasons, he is at #6. Here is the
list:
1 Ruth 190.72
2 Wagner 179.55
3 Mantle 178.37
4 Williams 176.39
5 Cobb 174.14
6 Bonds 171.58
7 Speaker 169.04
8 Johnson 167.06
9 Keefe 166.78
10 Mays 164.56
11 Musial 162.06
12 Gehrig 161.05
13 Nichols 160.99
14 Hornsby 160.30
15 Morgan 160.26
16 Collins 156.66
17 Young 156.21
18 Aaron 155.94
19 Alexander 151.66
20 DiMaggio 150.89
21 Schmidt 150.89
22 Lajoie 150.23
23 Robinson 148.47
24 Ott 147.72
25 Matthews 146.50
26 Henderson 145.64
27 Foxx 145.63
28 Yaz 142.83
29 Grove 142.57
30 Mathewson 141.90
31 Rose 139.58
32 Brett 138.96
33 Jackson 136.39
34 Crawford 135.58
35 Ripken 135.49
36 Yount 134.22
37 Waner 132.97
38 Murray 130.83
39 Molitor 128.01
40 Kaline 127.11
41 Winfield 126.07
42 Spahn 122.72
The problem with his formula is that it weighs peak value too heavily.
If you read the original edition of the Historical Baseball Abstract,
you'd notice that he rated players based on peak skill and career
value. The formula given above from the new edition has 4 statistical
components plus an adjustment for year of birth. Of the 4 components,
two of them are directly related to peak value (the top 3 *IS* peak
value while the top 5 consecutive seasons component is peak value +
consistency) and a third (average value), is also related to peak
value (though it also serves to balance out injury seasons.)
Thus, a player like Tim Keefe can rank very highly (#9 all-time). Why?
Because he had an incredibly high peak (highest top 3 score of
all-time, second highest top 5 score). The fact that he had a
relatively short career and had a relatively low average win shares
doesn't make that much of a difference (he had a low average because
the average win shares is based on a 43 start season, but Keefe
started an average of 58 games per year during his 5-year peak). The
case of Kid Nichols (#13) is similar. Since they both pitched in the
19th century, James makes an adjustment in the "subjective" category,
but that is not something we can quantify.
The career value (total win shares) is a much smaller factor. For
example, the reason Bonds jumped up from 10 to 6 wasn't because his
career win shares was increased, but because his peak value (top 3
seasons) had a large increase (his 2001 season is the 4th best of
all-time according to Win Shares). Unless Bonds surpasses 40.78 win
shares in the next few seasons, it is unlikely that he will increase
his ranking. In fact, if he deteriorates (as most players do), he may
drop down, as his average win shares per 162 games will decrease. His
average would have to drop by two win shares per 162 games for him to
drop to #7, so that may be unlikely to happen.
It should be understood that this list isn't THE top 42 players of
all-time, as it only includes players with a lot of career win shares.
Because of the categories, one with a short career that didn't have
enough career win shares to be included in the top 40, but had a high
peak value, can sneak into to the top 42.
Anthony Giacalone
I think that you miss the boat here. First, I would quibble as to whether
13 year career was "relatively short" as you say about Keefe. Secondly, if
Win Shares was really biased towards high peak players wouldn't we have seen
a very different list. Where is Koufax, Dean, Greenberg? They don't make
the list. However, we get guys like Aaron, Yaz, Rose, Ripken and Spahn on
the list, who had long, long careers with relatively low peaks. I can't
remember if that is Jackie or Frank at #23 but aside from Keefe and Jackie,
if that is him, then who exactly are you thinking about that gets a boost by
the alleged bias towards peak?
Anthony
"Gus Mahler" <gusm...@hotmail.com> wrote in message
news:E17LZIw-...@ns15.u-build-it.net...
tom viscelli
Part of the problem--that James does admit--is the arbitrary choice of
"best three" and "best consecutive five" seasons. I don't have the
books in front of me so I can't go into individual cases, but the
system defines a bias towards players with exactly those strengths on
their resumes. While those players you mentioned do appear towards
the bottom half of the list (I know, I know. "Bottom half" of the top
echelon of all-time greats doesn't mean much), the top of the list is
dominated by players with slightly shorter, brighter careers. It's
only a bias though, not enough to keep long-term greatness completely
off the list.
tom
"Anthony Giacalone" <giac...@ameritech.net> wrote in message news:<E17ZIDr-...@ns15.u-build-it.net>...
> I think that you miss the boat here. First, I would quibble as to whether
> 13 year career was "relatively short" as you say about Keefe. Secondly, if
> Win Shares was really biased towards high peak players wouldn't we have seen
> a very different list. Where is Koufax, Dean, Greenberg? They don't make
> the list. However, we get guys like Aaron, Yaz, Rose, Ripken and Spahn on
> the list, who had long, long careers with relatively low peaks. I can't
> remember if that is Jackie or Frank at #23 but aside from Keefe and Jackie,
> if that is him, then who exactly are you thinking about that gets a boost by
> the alleged bias towards peak?
>
> Anthony
> >
Gus Mahler
> "Anthony Giacalone" <giac...@ameritech.net> wrote in message news:<E17ZIDr-...@ns15.u-build-it.net>...
> > I think that you miss the boat here. First, I would quibble as to whether
> > 13 year career was "relatively short" as you say about Keefe. Secondly, if
> > Win Shares was really biased towards high peak players wouldn't we have seen
> > a very different list. Where is Koufax, Dean, Greenberg? They don't make
> > the list. However, we get guys like Aaron, Yaz, Rose, Ripken and Spahn on
> > the list, who had long, long careers with relatively low peaks. I can't
> > remember if that is Jackie or Frank at #23 but aside from Keefe and Jackie,
> > if that is him, then who exactly are you thinking about that gets a boost by
> > the alleged bias towards peak?
> >
> > Anthony
I just entered in Koufax's numbers. Can you spell Overrated? Here
are his win shares by season:
3
1
6
7
9
9
20
15
32
24
33
35
For his first 6 seasons, he was just dreadful. Below replacement
level. Next two seasons was a decent starter, followed by 3 great
seasons, that really aren't at the level of all time greatness (James
admits that batters get higher totals than pitchers. So his career
total of 194 win shares is less than the best 5-year total of Wagner,
Keefe, Ruth, Williams, Walter Johnson, Mantle, Cobb, Speaker, Nichols,
Cy Young, Joe Morgan, or Willie Mays. Even his best 3 years is only
an average of 33.3 per year, which is only 39th on this list. In my
latest list, in which I have added Koufax and Biggio to make this a
top 44 list, Koufax is #44.
Gus Mahler
> "Anthony Giacalone" <giac...@ameritech.net> wrote in message news:<E17ZIDr-...@ns15.u-build-it.net>...
> > I think that you miss the boat here. First, I would quibble as to whether
> > 13 year career was "relatively short" as you say about Keefe. Secondly, if
> > Win Shares was really biased towards high peak players wouldn't we have seen
> > a very different list. Where is Koufax, Dean, Greenberg? They don't make
> > the list. However, we get guys like Aaron, Yaz, Rose, Ripken and Spahn on
> > the list, who had long, long careers with relatively low peaks. I can't
> > remember if that is Jackie or Frank at #23 but aside from Keefe and Jackie,
> > if that is him, then who exactly are you thinking about that gets a boost by
> > the alleged bias towards peak?
> >
> > Anthony
I just entered in Koufax's numbers. Can you spell Overrated? Here