Well, this is a great debate--even if I don't convince anyone at least my research has made people reconsider their assumptions.

Le Samourai, it includes all NL starting non-catchers since 1893, plus other backup seasons "of interest."

Bob Dernier Cri, you know I didn't mean Concepción would be hitting 45 home runs. Seasons like Edgar Rentería's 2003 might be more accurate. But to address the substantive point: I imagine your PHoM does include some catchers and non-Honus Wagner shortstops in it. None of them will have the same BRAA+FRAA as even many mediocre first basemen. Clearly playing SS is harder than playing 1B, and we have to account for replacement level in our rankings. How much harder? Well, that's what my research attempts to answer. Why were infielders in general and SS in particular such poor hitters in the 1970s? I agree with 'zop that there's got to be some reason, it's not just random fluctuation, you see the exact same pattern at four different positions, all derived independently. Does anyone have league groundball/flyball data going back in time?

Joe Dimino, it would be great if you could send that along.

Does anyone have any better ideas for how to separate out replacement level for CF and corner OF? The worst three CF tended to hit equally or even better than the worst three corners all the way to about 1980. Using Nate's data, I've given CF a flat 0.4 win bonus relative to the corners consistnetly over time, but that might not be accurate.

Ultimately I can't see how that's much different from saying that the men's hoops MVP of the Little East conference should be as good an NBA draft prospect as the MVP of the Big East.

This mixes "value" and "ability". If the MVP of the Little East has even a shot at being an NBA prospect, then he probably has more value than the MVP of the Big East where NBA prospects are much more common (excepting high lottery picks, perhaps). This is under the principle of "a conference title is a conference title" ;-) The NBA scouts are looking purely at ability to play in the NBA; conference value, irrelevant. The related question to this is whether Ruth would dominate the 1990's in the same way as he did the 1920's? And if the answer is "No" does that diminish his greatness? Value is always with respect to a context.

Therefore, we have to consider the possibility that there was something inherently difficult about SS defense during the no-hit replacement era that made it a more difficult defensive position compared to other eras, and therefore limited the offensive value of those who played it. Maybe it's much harder to play SS on turf if you're a big guy, and the widespread turf fields forced larger players to the OF and 3B.

NL was THE turf league. IIRC, Montreal, Philly, Pittsburgh, St. Louis, Cincy, and Houston all had rugs. Half the league. The Mets had grass but played 30% of their games on phony-turf.

AL was the grass league. Originally, just KC, add Toronto and Seattle, then Minnesota. Don't remember any others, offhand (White Sox for a few years?). Even with 4 turf opponents, the Red Sox are playing only 15% of their games on plastic.

I find this very intriguing, but part of me is loathe to accept that a weak bottom level of SS's (or a strong bottom level of 1B) shoudl really effective how a player is viewed in a HOM context. As a GM or someone trying to critique salary disbursement, it is extremely valuable. But we are trying to figure out the best players of all time across eras, not necessarily within one.


Exactly, this is why Win Shares or WARP with a constant replacement level is more meaningful. Why Jeff Bagwell's HOM candidacy should be impacted by seasonal changes in the ability of AAA first baseman or the leagues bottom three regulars is tough to grasp. The idea of replacmement level is fine provided that all candidates are compared to the same standard. The idea that replacement level changes from season to season and therby impacts the "merit" of the leagues top players artificially doesn't pass the common sense test.

In terms of measuring excellence many would argue that WAAP is a better measuring stick and this would hurt long-term average plus performers such as Sutton and help players such as McGraw and Koufax. Probably a system that uses both WARP and WAAP and weighs them equally would be the most appropriate.

The adjustment for standard deviations to account for "different degrees of difficultness to dominate a league" is something that must be handled very carefully. Is this done on a league wide basis, or is each position looked at separately? There are many reasons that could cause the standard deviations to be higher/lower in some years than others: segregation, military service, expansion, changes in the ratio of major league players to the size of the available talent pool. There are also random year to year fluctuations in the distribution of talent at particular positions. Sometimes the talent at a particular position at a certain point in time simply has a bunch of very equal players. This is currently true of AL second baseman - Cano, Barfield, A. Hill, Pedroia, Iguchi, Ellis, Grudzielanek, Roberts, Lopez, Kinsler, Kendrick, Castillo, Polanco, & Wiggington. There is very little difference in overall ability of this bunch. The standard deviation of a graph of their WARP would be very low indeed. Some are good hitters, some are good defenders.....no one is truly bad. No one looks a real star although Cano, Kendrick, and Kinsler look like they could develop into stars. If we adjusted these players performances because of the low standard deviation it would be a big distortion. It would create an artificial spread of ability where none actually existed.

Clearly playing SS is harder than playing 1B

Maybe, though that's not the only way to put it. Hitting well enough to play regularly at first base in the majors is probably much harder than playing major-league-quality shortstop defense. Which is only to say that degree of difficulty is beside the point. It's not whether something is harder to do, but whether it makes as much difference to do it.

OK, here's my case for the third guy I will really push for, Ron Cey. Here I'll compare him to his contemporaries Nettles (who might be deserving himself, he's right on the edge) and Bando (who's not close), as well as old-timers (and HoM electees) Hack and Groh.

Ron Cey


Year BWAA/Yr BRWAA/Yr FWAA/Yr Rep WARP1/Yr SFrac WARP1 LgAdj WARP2 WARP2/Yr PennAdd Salary
1979 4.3 0.1 1.1 -1.6 7.0 .92 6.4 1.015 6.5 7.1 .091 $12,465,317
1976 4.3 0.0 1.1 -1.5 6.8 .95 6.5 .997 6.5 6.8 .091 $12,001,827
1975 3.6 0.1 1.4 -1.3 6.4 1.03 6.6 1.000 6.6 6.4 .093 $11,601,384
1978 3.4 0.0 0.4 -1.7 5.5 1.06 5.9 1.027 6.0 5.7 .083 $9,688,532
1980 2.4 0.0 0.4 -1.6 5.7 .99 5.6 1.036 5.8 5.9 .079 $9,568,419
1981 4.3 0.0 0.3 -1.5 6.2 .84 5.2 .981 5.1 6.0 .068 $8,600,617
1974 1.7 0.0 1.3 -1.3 4.3 1.04 4.5 .978 4.4 4.2 .058 $5,702,690
1977 1.8 0.1 0.7 -1.6 4.2 1.04 4.4 .985 4.3 4.1 .056 $5,493,979
1973 1.1 0.0 1.4 -1.4 3.9 .93 3.6 .967 3.5 3.8 .045 $4,204,312
1982 1.5 0.1 0.2 -1.3 3.0 .98 2.9 1.033 3.0 3.1 .038 $3,203,948
1986 3.2 0.0 -1.7 -1.2 2.6 .49 1.3 1.030 1.3 2.7 .015 $1,286,718
1984 0.8 0.1 -1.0 -1.2 1.0 .91 1.0 1.043 1.0 1.1 .011 $633,900
1983 1.8 0.0 -2.2 -1.2 0.7 1.03 0.7 1.047 0.8 0.7 .008 $424,771
1985 -0.2 0.0 -1.7 -1.1 -0.7 .90 -0.7 1.023 -0.7 -0.8 -.008 $0
TOTAL 2.4 0.0 0.3 -1.4 4.1 13.11 53.9 1.004 54.1 4.1 .728 $84,876,414
</pre>

Graig Nettles


Year BWAA/Yr BRWAA/Yr FWAA/Yr Rep WARP1/Yr SFrac WARP1 LgAdj WARP2 WARP2/Yr PennAdd Salary
1976 2.7 0.0 1.8 -1.9 6.4 1.05 6.7 1.014 6.8 6.4 .096 $11,996,881
1971 2.1 -0.1 3.2 -1.1 6.3 1.12 7.0 .959 6.7 6.0 .095 $11,373,132
1978 2.3 -0.1 1.1 -2.1 5.5 1.05 5.8 .959 5.6 5.3 .076 $8,502,257
1977 2.0 -0.2 1.1 -2.0 5.0 1.05 5.3 .932 4.9 4.7 .066 $6,891,502
1975 1.2 0.1 1.6 -1.7 4.7 1.02 4.8 .976 4.7 4.6 .062 $6,459,289
1973 0.8 0.1 1.6 -1.8 4.6 1.01 4.7 .954 4.5 4.4 .059 $6,041,237
1972 1.4 0.2 1.4 -1.3 4.4 1.06 4.7 .978 4.6 4.3 .060 $6,002,051
1981 1.3 0.0 1.0 -1.9 4.4 .96 4.3 .937 4.0 4.1 .052 $5,126,187
1970 0.7 0.0 2.4 -1.0 4.2 1.01 4.3 .926 4.0 3.9 .051 $4,910,163
1974 0.9 0.1 1.1 -1.7 3.9 1.01 3.9 .997 3.9 3.8 .050 $4,719,838
1985 2.5 0.0 -0.2 -1.1 3.4 .82 2.8 1.023 2.9 3.5 .036 $3,322,223
1984 1.4 0.0 0.2 -1.2 2.8 .73 2.0 1.043 2.1 2.9 .025 $2,124,494
1983 1.7 0.0 -1.2 -1.6 2.0 .82 1.7 .989 1.6 2.0 .019 $1,358,888
1979 0.0 -0.1 -0.4 -2.0 1.6 .93 1.5 .937 1.4 1.5 .016 $1,015,828
1980 0.9 0.1 -1.4 -2.0 1.8 .58 1.0 .964 1.0 1.7 .011 $759,689
1968 2.5 0.0 0.1 -0.9 3.5 .14 0.5 1.010 0.5 3.6 .005 $573,916
1982 0.1 -0.2 -0.8 -1.7 0.8 .72 0.6 .982 0.6 0.8 .006 $340,296
1986 -0.4 0.0 -0.4 -1.2 0.4 .63 0.2 1.030 0.2 0.4 .002 $121,261
1969 -0.1 -0.2 -0.4 -0.8 0.2 .41 0.1 .929 0.1 0.2 .000 $34,012
TOTAL 1.3 0.0 0.9 -1.6 3.8 16.13 61.7 .971 60.0 3.7 .789 $81,673,145
</pre>

kwarren--that is EXACTLY why I don't use the league's actual standard deviation, because there is plenty of random noise and certain eras may happen to have very many or very few great players. Instead, I use a *projected* standard deviation based on my multiple regression analysis, to see what the "true" spread of talent in the league should be based on things like runs per game, expansion, integration, etc. If you look at the graph in the historical standard deviations chart, places where the actual stdev is consistently below the regression line (1910's NL) are places where players simply happened to be very close together talent-wise, while places where it is higher (2001-02) are years where there really were a lot of exceptional performances (or steroids). My system won't turn Zack Wheat into Tris Speaker, don't worry. The analysis is done on a leaguewide basis.
As for your other point, positional depth does change over time, and that is what I am trying to measure. Unles you are putting Larry Doyle very high on your ballot, you clearly recognize that changes in positional rep levels throughout the game's evolution have to be taken into account.

Bob Dernier Cri, I agree with you, I was just taking a semantic shortcut.

Actually, just looking at these first two, I'm no longer so sure that Cey is better than Nettles. He certainly was a much better hitter and had a longer and higher peak, which the salary estimator likes, but I am not sure that outweighs Nettles' substantial advantage on career length and value. Perhaps both of them belong. It's worth noting that Nettles' 1970 and particularly 1971 seasons with the glove are absolutely extraordinary. Getting over 2 fielding wins above average per year is extremely hard to do, over 3 is Ruthian in its infrequency (but not in its value). Anyways, on to the rest of the crowd.

Looking at them once more, I guess what the salary estimator is telling me is that a lot of Nettles' work around 1980 was so close to replacement level that it was near worthless, even if it pads his career WARP total. Regardless, they're close.

Sal Bando


Year BWAA/Yr BRWAA/Yr FWAA/Yr Rep WARP1/Yr SFrac WARP1 LgAdj WARP2 WARP2/Yr PennAdd Salary
1969 5.0 -0.2 -0.6 -0.8 5.1 1.18 6.0 .929 5.6 4.7 .076 $7,846,300
1978 2.6 0.1 0.5 -2.1 5.4 1.00 5.4 .959 5.1 5.2 .069 $7.721,101
1976 2.3 0.2 0.3 -1.9 4.6 1.02 4.7 1.014 4.7 4.7 .063 $6,619,318
1972 2.2 0.2 0.8 -1.3 4.6 1.07 4.9 .978 4.8 4.5 .064 $6,493,102
1973 4.1 0.0 -1.5 -1.8 4.4 1.08 4.8 .954 4.6 4.2 .061 $5,980,766
1971 4.0 -0.3 -0.4 -1.1 4.4 1.03 4.6 .959 4.4 4.2 .057 $5,716,368
1970 4.1 -0.4 -0.6 -1.0 4.3 1.00 4.3 .926 4.0 3.9 .051 $4,904,403
1974 2.8 0.3 -1.4 -1.7 3.4 .94 3.2 .997 3.2 3.4 .041 $3,643,581
1968 1.2 0.1 0.4 -0.9 2.6 1.09 2.8 1.010 2.8 2.6 .035 $2,733,137
1975 0.3 0.2 0.4 -1.7 2.6 1.06 2.7 .976 2.6 2.5 .033 $2,470,377
1977 0.2 0.1 -0.1 -2.0 2.3 1.05 2.4 .932 2.2 2.1 .027 $1,908,376
1979 -1.0 0.0 -0.6 -2.0 0.5 .86 0.4 .937 0.4 0.5 .004 $207,232
1967 -2.0 0.1 0.4 -1.0 -0.5 .24 -0.1 .976 -0.1 -0.5 -.002 $0
1981 -1.8 -0.1 -1.1 -1.9 -0.9 .17 -0.2 .937 -0.1 -0.8 -.002 $0
1980 -3.0 -0.2 -0.6 -2.0 -1.6 .45 -0.7 .964 -0.7 -1.6 -.008 $0
TOTAL 2.1 0.0 -0.3 -1.5 3.4 13.23 45.15 .965 43.55 3.3 .569 $56,244,062
</pre>

Heinie Groh


Year BWAA/Yr BRWAA/Yr FWAA/Yr Rep WARP1/Yr SFrac WARP1 LgAdj WARP2 WARP2/Yr PennAdd Salary
1919 4.5 0.6 1.6 -1.6 8.5 .91 7.7 .913 7.0 7.7 .100 $14,375,503
1917 4.3 0.3 1.1 -1.5 7.4 1.09 8.1 .930 7.5 6.9 .108 $13,952,723
1918 4.1 0.3 1.0 -1.6 7.3 1.09 8.0 .875 7.0 6.4 .099 $12,320,254
1916 3.0 0.3 1.4 -1.4 6.2 1.05 6.5 .949 6.1 5.9 .085 $10,103,340
1915 2.3 0.3 1.7 -1.5 5.9 1.04 6.1 .957 5.9 5.6 .081 $9,424,997
1920 2.5 0.4 1.1 -1.7 5.8 .99 5.8 .931 5.4 5.4 .072 $8,331,593
1921 2.6 0.7 1.7 -1.7 6.9 .63 4.4 .882 3.9 6.1 .050 $6,597,617
1923 1.0 0.1 0.7 -1.8 3.9 .82 3.2 .872 2.8 3.4 .035 $3,171,107
1924 0.0 0.2 1.0 -1.7 3.1 .98 3.1 .891 2.7 2.8 .034 $2,722,609
1914 2.6 0.7 -0.9 -0.6 3.0 .85 2.6 .923 2.4 2.8 .029 $2,391,318
1913 0.8 0.8 0.6 -0.4 2.7 .71 1.9 .885 1.7 2.4 .020 $1,512,823
1922 -0.9 0.1 1.3 -1.8 2.6 .74 1.9 .871 1.7 2.2 .020 $1,472,581
1925 -4.0 0.0 -2.1 -1.7 -4.2 .11 -0.5 .894 -0.4 -3.7 -.005 $0
TOTAL 2.3 0.4 1.0 -1.5 5.3 11.02 58.8 .913 53.7 4.9 .723 $86,376,465
</pre>

Stan Hack (no wartime penalty)


Year BWAA/Yr BRWAA/Yr FWAA/Yr Rep WARP1/Yr SFrac WARP1 LgAdj WARP2 WARP2/Yr PennAdd Salary
1945 3.9 0.3 1.8 -1.0 7.3 1.06 7.7 .803 6.2 5.8 .086 $10,180,497
1938 3.7 0.3 0.5 -1.6 6.4 1.11 7.1 .847 6.0 5.4 .083 $9,286,903
1940 3.6 0.4 -0.1 -1.5 5.7 1.04 6.0 .855 5.1 4.9 .069 $7,413,914
1935 3.4 0.4 0.6 -1.8 6.5 .76 5.0 .829 4.1 5.4 .053 $6,397,114
1941 4.6 0.2 -1.2 -1.4 5.2 1.05 5.4 .850 4.6 4.4 .061 $6,132,319
1942 4.3 0.2 -1.1 -1.2 4.8 1.01 4.9 .960 4.2 4.2 .055 $5,409,585
1937 1.5 0.3 0.4 -1.6 4.1 1.03 4.2 .862 3.6 3.5 .046 $4,178,150
1933 7.0 0.8 2.4 -2.3 12.8 .11 1.4 .899 1.3 11.5 .015 $3,646,024
1946 3.6 0.1 0.0 -1.0 4.8 .63 3.0 .886 2.7 4.3 .033 $3,519,490
1943 2.5 0.1 -0.5 -1.2 3.6 .96 3.4 .838 2.9 3.0 .036 $2,968,885
1936 2.1 0.4 -1.3 -1.8 3.2 .98 3.2 .853 2.7 2.7 .033 $2,667,219
1934 0.5 0.3 0.9 -1.6 3.7 .72 2.6 .852 2.2 3.1 .027 $2,369,135
1939 1.0 0.3 -0.8 -1.4 2.3 1.10 2.5 .846 2.1 2.0 .026 $1,754,274
1944 1.3 0.2 -0.1 -1.0 2.7 .66 1.8 .835 1.5 2.2 .017 $1,285,788
1947 0.4 0.0 1.4 -1.0 2.8 .47 1.3 .959 1.3 2.7 .015 $1,243,133
1932 -1.4 0.3 -1.5 -1.6 -0.8 .30 -0.2 .878 -0.2 -0.7 -.003 $0
TOTAL 2.7 0.3 -0.1 -1.4 4.6 13.00 59.3 .848 50.3 3.9 .652 $68,452,430
</pre>

Hey Dan, any guesses as to why Bando is so far behind Cey and Nettles?

Right now my rankings go Bando, Cey, Nettles, though they are all in between 40 and 60, I have no problem with moving them around.

It also looks like Groh was a godo selection. I supported Hack and I still do. I take it that the replacement level for 3B was pretty high during this time? Odd because there were any other excellent 3B during his time either.

At least I seem to have figured out how to format.

After accounting for Groh's league environment (which wasn't, in fact, that easy to dominate), Cey looks almost identical to him, with a sllllightly less valuable peak and some hanging-on at the end. They were quite similar types of players, although Groh's fielding was a bit better at its best. Nettles is by far the worst hitter of the bunch, but was one of the best gloves at 3B ever and had an extremely long career, leading the pack in total WARP2 by a comfortable margin. I haven't done Brooks Robinson, but he would seem to me to be a good (if slightly stronger) comp for Nettles. Graig seems borderline to me.

Hack looks like a mistake to me. He was the best hitter of the group, but his offensive peak (41-42) was also his defensive nadir, preventing him from having any real MVP-caliber seasons. And the 30's/early 40's NL was quite an easy league to dominate--Hack was rarely among its true stars (Ott, Vaughan, Waner, Mize) despite his solid WARP1 totals. Moreover, his best year was a war year. I don't see him as close to HoM caliber, although I wasn't following the discussions when he was elected.

Bando isn't even HoVG.

Mark Shirk, I'm not sure if I can see anything in the charts that you can't...

Bando played the same amount of time as Cey. He was a somewhat worse hitter and a much worse fielder. And his only real big offensive year was during an expansion season (not that it makes a big difference). Nettles played for much longer than Bando and was one of the greatest fielders ever at his position. Bando just isn't near their class...I have no idea how Win Shares sees Bando as equal to Cey, but then again I think Win Shares are pretty terrible. WARP sure doesn't.

No, replacement level for 3B wasn't so high during Hack's time at all. But he wasn't anything special with the glove and only once finished higher than 10th in his league in OPS. A good-not-great hitter and average fielder who played in easy-to-dominate leagues and did a chunk of his damage, including his best year, during the war? I'll pass.

Right now my rankings go Bando, Cey, Nettles, though they are all in between 40 and 60, I have no problem with moving them around.

It also looks like Groh was a godo selection. I supported Hack and I still do. I take it that the replacement level for 3B was pretty high during this time? Odd because there were any other excellent 3B during his time either.

Mark, Dan - Here is how WARP (unadjusted) views all of these guys.


Graig Nettles....10.7, 10.2, 8.9, 8.4, 8.2...(46.4)..107.2
Stan Hack........11.1, 9.9, 9.8, 9.0, 7.5...(48.3)..101.8
Heinie Groh......11.5, 10.6, 10.4, 9.1, 8.9...(50.5)...91.4
Ron Cey..........10.5, 9.6, 9.2, 9.1, 9.0...(47.4)...96.6
Sal Bando.........9.3, 8.5, 8.4, 7.9, 7.5...(41.6)...83.2

It you treat longevity and peak as equal I think you have to give the nod to Nettles, particularly when you consider that Groh played in an all-white all-American league and Hack played a good chunk of his career in an all-white, all American watered down league during the war years.

When it comes to Hack, I think he was selected for two reasons,

1) There weren't many good 3B in his day. I think he is the only MLB 3B we have elected between Groh and Matthews (we did elected Jud Wilson and John Beckwith, but they also played a lot of 1B, especially Beckwith). How does Pie Traynor look in your system? If I am not mistaken, he is the only other 3Bman from that era we have looked at and he was a little before Hack.

and

2) He has a decent peak. Before your 'ease of domination' adjustment, Hack's peak looks better to me than that of Cey and Nettles and there are a number of voters, myself included, who care a whole lot more about peak than the filler seasons at the end of a player's career. His peak in WS is also stellar, and his peak in WARP1 isn't too shabby either. While it may have been easy to dominate his leagues (and we have a whole lot of 1930's players but have generally decided that it may simply have been an era with a glut of star level players) 3Bman certainly weren't doing it.

Could a league be 'easier to dominate' simply because there was a glut of star level talent? Would such an occurrence effect the adjustment that much? Or do you think that star gluts happen becuase a league is easier to dominate?

Traynor comes out terribly. $53M.

My salary estimator is very peak-focused as well, although it's kinder to higher-rate lower-PA peaks (John McGraw) than the opposite (Richie Ashburn).

I am so unimpressed with Win Shares (even if I use the fielding part in my system, for lack of a better option) that I don't know what to make of their take on anybody.

I don't think I've been clear enough about how the ease of domination adjustment works. To repeat: I am NOT using the actual standard deviation of the league--if I did, Zack Wheat and Tris Speaker would look the same. I am using a *regression-projected* standard deviation for the league, based on independent variables like run scoring, integration, and expansion. The residuals from my regression--the unexplained variance, the gap between my projected stdev and the real one--is due to two things: random fluctuation, and the actual distribution of talent, what you might call a "star glut" (2001 NL) or "star drought" (1910's NL). My system does *not* dock Bonds' 2001 because Sosa, L. González, and Aurilia had amazing years, nor does it extra-credit Gavvy Cravath's 1913 for being the best in a year when no one really distinguished themself from the pack. It looks at the *conditions* of the league, determines how easy it *actually* was to dominate, and adjusts accordingly, regardless of whether players "took advantage" of that ease of domination in any particular season or not.

Thus, there is no way that a league would be easier to dominate because there was a glut of star level talent--number of huge seasons or HoMers playing in the league or whatever are not part of my regression. The other hypothesis is indeed possible: that what *appears* to be a "star glut," as measured by OPS+ or BP WARP1 or WS, is actually just ease of domination. I think that is the case for Hack and Medwick.

kwarren wrote:

Exactly, this is why Win Shares or WARP with a constant replacement level is more meaningful. Why Jeff Bagwell's HOM candidacy should be impacted by seasonal changes in the ability of AAA first baseman or the leagues bottom three regulars is tough to grasp. The idea of replacmement level is fine provided that all candidates are compared to the same standard. The idea that replacement level changes from season to season and therby impacts the "merit" of the leagues top players artificially doesn't pass the common sense test.

WARP does not have a constant replacement level.

WARP does not have a constant replacement level.

Can or should any system have a constant replacement level? Seems to me that replacement is an ever-evolving thing, and it hits different eras and positions differently.

I'm not well enough versed in replacement theory to know this, but I'm not even sure whether there can be a constant relative replacement level.

To clarify:

WARP1 does not have a constant replacement level.

Dan R:
I don't see [Hack] as close to HoM caliber, although I wasn't following the discussions when he was elected.

One thing you can't do around here is check whether there is a discussion thread for the player and measure its length.

IIRC, Bill Terry and Stan Hack waltzed in with little discussion. There was notable resistance to Joe Medwick and lots of argument because some see him and Chuck Klein as equals while others see Medwick as superior in peak (1935-37?), prime (1934-41?), and padding. It's possible that Medwick benefited from proximity to the quirky careers of Klein (the quirk being Baker Bowl), Hack Wilson and Wally Berger.

Win Shares ratings have always been influential here, especially (I think) for a few decades after Joe Dimino stopped regular publication of Pennants Added (his tweak of Wolverton?) and we first learned of a major revision in WARP. --perhaps Fall 2003, before 1920?

I don't think I've been clear enough about how the ease of domination adjustment works. To repeat: I am NOT using the actual standard deviation of the league--if I did, Zack Wheat and Tris Speaker would look the same. I am using a *regression-projected* standard deviation for the league, based on independent variables like run scoring, integration, and expansion. The residuals from my regression--the unexplained variance, the gap between my projected stdev and the real one--is due to two things: random fluctuation, and the actual distribution of talent, what you might call a "star glut" (2001 NL) or "star drought" (1910's NL). My system does *not* dock Bonds' 2001 because Sosa, L. González, and Aurilia had amazing years, nor does it extra-credit Gavvy Cravath's 1913 for being the best in a year when no one really distinguished themself from the pack. It looks at the *conditions* of the league, determines how easy it *actually* was to dominate, and adjusts accordingly, regardless of whether players "took advantage" of that ease of domination in any particular season or not.

Dan:

I don't understand why Concepcion's WARP or WS should be inflated to account for the fact that he played in a league and era where there were a lot of weak shortstops. Both WARP and WS (which compares a player only to players at the same position) would already rate Concepcion artificailly high in such an environent. Similarly WARP and WS would both penalize Santo to some extent for playing at the same time as a lot of good 3B. Why create an adjustment that penalizes him more.

Your suggestion in other posts that Concepcion was actually a better player than Santo was based on the idea that traditional WARP under-values Concepcion and over-values Santo, whereas I believe that is exacxtly the opposite. When A-Rod, Jeter, Garciaparra, and Tejada were all top notch AL shortstops WARP would penalize all these players because they all happened to play at the same time and they would be costing each other wins that they would get credit for in a weaker field. Same thing for Santo and his contemporaries. But because Concepcion had the field to himself during his career his WARP totals would be artificially high. If anything the difference between Concepcion and Santo that WARP indicates is likely low not high.

Um, neither WS nor WARP (1 or 3) compare a player's *hitting* only to his positional peers. Both Batting Win Shares and BP's BRAR (both season and all time adjusted) are determined without regard for position.

Fielding Win Shares (FWS) and both versions of BP FRAR are of course position-specific. But James' method for allocating FWS doesn't come close to capturing the reality of the magnitude of defense in general, and of the offensive gaps between positions in particular--according to WS, the gap between Frank Howard's fielding and Roberto Clemente's is only on the order of one win per 162 games, which is nuts; similarly, replacement SS are consistently some 4 total WS per season worse than replacement 1B. And BP's FRAR are determined either by the number of chances per position (the season-adjusted version) or by a set constant (the all time adjusted version), neither of which correlate to the scarcity of hitting talent at the position. To be clear: neither system makes any adjustment for the offensive strength/depth or lack thereof of a position at a given point in time. I believe systems like RCAP do, although I'm not sure how they work.

WARP (both 1 and 3) most certainly do not penalize A-Rod, Jeter, Nomar, and Tejada for playing in the same league together; where did you get that idea from? They are compared to a leaguewide .230 EqA offensive replacement level, and a defensive replacement level determined by SS chances per game (WARP1) or a timeless constant (WARP3). I think you should acquaint yourself a bit more with the systems' methodology before making these kinds of incorrect assertions about them.

WARP (both 1 and 3) most certainly do not penalize A-Rod, Jeter, Nomar, and Tejada for playing in the same league together; They are compared to a leaguewide .230 EqA offensive replacement level, and a defensive replacement level determined by SS chances per game (WARP1) or a timeless constant (WARP3).

Dan

So why do you believe that Concepcion have to be adjusted upwards, and Santo downwards? And why would the strength of the players playing the same position as them be such a factor in your mind?

I feel like a broken record...but here goes again.

1. Do you believe that a player's position matters? Or should we have a HoM consisting entirely of 1B and OF? If it matters, why does it matter? Because every team has to field a player at every position, and you can't have a whole lineup of Jason Giambis. How much should we credit Miguel Tejada for putting up offense at the SS position that would be good enough for the outfield? We have to look to the replacement level. You don't know how much any player's contribution is worth until you know what a team would have to put in his place if he went down.

To take an oversimplified example, let's say a team has a 100-run 1B and a 51-run SS, in a league where the replacement level is 80 runs for 1B and 50 for SS. It receives a trade offer of an 80-run SS in exchange for the 100-run 1B. If it accepts, it can hire an 80-run 1B for the league minimum salary, while adding the 80-run SS to the roster, so it improves from 151 to 160 runs overall, despite the fact that the SS produces fewer runs than the 1B. This is why knowing the strength of the players playing the same position is absolutely critical to understanding the value of any given player, and it is very, very basic stuff. If I've lost you already...well, I'll assume I haven't.

2. If position matters, and therefore we need a replacement level, how should we determine it? Replacement level is not some abstract concept, but rather an empirical, existing degree of ability that we can measure by looking at which players are actually freely available. My worst-three-regulars method attempts to do this over all of baseball history since 1893.

3. Certain league conditions (e.g. high run scoring, expansion, segregation) make it easier for players of a given performance level to accumulate wins above a replacement player than others (e.g. low run scoring, no expansion, integration). If we don't correct for these factors, our WARP1/WS based HoM will consist almost entirely of guys from the 1890s and to a lesser extent the 1920s and 30s. My regression-based standard deviation correction properly adjusts for these factors--but NOT for the actual distribution of talent. If nobody happens to be good in a given year, or if a lot of players do, that will still show up in the WARP2 totals.

All that said--Concepción has to be adjusted upwards because he played in the worst era for SS depth in MLB history, meaning that his contribution above average produced more wins than the same contribution at SS would produce in, say, the 1950s. Furthermore, he played in an extremely low-stdev era--a low-scoring, integrated league many years removed from expansion--in which the greatest players were not able to exceed replacement by the same amount that they did in, say, the 1930s. By contrast, when Santo played, good-hitting 3B were fairly easy to come by, meaning that if he went down, Chicago would not have suffered nearly as much as if Brett or Schmidt had gone down around 1980. And his performance relative to average looks better because he got to beat up on expansion pitchers and is being compared to expansion hitters in OPS+, BP WARP, and WS. The standard deviation adjustment corrects for this.

If I haven't made this clear by now, I just never will.

I just wanted to chime in here and thank Dan for this and say that I'm really impressed by it and I think it's extremely well-thought out and insightful. So thanks.

Thanks, Kiko--I've put god knows how much time into it, so I'm glad to hear that it's been helpful. Nothing would make me happier than to see this change people's ballots one way or another. It doesn't have to be Cey, RSmith, and Concepción--if I've made the electorate take account of changes in historical replacement levels and standard deviations in their voting, I'll be more than satisfied. That said, I would be pretty happy if Concepción got in. He won't.

according to WS, the gap between Frank Howard's fielding and Roberto Clemente's is only on the order of one win per 162 games, which is nuts

Dan, do you mean that sounds prima facie nuts? I agree that one win sounds like a tiny margin, but when we are talking about the difference between two major-league right fielders, that might be a huge margin indeed. How much impact can a right fielder's defense have on a game, anyway? Let's say it's 1964; Clemente's raw range factor in RF is about 2, Howard's is about 1.5; of course 1.5 of both numbers consists mostly of putouts that any right fielder would make. Clemente makes 11 more assists and six more errors. And then of course there are a number of other plays on base hits where Clemente discourages a runner from going from first to third or scoring from second. But the tie in the standings between LA and Pittsburgh that season is produced by the two clubs' pitching staffs and their entire offense and the defense that they get from eight other positions. Is it nuts to think that if they were traded even-up (their batting aside) LA would have finished only one extra game ahead of Pittsburgh?

Win Shares, as I take it, credits Howard with defensive value in making those 1.5 putouts per game, even if any outfielder -- Greg Luzinski, Kevin Reimer, whoever -- would have been likely to make those 1.5 putouts. Likewise, it sees most of Clemente's value in those same 1.5 putouts. Essentially in Win Shares (correct me if I'm wrong cause I probably am) the "replacement value" for a defensive right fielder is not having a right fielder at all. Now, that doesn't sound very common-sensical, but it recognizes that Howard did catch 183 fly balls that year, and decides to credit him with those catches, despite the fact that it was the least he could do. But in any case, the residual difference between Clemente and Howard might well be worth no more than a single win per season; right fielders, provided they can catch the absolute routine fly outs, are perhaps not all that greatly distinct from one another. A single win is therefore gigantic. (I don't know if I believe any of this, I'm just arguing for some credence for the Win Shares concept.)

Bob Dernier Cri, that is most definitely the Win Shares concept. I say it is nuts only because we know that guys do have double-digit FRAA seasons every year at every position (no matter which system you use), that Clemente was consistently one of the best at his position and Howard consistently the worst, and therefore that the gap between them should be at least 20 runs/2 wins in a typical year. And I'm saying this as someone who generally argues for conservative assessment of defense (see the Mazeroski thread).

Dan, very interesting piece. Just a comment about Win Shares and how it evaluates defense. If I understand you correctly, you describe Win Shares as "nuts" because the spread of value it allocates to individuals is so constricted. For example, Frank Howard and Roberto Clemente in right field in the 1960s.

I don't have the Win Shares book handy, but I believe James intentionally restricted the spread of Win Shares defensively. I understand James did this for several reasons.
One, measuring defensive contributions to team success at the time James developed Win Shares (the mid 90s) was in its infancy compared to measuring offensive and pitching contributions to team success.
Two, James attempted to create a system to evaluate how much a player contributed to his team winning. As such, just showing up has tremendous value to a team. Philosophically, Win Shares believes players have most of their value by simply showing up because teams do not start out as .500 teams.
Three, Win Shares attempts to use the statistics available throughout time to estimate a player's contributions. Batted ball statistics / hit location / line drive-fly ball-groundout breakdowns are not available throughout time. I don't know for how many games Retrosheet had play-by-play available in the early 90s, but I believe it to be a very small number. With this lack of detailed information available, it makes sense to me to restrict the spread of wins awarded.
Fourth, Win Shares are only about contributions to wins. With the .200 winning percentage baseline (or 32 wins per 162 games) and the fact that almost every team will finish within a 54 win spread (108-54 to 54-108), there are not that many wins to spread around once you allocate wins for hitting contributions, pitching contributions, and fielding contributions.
Fifth, James admits he should have finished his work on Loss Shares. If there were Loss Shares, then a player's failures could be included.
Going back to Howard and Clemente, if we had batted ball statistics and hit locations then we could make much better approximations of their defensive contributions. If James had access to the runners advancing/holding on various hits, then that could create a bigger spread. If James had access to the above fielding information and had Loss Shares, I believe there would be a much larger spread defensively between the two. We could see the effect of extra bases taken or prevented. We could see doubles held to singles or singles played to doubles. This would more accurately describe the defensive contributions of Howard and Clemente. But because of available data and an inability to get the kinks worked out in Loss Shares, James chose to restrict the wins awarded for defensive performance.

My two cents.

Here's a question on the Sinins' encyclopedia (and this seems as good a thread as any to ask it, since I found the query while checking out how bad the SS of Concepcion's time were):

I would think that the difference between RCAA and RCAP among players at one position, given they played about the same amount, should be the same.

Here are the RCAP leaders for shortstops in 1984. Why is Concepcion's RCAP 23 runs more than his RCAA, more than any other player's?

player......... RCAP RCAA PA
1 Cal Ripken...... 63 49 716
2 Alan Trammell. 48 36 626
3 Robin Yount..... 39 26 702
4 Ozzie Smith..... 17 .1 484
5 Craig Reynolds. 14 -5 571
6 Tony Phillips.... 10 ..0 505
7 Dave Anderson. 7 .-8 433
8 Garry Templeton 6 -12 535
T9 Ron Washington 5 0 206
T9 Tom Foley....... 5 -7 304
11 Rafael Santana 4 -2 162
T12 D Concepcion 3 -20 600

the gap between them should be at least 20 runs/2 wins in a typical year

Dan, that certainly makes sense. I guess the major conceptual choice is whether to see Howard's defense in RF as a replacement level or not. In one sense it is because he's out there in right (and in Win Shares' mind, he actually has value above replacement merely by being out there). But in another, Howard's fielding is certainly below "replacement" level in that there must have been any number of guys in the minors in 1962-64 who could play better defense than Frank Howard (none of whom could remotely aspire to carry his bat, of course). In fact that was one of my initial gripes about Win Shares, that it credits Greg Luzinski with having won a fair handful of games with his glove in the course of his career as a left fielder.

There is no such thing as "replacement level offense" or "replacement level defense" as far as I am concerned. There are no replacement statistics--only replacement players. Replacement level isn't some arbitrary abstract concept to philosophize about; it is the real and empirical freely available talent level (which, I repeat, is not steady over time). And you can't divide up players and take Frank Howard's offense but "replacement level defense"--you have to choose between Frank Howard (summing his offensive and defensive contributions relative to average) and a real replacement corner outfielder (offense and defense combined relative to average).

I've tried to do some of my three worst regulars averages for WS, but since corner outfielders virtually never differ by more than two Fielding WS, it made defense almost irrelevant. I imagine if there were such a thing as Loss Shares, Luzinski or Howard would be credited with some defensively, thus more accurately representing the spread of fielding performance.

TomH, I have no clue--I don't know how RCAA and RCAP are calculated. That certainly wouldn't happen in my system.

you can't divide up players and take Frank Howard's offense but "replacement level defense"

Well said.

Did Hack have to go head to head against Bob Elliott? I think Elliot looks better in comparison, especially since he was one of the top RH sluggers in the league and Hack wasn't.

Dunno. I have Elliott as worse than Hack, though, and quite far from electability.

David,

Since the spreadsheet is year by year and not player by player, are there any 3B bewteen Groh and Matthews (mid twenties to mid fifties) that your system thinks should be elected? Traynor, Hack, and Elliot all come out as a no. Have you done AL versions for Kell (highly unlikely) or Rosen? It just seems weird to have a gap that large. Even our 1B gap (mid 1890's to mid 1910's) isn't that large.

Hahahah, I got confused, I thought you were talking to someone else named David! I forgot that's my new handle.

The spreadsheet can be sorted any way you want...

No, there aren't any in the NL at least. There were a number of excellent 3B seasons by guys who weren't primarily 3B (Frisch, Ott etc.), but no one distinguished themself from replacement level enough to meet HoM standards in my opinion. (I am incorrectly accused of supporting the "best at X position in Y decade" argument--"best of a bad lot" is not good enough for me. You have to be *a lot* better than a bad lot to make my Hall of Merit!) I don't really have a problem with not having a 3B from the 1925-50 period--where is it written that talent must be evenly distributed to all positions in all eras? Random "star gluts" or "star droughts" are to be expected, both on a leaguewide basis (1910s NL vs. 2001 NL) and particularly at a given position (1930s/40s 3B, 1990s AL SS). My HoM has no quotas --I simply look for guys who exceeded the replacement level of their time at their position by the greatest number of standard deviations. If no 3B in the 30s and 40s did that, so be it.

I haven't done Kell. Rosen's career was just too short.

To clarify--if anyone can give me a *reason* why the stdev for third basemen in the 30s and 40s was so much tighter than for other positions during that time period, I'm happy to be convinced. (Thanks to the wonders of regression, I can give precise quantitative weightings for the reasons why *leaguewide* standard deviations move up and down in particular seasons, and therefore why they should be corrected for). Until then, I have to accept the null hypothesis: that there was simply a "star drought" at 3B in that era.

But in another, Howard's fielding is certainly below "replacement" level in that there must have been any number of guys in the minors in 1962-64 who could play better defense than Frank Howard (none of whom could remotely aspire to carry his bat, of course).

I think this leads to the flaw in Win Shares that forces the spread in defense to be too narrow for the reasons outlined by mulder & scully (#127). The idea that all players must have "non-negative" value by definition is fine, but there's no reason why a player couldn't contribute negatively in some aspect (e.g., fielding) but offset that with positive aspects elsewhere (e.g., hitting). James recognizes this for pitcher hitting, by allowing negative win shares for pitcher hitting (although oddly putting them on the "pitching" side of the ledger), but doesn't recognize that the same thing applies to fielding (and, to a lesser extent, to hitting).

I just think my methodology is so much simpler, more straightforward, and better reflects the real game. Calculate total wins above/below league average (offense and defense together) for every player, determine who are replacement players at a given position and average their rate performance (in wins above/below average--pre-DH replacement 1B were actually above average), and subtract that replacement rate (times playing time) from the player in question's performance, and you've got it. Why make things any more complicated than that? What do you gain by using a different methodology?

I agree, Dan.

bump

also, I had a question for you Dan/David. If you had to ballpark an approximate "replacement level" for the 3 worst players at a position, what would that level be? About .350 winning percentage? .400? Or somewhere in between/higher/lower?

I had a couple of questions for you about your WARP:
1. Does it work for teams? Do teams have their own totals, or is it the sum of their players? In other words, Win Shares builds from the team down to the players, does yours build from the players up to the team or is a team WARP non-sensical?
2. I think I missed this, but how do you determine pitcher "value?"
3. Is a team WARP based on comparing the totals of a team as a whole to the sum of the individual replacement levels for each position?

For example, I looked at the 1974 Reds. They finished 98-64. I tried to find WARP scores for their starters. Excluding Bench at catcher, I came up with 27 WARP for the other 7 regulars. Looking at Bench's numbers and comparing him to Concepcion and Morgan and Win Shares, I guessed he would have 8 to 9 WARP. That gives the 8 starters about 35 WARP. Compared to a .350 replacement level, the Reds at 98-64 or 96-66 by pythagorean were 40 to 42 WARP. Compared to a .400 replacement level, the Reds were 32 to 34 WARP. That would mean the pitchers and back-ups contributed between nothing to 7 WARP. Or would just compare the Reds to the Padres: worst ERA+ and OPS+ (77 and 86), 3rd worst home ERA, worst road ERA, worst runs scored at home and on the road. The team had the worst record in the league by 6 games, but were still 9 games better then their pythagorean record.

So, is that sort of analysis allowed? Or totally not what WARP should be used for?

I greatly appreciate your efforts.

I was going to mention pitching also. Does your methodology offer any insight into pitchers or is there little to add vs. regular WARP? I think the HoM is short on pitchers and it would be very interesting to see if players like Newcombe or Tiant would benefit/suffer due to low std devs.

Sorry for the delayed response. mulder & scully, your questions are extremely thoughtful and helpful, and are leading me to make some refinements to the system (and also helped me catch a typo that may have thrown some numbers off). As soon as I have an update (which I'm doing A-S-A-P), I'll be able to answer. You most certainly can do WARP for a team, building up rather than down (the building-down problem is why some teams have more BP WARP1 than actual wins, if I'm not mistaken).

mulder & scully, Trying to make these numbers add up on a team level is driving me crazy. I'm going to post the problem I'm having, and hopefully someone can help me think it through. Basically, I can't make a league-average team come out league average. What I mean is:

These are the average OBP's and runs produced per 25.5 batting outs in the NL from 1985-2005: (P includes pitchers and pinch hitters in the nine-slot)


Pos OBP RC/25.5
C .318 4.26
1B .354 5.66
2B .334 4.54
3B .333 4.91
SS .316 3.98
LF .349 5.50
CF .335 4.84
RF .345 5.44
P .232 1.61
</pre>

If you take a straight average of those nine rates, you get a .324 OBP and 4.53 runs per 25.5 outs, which were indeed the league averages for that time period. So far, so good.

Now what happens if you actually stick them on a team together? Ignoring lineup effects for simplicity, we'll give them all the same number of plate appearances. The average NL team in the time period had 670 PA per lineup slot and 4,073 batting outs per 162 games.


Pos PA Out RC
C 670 457 76
1B 670 432 96
2B 670 446 79
3B 670 447 86
SS 670 458 71
LF 670 436 94
CF 670 445 84
RF 670 439 94
P 670 514 33
TOTAL 6030 4073 714
</pre>

But wait...the league average (and the average of the players' rates) is 4.53 RC per 25.5 outs, which in 4,073 outs is 724 runs! So our team of league-average players is suddenly ten runs below average!

I understand the reason why this is the case. The players are all positional average on a per-out basis, but because they have different OBP's, they don't generate the same number of outs. The low-OBP guys, particularly the P, will gobble up more of the outs, leaving fewer available for the productive hitters, thus reducing the overall run output of the team.

But how can I combine players into a team if a team of league-average players isn't average? Is there a better way to do this? If not, I'm sort of at a loss...

But how can I combine players into a team if a team of league-average players isn't average? Is there a better way to do this? If not, I'm sort of at a loss...

Can't you assume that the P will bat 9th? The 9th spot will have something like 70 few PA over the course of a season than the average spot in the lineup. Maybe this will get the totals close enough that you can ignore other lineup effects. (Although, in reality, things are more complicated because pinch hitters use up a lot of the P plate appearances.)

Well, those pinch hitters are counted in my nine-slot positional average.

That might be a brute-force approach to get the right answer--that a team of league average players is indeed league average--but I'd like to make sure I have the concepts right and then apply them if possible. It's sort of like preferring BaseRuns because of its theoretical elegance even if linear run estimators are just as good for most mundane tasks.

OK, I think I've figured it out--you *can't* ignore lineup effects, the league average is what it is only because pitchers hit less frequently than position players. So mulder and scully, this can get fairly complicated fairly quickly. Take the case of Bench on the '74 Reds. NL catchers from 1985-2005 averaged 634 PA per season; Bench had around 700 that year. My system (once I get a catcher replacement level) will credit him with 700 PA above the catcher replacement level. But in fact, Bench had about 65 PA that *weren't* taken from a catcher--they were taken from whichever position player would otherwise have hit in his spot closer to the top of the lineup. The replacement level for those marginal PA should thus be higher. How much higher, I have no idea.
I'll keep working on this and post as soon as I have something useful to offer.

Great stuff Dan. Thanks for all of your work.

By the way Dan, I also found that determining 2B replacement level was very difficult.

My study was flawed so it's not worth presenting, but it was striking that other positions were pretty easy to peg, and 2B was just all over the map.

Yep. Well, Nate's research suggests that 2B average is closer to 2B rep level than any other position (maybe C is down there too). Basically, there's a pool of middle infielders, the good fielders go to SS, the good hitters who can't handle SS become star 2B, and then there's just sort of a glut. What my research shows is that apparently that wasn't always the case--from 1916-1985 or thereabouts, 2B rep level *was* fairly stable, and at more or less the same proportion of positional average as other positions. Then, something happened in the 1980's causing the position to become much, much deeper. Could it possibly be an offshoot of the Ripken effect--even if there was only one Cal Ripken, managers realized that big guys who used to play OF or 3B could indeed play the middle infield, even if they couldn't meet the defensive demands of SS?
As for pre-1920, we all know 2B was deeper then. But it's an oversimplification to say it "switched places" with 3B. Basically, from 1907-15, 2B was more like corner OF than anything else--the worst three regulars had a similar level of performance. The position was that easy, which calls into question whether Collins (and to a lesser extent Lajoie) are really first-tier inner circle Hall members. However, from 1893-1906 it appears to have been about equally scarce to 3B. Anyone have any ideas about what changed around 1906-7 that would cause 2B to suddenly become a much deeper position?

mulder & scully, I'm still working on team-level WARP, hope to have something to post soon.

My theory had been that in the real old days, 2B was a premium defensive position (see Bid McPhee). As gloves got better, the importance decreased, then fell off a cliff with the live ball (see Rogers Hornsby). I remember in little league, you could hide a 9 year old sometimes at 2B, since you weren't going to turn two anyway, and there weren't many left handers ripping balls in that area. Then around WW2 some teams realized that turning two could be done a lot more often than had previously been common. That made a pivot 2B almost as valuable as a good shortstop (except they kept getting hurt). And recently some teams have gone back to the Jeff Kent/Hornsby model.

I think the truth is something like that, but the data again is all over the map. I am pretty sure that post WW2, up through the 80s, you needed someone who could turn two. Other than that there's an opportunity to trade defense for offense (Cutshaw/Doyle, Hornsby/Frisch, Kent/others).

Then again Hornsby started at short, maybe he was better than the metrics say.

Maybe also the young 2B getting hurt in the pivot is putting a bias in the data that is non-trivial to remove.

2B replacement level was not particularly high in the 1920's. Take a look at, say, Cotton Tierney in 1924 or Jay Partridge in 1927--both were given full-time starting jobs despite EqA's around .230 and putridly awful fielding. It would be hard to imagine they could have kept those regular roles all year if the freely available talent level were much higher. Moreover, people tend to conflate Hornsby's clubhouse jackassness with poor fielding (probably because good fielding is somehow considered by sportswriters to be morally superior to slugging). He was, by all quantitative measures, a perfectly decent if unspectacular second baseman.

He was, by all quantitative measures, a perfectly decent if unspectacular second baseman.
?????
By Win Shares, of the 71 men who played at least 10000 innings at 2B, Hornsby was ... 71st.

And it was commmonly said that Hornsby was very poor at fielding popups behind him.

Well, I overstated the case, but not by much. Here are my defensive scores for Hornsby from 1920-29--I take FRAA, convert Fielding WS to FRAA, regress both so their standard deviation is equal to Chris Dial's (quite conservative), and take the average. These are expressed as a rate per 162 games.


1920 5
1921 6
1922 -5
1923 -9
1924 4
1925 -14
1926 -9
1927 2
1928 -13
1929 2
</pre>

So he was a stinker in '25 and '28. He was hurt in '23, and generally bad in '26. He was averageish in the field for his huge offensive years of '20, '21, '22, '24, '27, and '29. I'd say that is only slightly worse than "perfectly decent if unspectacular"--I should have said "generally solid when healthy, with occasional year-long bouts of atrocity."

Of course awful second baseman dont' usually play 10,000 innings at the position. I am pretty sure that James mentions that somewhere as well.

mulder & scully: OK, I've made a bit of progress on team-level WARP. There are a few slight complications:

1. If you stick a full team of replacement players together, the whole is worse than the sum of the parts. To assess any individual player's value, you put him on an otherwise average team and see how many wins he adds/subtracts. But if you put 9 replacement players together, each of whom would make an average team worse by say two wins, the team will be about 20 wins below average instead of 18, since the replacement players' poor OBP's compound each other. That said, I think the proper baseline in this hypothetical is 18 wins below average, not 20.

2. On an average NL team, a replacement player would generally hit in the 7 or 8 spot. So if Rafael Furcal went down, LA wouldn't actually have to eat 740 PA of replacement shortstop hitting--it would dump the replacement shortstop in the 8 slot, and move everyone else up. It seems to me the actual proper way to address this is to use the positional replacement level for the first 95% of 1/9 of the league average team PA, and then the league average of the other 8 positions in the lineup as the replacement level for all subsequent PA. But this correction never changes things by more than 0.2 wins, so I'll let it slide for now.

3. Nate didn't do FAT for pitchers, so I took it upon myself to do my own study using his methodology. The numbers were pretty much what I would have expected, but the gap between starters and relievers is very big, and one would imagine that that's changed over time with pitcher usage. Until I actually do WARP for pitchers going all the way back in time, I won't be able to say how, or how much.

4. I can't quite get the replacement levels for AL and NL teams to line up--an NL replacement team winds up being about 9 runs worse than an AL one. It's only 0.1 win per lineup slot, though, so for now I'll ignore it.

That said, here's how I'd look at the 1974 Reds.

1. The replacement levels for the seven positions I have in 1974 sum to 10.7 wins below average. Assuming that the relationship between rep level for catcher and the other positions didn't change between 1974 and the 1985-2005 average (which is unlikely, it is most likely lower), catcher rep level would be 2.3 wins below average with an average lineup spot's playing time, which is reduced to 2.2 because catchers tend to hit low in the lineup. This brings the position player total to 12.9. I'll assume that the pitchers and nine-slot pinch hitters both hit at their positional average, and that those averages were the same (relative to the rest of the league) in 1974 as they were from 1985-2005 (again, probably an incorrect assumption). The nine-hole hitters subtract a further 5.6 wins below average, making the lineup (offense plus fielding) of replacement players 18.5 wins below average.

2. The 1974 Reds had 70% of their innings thrown by starters. Assuming that the replacement levels for pitchers were the same in 1974 as for the 1985-2005 average (an incorrect assumption), replacement starters would have been a further 7.6 wins below average, and the relievers 0.7 wins more, bringing the replacement team to 18.5 + 7.6 + 0.7 = 26.8 wins below average, or 54.2 wins.

3. The six players I have on the 1974 Reds were 27.5 WARP. As an average fielder, Bench would be about 6.5 WARP, add say 10 FRAA/one win with the glove and he's at 7.5, bringing the team to 89.2 wins.

4. For convenience's sake, let's assume that as a team, the 1974 Reds' pitching staff had a league-average ability to prevent hits on balls in play. The HR/(AB-K) rate at home was 92.7% of that rate on the road. The staff faced 6,190 batters, striking out 875, walking or plunking 553, and allowing 126 home runs. Of those homers, 62 were hit against them at home and 64 on the road. So we divide the 62 by .927 and multiply the 64 by (.927+11)/12 = .9939 and add them up, to get 130 park-adjusted HR allowed. The league batting average on balls in play (BABIP) in 1974, including sacrifice hits and flies, was .277. So we have 6190-875-553-130 = 4,632 balls in play allowed, at a .284 BABIP is 1,281 non-HR hits. 17% of non-HR hits in 1974 were doubles and 3% were triples. The league attempted to steal 8.2% of the time it had runners on base, and was successful exactly two-thirds of the time. So we have the pitching staff, with a neutral defense and park and league-average BABIP ability, giving up 1,025 singles, 218 doubles, 38 triples, 130 home runs, 553 BB+HBP, 100 SB, and 50 CS in 5,637 opposing AB. That comes out to 666 eXtrapolated runs allowed. (They actually gave up 631, exactly in line with their true component stats allowed, suggesting that the team was 35 fielding runs above average, assuming a pitching staff with no BABIP ability and a BABIP-neutral home field. I am crediting their starters with 31 FRAA, including a guess of 10 for Bench. Isn't it nice when things add up like that?)

5. An average NL team in 1974 allowed 670 XR. So the pitchers were 0.4 wins above average, plus 8.3 for the gap between a replacement and an average pitching staff, makes a total of 8.7 WARP for the pitching staff. 89.2 wins for the position players, plus 8.7 for the pitchers, is 97.9 wins.

6. The 1974 Reds won 98 games.

:)

sorry, typo, that should have read "4,632 balls in play allowed, at a .277 BABIP is 1,281 non-HR hits." Duh.

Well, thanks to the amazing spreadsheet-merging mastery of David Foss, I have good news to report: I now have WARP for both leagues, all non-pitcher positions, for every season over 50 PA since 1893. I'm posting the new spreadsheets at the same URL now. I've made some slight methodological improvements but nothing substantial. Here are the main caveats:

1. My system still can't really handle guys who played multiple positions of differing difficulty in the same season. Players who played the majority fo their games at an easier position but a large minority of games at a more difficult secondary position (eg 80 games at 3B, 60 games at SS; or 75 at LF/RF and 70 at CF) will be notably overrated, since they accumulate Fielding WS at a tough position for many games but are compared to the positional average of the weak position. Conversely, guys who played a lot of games at an easier secondary position (eg 80 games at SS, 60 games at 3B or 75 at CF and 70 at LF/RF) will be underrated. So no, Toby Harrah won't be high on my ballot next year, for example--his case depends on his monster 1975 being played entirely at SS, which my spreadsheet thinks he did but in fact he didn't. A lot of guys are affected by this--a few that leap to mind are Kiki Cuyler in various years and Andruw Jones in 1997. I may try to go back in and correct the most egregious of these errors by hand.

2. Since I don't have UZR, PMR, or Fielding Bible data in spreadsheet form for 2004-05, I'm just using straight Chris Dial defense numbers for now. Since his data aren't park-adjusted, that's why Manny shows up as a sub-3 win player in 04 and 05. He wasn't.

3. I still haven’t been able to figure out how to get a floating replacement level for CF (since the worst starting CF actually outhit the worst starting corner OF pre 1960) or DH (since so few teams have full-time DH’s that I can’t get a good worst-regulars average). So I’m just using the flat FAT levels across time: CF as 0.4 wins per season harder than corner OF, and DH’s at league average.

4. Replacement levels for outfielders in the 1890’s will be a few ticks too low, due to teams’ proclivity to stick ungodly atrocious hitters in right field.

5. Due to various quirks in my standard deviation regression, a few seasons are coming out with too-low projected stdevs: the 1955 and 56 NL in particular, and both leagues in 1987 slightly. I'd subjectively reduce the '55 and '56 NL WARP by about 5%, and the '87's by 2-3%.

Other than that, I think they're good to go. I'll post an updated WARP PHoM over the weekend.

Never mind, the Mindspring site appears to have died. I've posted it to the Hall of Merit Yahoo! group.

OK, I have now calculated all adjustments--estimates for the pre-1893 portion of the careers of players I'm considering, war credit and deductions, Negro League and minor league credit, and corrections for the multi-position and 55/56/87 league adjustment errors in the raw data. Here is my quite surprising list of career salaries, with all relevant comments. All 1997 ballot comments refer to ML position players, since I haven't dealt with pitchers and Negro Leaguers yet.

1. Babe Ruth, $446,010,281. I wasn't quite sure how to handle the 1918-19 crossover seasons in my salary estimator, he should probably be a couple mil higher since I'm counting the pitching and the hitting as separate seasons.
2. Ted Williams, $436,918,360. This was *not* the #2 I was expecting. But giving war credit at the average of the four surrounding shoulder seasons for 43, 44, 45, 52, and 2/3 of 53, this is where he winds up. He *is* clearly the second-greatest hitter of all time...and I guess that's what really matters. He benefits from playing in a lower standard deviation era than Ruth as well. If your research never surprises you, why do it in the first place?
3. Barry Bonds, $422,642,003. Very surprised he's still behind Teddy Ballgame. But while Bonds' top 3 seasons are better than anything Williams did, Williams (with war credit) takes the lead for the next batch and doesn't look back. Anyways, it's close, and any slight tweak to replacement levels, standard deviations, or war credit could switch them.
4. Honus Wagner, $376,242,220
5. Ty Cobb, $367,071,500
6. Tris Speaker, $310,418,921. Higher than I would have thought. But clearly in the second tier of immortals.
7. Willie Mays, $309,909,109--including war credit at an MVP level for '53, and a high All-Star level for '52.
8. Rogers Hornsby, $301,654,210
9. Stan Musial, $281,713,108--includes deductions for '43 and '44, and credit for '45.
10. Eddie Collins, $269,218,073--Top 10 all-time. Wow.
11. Mickey Mantle, $267,155,123
12. Hank Aaron, $260,783,696
13. Nap Lajoie, $255,874,641
14. Joe Morgan, $241,002,791
15. Mike Schmidt, $239,397,498
16. Lou Gehrig, $231,474,434
17. Rickey Henderson, $217,430,970. Does this surprise people?
18. Mel Ott, $204,155,349--with war deductions
19. Alex Rodríguez, $197,074,720. And counting. He could crack the top 10, but I don't think he'll make it to $300 million, unless he's got some more 2005's in him.
20. Cal Ripken, Jr., $195,765,669.
21. Jimmie Foxx, $190,430,491
22. Arky Vaughan, $188,451,957. Had he played from 44-46, he might be around 230.
23. Frank Robinson, $185,529,647
24. Joe DiMaggio, $178,901,483. With war credit.
---Inner Circle---
25. Ed Delahanty, $169,429,174. Anyone with a pre-1893 career will be less reliable, as I'm relying on BP numbers here, but this looks about right. He'd probably have gotten over 200 had he not fallen over Niagara Falls.
26. Barry Larkin, $166,938,847. What a stunner--on the inner circle's doorstep. A truly remarkable player--in the 13 seasons between 1988 and 2000, he had 7 seasons worthy of MVP consideration, another four that would have been had he not missed time due to injuries, and the remaining two at as a strong All-Star and the best SS in the NL. HoM peak-level performance for more than a decade. Criminally underrated.
27. Robin Yount, $166,666,662. Hit like A-Rod, but in the scarcest era for SS and hardest-to-dominate league in history. And a good All-Star OF in the later 80's too. One of the greatest non-SB baserunners of the Retrosheet era.
28. Johnny Mize, $165,696,975 with credit.
29. George Brett, $164,601,076
30. Alan Trammell, $161,786,277. Not in the Hall of Fame. My God. The early 80's SS position was more "feast or famine," as Nate Silver would put it, than any other in history as far as I can tell. You had Ripken, Yount, and Trammell--who, fittingly, each won an AL pennant from 1982-84--and then you had Jackie Gutiérrez, Houston Jiménez, Alfredo Griffin, etc. It's hard to see how you could lose with one of those guys on your team--the rest of the league was down a good 7 wins to you at the position every damn year. Fewer monster MVP-type years than Yount, but a longer career at SS.
31. Bill Dahlen, $161,205,999, using BP for pre-1893.
32. Luke Appling, $160,307,671--with both war credits and deductions.
33. Billy Hamilton, $157,859,904--using BP for pre-1893.
34. George Davis, $153,104,353--using BP for pre-1893.
35. Eddie Mathews, $149,565,731
36. Hank Greenberg, $149,059,368. Higher than I thought--he'd comfortably make my HoM even before war credit, of which he has Four years.
37. Wade Boggs, $148,612,495
38. Joe Cronin, $148,612,495
---current PHoM average---
39. Charlie Gehringer, $145,555,805
40. Ozzie Smith, $144,185,780
41. Bobby Grich, $143,555,805, and not in the Hall of Fame.
42. Jeff Bagwell, $139,381,942--calculating the rest of 1994 at his three-year average.
43. Mike Piazza, $138,139,241--including a 20% catcher bonus, and using Chris Dial's extremely tight standard deviation for catcher defense. If you think Piazza cost his teams 10 runs a year in the field, he'd be lower.
44. Gary Sheffield, $137,721,620--does anyone realize how good he's been, and for how long?
45. Paul Waner, $137,516,232
46. Tim Raines, $136,705,988
47. Ken Griffey, Jr., $135,937,359
48. Monte Irvin, $133,916,682--going by the MLE's. This looks awfully high to me.
49. Al Kaline, $133,143,215
50. Lou Boudreau, $132,782,038 with war deductions.
51. Reggie Jackson, $131,725,453
52. Jesse Burkett, $131,501,414
53. Fred Clarke, $130,407,484
54. Frank Thomas, $129,874,318
55. Frankie Frisch, $128,985,272--that 1927 fielding year is something else.
56. Rod Carew, $128,691,341--would be higher if I had non-SB baserunning for pre-1972.
57. Mark McGwire, $127,732,265--how could you possibly argue this guy isn't a Hall of Famer on the merits?
58. Hughie Jennings, $126,530,581--remember, the salary estimator looves its peak.
59. Carl Yastrzemski, $126,366,379
60. John McGraw, $125,400,223--my first PHoM-not-HoM. The salary estimator is all about peak rate, which McGraw has by the bucketful.
61. Joe Jackson, $124,479,535
62. Jackie Robinson, $122,023,159 WITHOUT minor league, war, or Negro League credit. I'm still waiting to get this data from someone--help!
63. Johnny Bench, $120,846,046--including 20% catcher bonus.
64. Joe Kelley, $120,105,821--surprised he's this high. His 1896 looks like it was really something, though, with those 87 steals. A terrific fielder according to WS and WARP.
65. Tony Gwynn, $119,903,872
66. Ernie Banks, $118,736,304
67. Manny Ramírez, $118,412,379.
68. Roberto Clemente, $116,554,930
69. Larry Doby, $115,668,825
70. Carlton Fisk, $115,479,803--including 20% catcher bonus. Should be a bit lower since he didn't play his whole career behind the plate.
---current PHoM median---
71. Al Simmons, $113,908,466
72. Gabby Hartnett, $112,440,055--with 20% catcher bonus.
73. Lou Whitaker, $111,846,510. 2B was a lot deeper when he played than in the preceding decade, but he was good enough for long enough that it didn't matter.
74. Gary Carter, $111,536,658--with 20% catcher bonus.
75. Bobby Wallace, $111,280,809--I wrongly opposed his election way back when.
76. Jim Edmonds, $110,685,408
77. Sam Crawford, $109,648,958
78. Pete Rose, $109,612,887
79. Scott Rolen, $109,347,982--a historically great fielder. Not as many studly MVP type seasons as Santo, but many more at a high All-Star level.
80. Harry Heilmann, $108,891,927
81. Enos Slaughter, $108,776,528 with war credit.
82. Eddie Murray, $108,238,194. Lower than I thought for a guy who compiled 500 HR and 3,000 hits in a low-stdev, pitcher's era.
83. Heinie Groh, $107,832,563
84. Bill Dickey, $107,777,114 with 20% catcher bonus.
85. Dwight Evans, $107,271,272 after regressing 1981.
86. Chipper Jones, $107,086,921--another underrated contemporary basher in my opinion.
87. Willie Keeler, $106,793,743
88. Larry Walker, $106,549,726. Unfairly written off for Coors Field--it's a 25%-30% deduction, not a 100% one. Just a whale of an all-around player. His 140 OPS+ is plenty juicy, but buoyed by being a terrific fielder and a phenomenal baserunner. An unsung hero.
89. Roy Campanella, $106,474,778 with Negro League credit and 20% catcher bonus.
90. Mickey Cochrane, $105,699,138 with 20% catcher bonus.
91. Sammy Sosa, $105,669,729
92. Ryne Sandberg, $105,483,441
93. Dick Allen, $105,266,682
94. Duke Snider, $105,234,777
95. Jimmy Sheckard, $105,188,695--gets dinged a lot for playing in the high-stdev early-aughts NL.
96. Elmer Flick, $104,730,736
97. Jim Thome, $104,247,324
98. Charlie Keller, $103,992,411 with war and minor league credit.
99. Yogi Berra, $103,939,865 with 20% catcher bonus.
100. Darrell Evans, $103,706,796.
101. Albert Pujols, $101,899,410...already.
102. Joe Sewell, $101,826,087.
103. Ron Santo, $101,668,85
104. Bobby Doerr, $101,505,988 with war credits and deductions.
105. Reggie Smith, $101,086,221 with Japan credit. Why him over Wynn? 1. Win Shares has him at about +60 FRAA for his career. 2. The Japan year was a good one. Together those are worth about $12 million.
106. Billy Williams, $100,110,931

107. Craig Biggio, $99,699,868
108. Dave Winfield, $99,459,562
109. Willie McCovey, $99,150,595
110. Willie Stargell, $99,046,753
111. Frank Baker, $99,026,698
112. Jimmy Collins, $98,835,308
113. Pee Wee Reese, $98,818,015 with war credit.
114. David Concepción, $98,464,674. Say no more, mon amour. The simplest way to put it: for a decade (1973-82), he was Ozzie Smith. Hitting, fielding, defense, the whole package, the same player. Ozzie had 13 years at that level, Concepción only 9. But 9/13 of Ozzie Smith is still a Hall of Meriter in my book. His offensive rate stats are dragged down by being called up to the majors early and six hanging-around years. The prime is delectable.
115. Iván Rodríguez, $98,082,720 with 20% catcher bonus. Would be higher with a higher stdev for C defense.
116. Sherry Magee, $97,855,252
117. Cupid Childs, $96,865,986
118. Paul Molitor, $96,482,641. Not as great as advertised...
119. Saturnino Orestes Armas Miñoso Arrieta, $95,480,277 with Negro League credit.
120. Phil Rizzuto, $95,468,507 with war credit. He came into the majors like gangbusters in '41 and '42, with superlative lights-out defense complementing above-league-average hitting. Fought for three years, had trouble readjusting in '46 apparently, but then picked up where he left off, saving 15 runs a year in the field and continuing to hit at or above the league average. And put it all together in 1950, when he won a much-deserved MVP award. Welcome to my PHoM, Scooter.
121. Will Clark, $95,391,756. Dang!! Will Clark?!?!? But there he is. Absolutely lights-out through 1992, and did more than just hang around in the '90s. The (consecutive) peak is high enough for anyone, and he's got a bunch of fill-in years for the career voters. Will the Thrill.
122. Derek Jeter, $95,001,337
123. Billy Herman, $94,556,834 with war credit.
124. Roberto Alomar, $94,214,814
125. Rafael Palmeiro, $93,901,425
126. Joe Gordon, $93,816,904 with war credit.
127. Brian Giles, $93,604,270--and would be higher if he hadn't been blocked at the beginning of his career.
128. Jason Giambi, $92,860,474. Back-to-back seasons among the top 20 all-time for first basemen, and 2002 was an MVP-type year as well. There are a bunch of juicers on this list. They won games for their teams.
129. Roger Bresnahan, $92,658,855 with the 20% catcher bonus. Catchers of his era just couldn't hit. He did.
130. George Sisler, $92,421,756
131. Dave Bancroft, $92,204,345
132. Willie Randolph, $91,672,368. Yowza! Is he even *mentioned* as a Hall of Fame candidate? He seems to be a Whitaker clone...if there's consensus on Whitaker (and there is, isn't there?), why not Willie? Plus he's finally wisening up and putting David Wright in the 2-slot this year.
133. Max Carey, $91,401,410
134. Graig Nettles, $91,207,958.
135. Johnny Pesky, $89,546,804 with war credit. You could almost copy and paste my remarks on Rizzuto.
136. Ralph Kiner, $88,919,074 with war credit.
137. Keith Hernández, $88,525,605. I didn't vote for him last year, but I would this year if he hadn't gotten in already. This is with a very tight stdev for 1B defense; he should probably be higher.
138. Toby Harrah, $88,407,566. Yep, you heard me right. He *was* a butcher in the field, but not every year--he was just average or slightly below in many of his big offensive seasons like his 75, 76 and 82. Again, in a low-stdev era when finding competent infielders was like finding WMD in Iraq, getting this kind of production from that spot was a gigantic advantage, even with the sometimes ghastly D.
139. Buddy Bell, $88,272,126. I sound like a broken record with these 70s and 80s infielders. But that's my system's main finding, I think. Incredible defensive longevity--he was saving double-digit runs in the field for 15 straight years. Very similar to Nettles.
140. Zack Wheat, $88,157,465
141. Goose Goslin, $88,044,560
142. Kiki Cuyler, $87,780,398. Terrific fielding and HoM-caliber hitting for an OF.
143. Joe Medwick, $87,743,989
---PHoM in/out line through 1996---
144. Jimmy Wynn, $87,713,690.
145. Edgar Martínez, $87,432,236
146. Harmon Killebrew, $86,812,972--Gasp! Shock! Horror! That's right, I really don't have Harmon Killebrew in my PHoM. I could barely believe it myself, but I dug into the numbers, and there are no typos here. His raw league-and-position-adjusted offense is indeed an eminently HoM'able $124,225,509, which would put him above the median HoMer. However: a. He played in hitters' parks, which subtracts $13,537,714. b. He hit into 35.5 double plays more than an average hitter would have in his opportunities, removing a further $4,940,760. c. He cost his teams 7 extra runs on the basepaths, dropping him another $1,513,265. and d. He was a really, really, really bad fielder. I use Chris Dial's standard deviations for defense, which are very tight (much less weight to defensive impact than, say, FRAA or UZR), and even I have him at minus 70 runs in the field, gobbling up a whopping $17,420,797. The remainder: a paltry $86,812,972. That said, $86M will probably be good enough to make my HoM by the time 2007 rolls around.
147. Stan Hack, $86,056,439 with war deductions. He's only a legit HoM'er if you don't deduct for wartime competition.
148. Jake Beckley, $85,551,356 using BP for pre-1893. Higher than I would have thought in this peak-friendly salary estimator. 149. Bert Campaneris, $85,219,785. A 70's shortstop. This is getting old by now.
150. Tommy Leach, $85,158,323
151. Ron Cey, $85,145,371. Fell in comparison to his peers in this version of my WARP, not sure why.

Other candidates getting votes, HoM-not-PHoM's, and other interest:

Richie Ashburn, $84,376,538. A nice little player. Would fare better if I distinguished between in-season durability and overall longevity. Even with his defense, his rates weren't that special.

Norm Cash, $84,046,053
Chuck Klein, $83,863,735
Bernie Williams, $83,357,460
Andre Dawson, $83,294,473 after regressing 1981
Vladimir Guerrero, $83,292,970
Rabbit Maranville, $83,200,555

Brooks Robinson, $82,924,926. Yep, I have room for Buddy Bell but not Brooks Robinson. How do I figure? 1. On raw wins above league average, Robinson leads 36.7-30.2. 2. Robinson played in a high-stdev expansion era. Bell did benefit from the 1977 expansion, but not as much. That chips 1.8 wins off of Robinson's total, and adds 0.1 to Bell's. 3. Bell played in a DH league; Robinson didn't. That adds fully 7.1 wins to Bell's total, and just 1.8 to Robinson's. (I repeat--if you don't correct for the DH in WS and WARP1, you will dramatically underrate post-73 AL players). That actually gives Bell *more* wins above average than Robinson, 37.4 to 36.7. 4. Third basemen were slightly tougher to find (lower replacement level) in Bell's era than in Robinson's. That adds about another three wins to Bell, expanding his advantage over Robinson to 3.7 wins. 5. Robinson had a much longer career. Comparing to replacement rather than to average credits him for this, and gives him the exact same career value over replacement (57.3 WARP2) as Bell. 6. But since they had the same career value, and Bell packed that value into fewer seasons, he had a higher peak and prime, as reflected by the salary estimator. Although Brooks has the single best season of the two, his 1964 MVP season, Bell has a sizable lead on seasons 2-4 that put him ahead. The gap between them is only $6 million, small enough that it could easily be reversed in a subsequent version of my WARP. But for now, Buddy Bell holds the day.

Gene Tenace $82,639,472 with catcher bonus
Bob Johnson $81,932,028 with war deduction--what is his case for minor league credit again?
Hugh Duffy $81,881,015
Albert Belle $81,637,984
Tim Salmon $81,620,279, wow! His 1995 was crazy good thanks to defense and baserunning.
Bobby Bonds $81,610,403
Fred Lynn $81,018,713
Frank Chance $80,946,849
Robin Ventura $80,910,018
Gavvy Cravath $80,818,076 with 1906, 7, 9-11 minor league credit--he was just a really bad fielder. Think Frank Howard or Greg Luzinski.
Mike Griffin $80,687,067
José Cruz, $80,465,001
Edd Roush, $80,454,332 with holdout credit
Dick Bartell, $80,287,956 with war credit
Chet Lemon $80,036,084
Ken Singleton $77,634,501
Ken Boyer $76,980,804--there are a ton of 3B I prefer to him. The same type of peak as Bell, Nettles, or Cey, but without the career.
Earl Averill $76,826,301 with minor league credit--I don't get this one at All. As much as Win Shares likes the early-career D, FRAA hates it.
Bill Terry $76,705,725--Mistake.
Ted Simmons $76,589,060 with 20% catcher bonus--I wish I could take back my vote for Simmons, not that would have mattered. From about 1950 to 1985, catcher was between 2B and 3B on the defensive spectrum. You can see it in the number of big-hitting catchers in those years--Berra, Campanella, Bench, Fisk, Carter. Compare that to nobody in the 19