I believe this is the first time we had a 200 post discusion thread in a long time

Either way I see no boost for Roush based on his higher per game rates. He played less games per season. The difference between Cravath and Roush on my ballot is the difference from #24 to #34, I think we all know how close that is on most of our ballots. I just disagree that Roush should get a boost for his per game rates when he was not playing as many games in season.

Js - this reminds me of the NL MVP ballot a couple of years ago among the Primer authors when I voted for Pujols and everyone else voted for Bonds. I took Pujols because he had more total value, and I remember Dimino speicifically telling me he took Bonds because you had to figure in the value of the leftfielder that was in there when Bonds wasn't.

Now I have to defend Joe's side of the arguement!

Here is my question though, does it bother you there are 2 very similar valued players, one who has actual major league numbers, and one who has MLEs as accurate as they may be, and you put the MLE player on top?

I personally just can't do that.

I definately understand though that the difference between 24 and 34 is no more than a hair.

I'd be curious to know your answer to David's question. How short of a career (no matter how good) can you support? The question doesn't arise from the players you evaluate above. Is there a threshold, a lå the good Doctor, below which players do not even get consideration?

I don't have a threshold for consideration. My system strives to give equal weight to peak, prime, and career values, so as career value becomes lower, peak and prime values have to become higher for a player to be a serious candidate. In practice, I have not yet supported for election any outfielders or first basemen with fewer than 300 adjusted career win shares, any infielders with less than 270 adjusted career win shares, or any pitchers with less than 250 adjusted career win shares, as my system sees them.

As we go forward, these lower bounds could change, as could my system. One of the things that I am doing with outfielder study in addition to seeing if I have the outfielders in the correct order is assessing whether my system is weighting any component of peak, prime, or career too heavily.

There is about one handful of players who are not valued highly by my system whose omission troubles me: McGraw, Chance, Moore, Bell, and Beckley. But I'm not so troubled that I'm prepared, at this point either to change my system or to make flagrant exceptions in these cases.

I supported Bonds that year, not because of the extra LFer that would play in his spot (I dont' think that matters in this circumstance) but because I believed that he had more overall value as well as per game value. I dont' know what the WARP or WS numbers were/are in fact I forget if we are talking about 2003 or 2004, but I do recall that Bonds lapped the field in RARP that year, beating Pujols by something like 30 runs over replacement. Did Pujols defensive advantage amount to 30 runs?

Re: Cravath v. Roush

With MiL credit (one year in teh PCL and his AA years) I have Cravath top five years as 37,36,32,30,29. Roush is at 34,34,31,29,24. All totals are adjusted to 162 games, though my adjustments give more weight to WS seasons of 34 or above (1.05x WS, rounding up at .7). This is a quirk, but when doing this translation there has to be somewhere that benefits one whole WS more than the number below it.

I believe that Cravath gets MiL credit for the three years that I mentioned and we have his stats and reliable conversion rates. I am a peak voter and Cravath seems to have a clear peak advantage. For instance WS>15 is 121 for Cravath and 113 for Roush (this and seven year non consecutive is how I measure prime). WS>25 is 43 for Gavvy, 29 for Roush. Cravath also stands out more amongst corner outfeidlers than Roush does amongst CFers.

I am not using WARP here because 1) I trust WS a little more and 2) while I do have rough translations for Cravath MiL years in WARP, I don't trust them enought to argue with them. It is mostly WS/3 with a few adjsutments. Nothing scientific.

Again, Roush's obvious career advantage and a good argument that their primes are roughly even makes this comparison closer. Their rankings are acutally #27 and #36. Veach is at #31, Burns at #49. Burns may be too low but I dont' know if he is better than anyone above him.

Chris,

You put Roush ahead of Burns, Veach, and Cravath because of Roush's higher per game WARP and per game WS despite his lower five year totals. However, we know the reason he has lower five year totals is that he happened to miss a lot of games, either through holdout or injury. I am not willing ot give him credit for this.

Therefore, I believe that his per game rates, while superior, are easily balanced out by his lack of in season 'durability' (games played). He may have been better when he was in there, but since he wasn't playing as often as the others his value goes down overall.

Mike has already given a good answer to this, but I think this point bears further consideration.

First, I am not giving Roush "credit" for games he didn't play. I am giving him credit for being a better ballplayer, when he was on the field during his peak, than either Cravath, Veach, or Burns was during their peak. This is in itself meritorious, though it's only one aspect of merit. But I do not blush for including a rate stat in my assessments, and I encourage _judicious_ use of rate stats.

Second, _if_ WARP or Win Shares had "replacement level" pegged accurately in their systems, then WARP or win share totals for a season would be a sufficient guide to value for that season. If we are being shown the value a player earned above replacement level and if we assume that when that player didn't play, he was being replaced by a replacement level player who, by definition, earns 0 value above replacement, then a tally of win shares or WARP would tell us everything we needed to know, elegantly combining rate of performance and durability.

I don't believe that WARP's "replacement level" is really replacement level, and I know that the zero point in win shares does not attempt to find a real replacement level. In both cases, I believe that real replacement level is higher than the zero point in these systems. Therefore, these systems will, to a small extent, overrate the value of sheer durability. Using a rate stat as a tie-breaker in close cases is, I believe, a good way to correct for this problem.

P.S. I am here only discussing the way I would rank the five-year peaks of this players. I'm not giving my overall ranking of them.

In the past, I have had Roush ahead of Cravath, but both were on my ballot in 1959 (am I unique in supporting both of this pair?). My 1960 ballot is obviously still in progress, so I'm not committing myself on how I will rank these two this year.

20 years ago I had Roush #10ish and Cravath outside my top 50. Now I have Cravath around 20-25 and Roush around #40-50. Go figure.

Obviously I had to decide to give Cravath MiL credit, and then I realized that Cravath has the higher peak. But I still (always did) have Leapin' Larry Doyle ahead of both of them, if it's a 1910s NLer you want.

More on yest's observation above....

Who knew? But this seemingly inconspicuous backlog election has generated some of the most thought-provoking, assumption-challenging, and ballot-changing discussions of methodology and voter-preference in several voting cycles.

Thanks everyone for the great posts, it's absolutely helping me to rethink, hone, and sometimes change my positions on things. Yeah, it feels unsettling, sort of, to be unsure of things and to see how much my own opinion has changed with time, but it's also great to see that everyone's grappling just as hard with this stuff.

Uh, the second paragraph of my last post came out funny. I hope no-one thinks the line "it's absolutely helping me to..." smacks of solipcism in the way it appears to me to have come off. I'm not really that self-centered, though my wife might disagree. ; )

Moved from Ballot thread

I wrote, then Chris J answered starting with huh, which is a response many of my question seem to generate.

13)RUBE WADDELL – Another big peak pitcher, though maybe it should be his teammate Vic Willis based on recent discussion. Chris J?

Huh? I'd be more than willing to help (time permitting), but I have no idea if you're asking for info on Waddell or Willis or comparing the two, or whatever. If you can clarify I can try to do some work on them by the '61 election.

I guess what I am aksing is this, they are about the same age, and Willis WS 3 year peak is better and his career WS number is better. Willis pitched about 1000 extra innings. But the ERA+ favor Waddell 134 to 118.

So you are known to look at these things closely Chris J how do they stack up for you?

I thought they were teammates in Pittsburg one year, but I was wrong.

I would rank them Roush, Veach/Burns, Cravath. (peak)

One aspect that Chris Cobb's excellent analysis does not take into consideration is "league quality".

When that is taken into consideration (i.e. the strength of the AL vs the NL during this period), Bobby Veach clearly moves to the front.

I use "All-Star" appearances as part of my analysis for prime, and this is a significant factor in my support for Veach.

"All-Star" OF 1913-1925 (based on WARP1 and WARP3)

1913
LF (burns, wheat, carey)
CF COBB, SPEAKER
RF JACKSON, Cravath, Crawford
1914
LF BURNS, TWALKER, WHEAT, Magee (veach)
CF SPEAKER, Leach, Kauff (cobb, zwilling)
RF CRAWFORD, Evans (cravath)
1915
LF VEACH, Shotton (cooper, killefer)
CF COBB, SPEAKER, KAUFF, Zwilling (magee, twalker)
RF CRAVATH, Hinchman, Crawford
1916
LF VEACH, WHEAT, Shotton, Jackson (burns)
CF COBB, SPEAKER, CAREY, STRUNK, Felsch (kauff, paskert, cwilliams)
RF Hooper (cravath, hinchman)
1917
LF VEACH, BURNS, Jackson, Lewis (bodie)
CF COBB, SPEAKER, CAREY, Felsch, Kauff, Roush, Cruise
RF (hooper, stengel, cravath, (roth))
1918
LF Burns (mann)
CF SPEAKER, COBB, Roush, Paskert, Carey (tobin)
RF HOOPER (youngs,flack)
1919
LF RUTH, VEACH, BURNS, Jackson
CF SPEAKER, COBB, ROUSH, Felsch, Myers, Kauff (bodie)
RF Youngs (rice)
1920
LF JACKSON, Veach, Burns, Wheat
CF SPEAKER, ROUSH, FELSCH, JACOBSON, CWilliams, Myers, Rice
RF RUTH, YOUNGS, Hooper
1921
LF RUTH, VEACH, McHenry, KWilliams (twalker, burns, bigbee)
CF COBB, SPEAKER, Jacobson (cwilliams, carey, rice, powell)
RF HEILMANN, BMeusel (tobin, youngs)
1922
LF KWILLIAMS, VEACH, RUTH, Bigbee (wheat, imeusel)
CF COBB, SPEAKER, CAREY, Miller, Jacobson
RF Heilmann, Youngs (cwalker)
1923
LF KWILLIAMS, JAMIESON
CF SPEAKER, MOSTIL, Carey, Statz, Witt, Cobb (roush)
RF RUTH, HEILMANN, Youngs (rice)
1924
LF WHEAT, Goslin (jamieson, cuyler, falk)
CF COBB, Jacobson, Speaker, Carey (roush, flagstead, cwilliams)
RF RUTH, HEILMANN, YOUNGS, Rice (hooper)
1925
LF GOSLIN, Wingo, Blades (wheat)
CF SIMMONS, SPEAKER, MOSTIL, CAREY, Roush, Cobb
RF CUYLER, HEILMANN (rice, cwalker)

BOLD (top-4); ALLCAPS (top-16); 
Mixed (top-32); (lowercase) (within 10% of top-32)

Top-4 Top-16 Top-32 Close Total
   1     5      1     1     8   Veach
   0     3      2     3     8   Burns
   0     2      3     3     8   Roush
   0     1      1     3     5   Cravath

Would Cravath have made MLB-All-Star teams during his time in the minors?  
Possible but unlikely more than two.

I guess what I am aksing is this, they are about the same age, and Willis WS 3 year peak is better and his career WS number is better. Willis pitched about 1000 extra innings. But the ERA+ favor Waddell 134 to 118.

So you are known to look at these things closely Chris J how do they stack up for you?

Willis is a guy I keep meaning to look at a lot more closely, but just haven't done so.

To me, he appeared to be a borderline candidate, and I remember being really curious to see how his RSIs would turn out and how that would affect his W/L record. Turns out he had average run support (RSI of 100.13) - not what I was expecting from him. This is especially interseting because he was not a good hitting pitcher - OPS+ of 18 was about 20 points below average for his era.

How to explain a 118 ERA+ and a win/loss record of .548? Was it like Waddell - a case of unearned runs? Nope. Adjust him for UER and he comes out 10 UER fewer than he should've. His career ERA would be changed from 2.63 to 2.61 and his ERA+ should go up to 119.

Uh-huh. So how do you explain this? The only thing I've gotten is that he underachieved in real life by 10 wins based on his real life run support and real life RA/9IP. Was he an underachiever? Just unlucky? Bad run support dispersal? I dunno, but you can get a handle on some of his run support dispersal if you're curious here. The whole thing (well, the whole thing except the header row) can be copied & pasted into excel if you want to. Not too much trouble to create the header row if you do so. I really suck at figuring if run support dispersal is off unless it's extremely obvious though.

The less interest one has in W/L, the better a candidate he would be.

Looking at the MOWP info, he wasn't used as an ace. The only time he had more starts than he should've in consecutive seasons was 1908-9. He only had four such seasons in his career. His AOWP of .496 is below average, but that doesn't change things too much.

I reckon he had fantastic defensive support early and late, but rotten support in the middle. Adjust ERAs accordingly for those periods.

A workhorse with a nice ERA but an unexpectedly bad W/L record. I really have nothing brillant here except to say that he's the sort of candidate that needs some bit of info or research to hook to kick him out of the pack and onto the ballots; and he just doesn't have it. I think he might belong on more ballots than he's on, and probably deserves some sort of research -- maybe something along the lines of what was done for Mickey Welch to determine why Vic lost as many as he did.

Right now I don't even have Willis tentatively ranked. I just think there's no way I'd put him on my ballot. If he gets closer, hopefully I'll look closer at him.

Rube Waddell? I began as a supporter of his. The first references to RSI in this project were discussing him. Someone - I forget who - said his UERs looked high so I checked and they were very high. I (incorrectly) labeled him the King of the Unearned Run, and dropped him from my ballot. The innings aren't great for him, the ERA+ overrates him, it was a shorter career, his own bat contributed to his run support (though it was bad run support no matter what he did), and there's questions if his personal character could've hurt the team over the course of the season.

He's still got the K's. Going by Dr Memory's K#, he's the fourth best ever, behind Dazzy Vance, Cy Seymour, and Brandon Lidge. So he's defense independent. His ERAs are still good, but I don't see what he has as being enough. I have him below Clark Griffith, Eppa Rixey, Hal Newhouser, Wes Ferrell, Mickey Welch, Urban Shocker, Dolf Luque, Red Ruffing, Bobo Newsom, Tommy Bridges, & Jose Mendez. Addie Joss should be in there somewhere, but I can't find him. That might be too harsh, but I'm fairly lax at how I rank guys the further I get from the ballot, and Waddell's safely off the ballot.

Can't really compare him to Willis too well because I don't have a good handle on Willis. Right now, I'd take Willis over Waddell. Ask again in 10 minutes, you might get a different answer.

A really extensive study on Willis might yield some evidence that Willis belongs on the ballot now, but aparently I don't think that's likely, given that he's stayed so far off my radar.

jimd,

How do you combine WARP1 and WARP3 in your all-star rankings?

I have some scripts which take the team pages for each league, and run correlations between the batting, pitching, and fielding parts (each position separately) of WARP1 and WARP2. With these correlations established, I can then project the WARP2 numbers into the other league(s), and then average each player over all leagues for the season. This effectively allows me to apply the "difficulty" portion of the transform without the "all-time fielding" portion.

Jingoist here again, lobbying hard for my buddies Duffy, GVH and Ryan.

However it's sliced these 3 guys are better HOM candidates than Cuyler who the electors saw fit to induct.

Nobody addressed this, but we didn't elect Cuyler, did we?

Daryn,

I think he means the HOF. However, isn't Duffy in the HOF?

Thanks again to Chris Cobb for the “hitters” comparison. Your work is consistently top rate. I wanted to check something for myself about the 5 and 10 year material so let me add to the win share stuff. I think I have the correct years for the peak/prime based on previous posts. To this I’ve added the cumulative team Win Percentage for the years of interest. I had to expand 1918,1919 to 154 games W/L records and extrapolate those short season win percentages.

5 year consecutive peak (with WWI prorated totals)

Name    Years  Total WS    WinPct of teams
Medwick  33-37 155.3   .574
Burns    16-20 145.2   .591
Veach    15-19 143.6   .545
Roush    17-21 143.6   .544
Cravath  13-17 142.9   .564
Averill  30-34 142.6   .531
Sisler   18-22 141.5   .516
Johnson  36-40 108.4   .353

10 year prime (with WWI prorated, Cravath has WinPct for 7 years)

Name     Years   Total WS    WinPct of teams
Averill    29-38  266.9   .535
Burns      13-22  262.0   .579
Cravath    08-17  255.5   .541
Medwick    33-42  248.5   .584
Roush      17-26  244.2   .550
Veach      14-23  239.9   .516
Sisler     15-25  236.4   .490
Johnson    33-42  217.7   .390

So which of these guys is not like the other? People say that how the total team wins doesn’t affect the individual player totals when it comes to Win Shares. But what about the extremes? Bob Johnson in the 5 year peak was on the equivalent of the 2004 KC Royals. And Joe Medwick and George Burns, when compared to Indian Bob, had things on easy street.

Anybody else consider this a possible valid reason why Bob Johnson WS totals lag behind?

Al,

Thanks for working this out. I certainly agree that this is a major reason why Johnson's WS totals lag behind.

Correction to 10-year prime numbers for Medwick from preceding posts

In going over the numbers for an Oms-Brown addition to the outfielders study, I discovered that I had inadvertently intemixed stats for Medwick 1932-41 and Medwick 1933-42. I had initially focused on the first set of seasons as Medwick’s top 10 years for batting rates (OPS+ and EQA), but as my study shifted, it became clear that top 10 years should be defined rather in terms of top WS and WARP1 totals. I switched some of Medwick’s totals to the 1933-42 ten-year stretch in consequence, but not all of them. Here is a reprint of Medwick alongside Averill, Sisler, and Johnson, with his changed spots in the rankings by WS and WARP1 totals listed below, along with the slight change this makes in the combined rankings, in case anyone is interested:

Revised 10-years (primes)

Medwick      Averill      Johnson     Sisler
6379 PA      6647 PA      6323 PA     6105 PA*
1476 g       1485 g       1459 g      1374 g*
139 OPS+     136 OPS+     137 OPS+    142 OPS+
.305 EQA     .307 EQA     .308 EQA    .305 EQA
-12 FRAA     15 FRAA      -8 FRAA     53 FRAA
195 FRAR     309 FRAR     183 FRAR    206 FRAR
--           --           --          30 PRAR
536 BRAR     548 BRAR     528 BRAR    509 BRAR
226.8 bws    215.0 bws    189.5 bws   209.8 bws
38.2 fws     51.9 fws     28.2 fws    19.1 fws (7.5 pws)

Overall, his PA go up, his g stay the same, his OPS+ and EQA drop, and all his WARP and WS totals go up. With this change, however, it becomes clear that bws correlate more closely with OPS+ than with EQA. Averill leads both Medwick and Sisler in EQA while trailing them in OPS+, and his rate of bws trails theirs, which are nearly equal. Johnson’s bws rate is sabotaged by his teams, of course, but WARP, interestingly, gives him fewer brar/g than Averill, despite Johnson’s higher EQA. I don’t understand that. I had better check my math.

Integrating this group with the other:

By WS totals
Averill 266.9
Medwick 265.0
Burns 262.0
Cravath 255.5
Roush 244.2
Veach 239.9
Sisler 236.4
Johnson 217.7

Medwick moves from fourth to second by win shares, just behind Averill.

By WARP1 totals
Averill 857
Veach 808
Roush 897
Burns 786
Cravath c. 750??
Sisler 745
Medwick 731
Johnson 711

Medwick moves from eighth to seventh, passing Bob Johnson.

These shifts in Medwick’s 10-year prime values change the ranking based on adding up their placements to the following:

Averill 16
Roush 20
Burns 21
Medwick 25
Veach 27
Cravath 30
Sisler 33
Johnson 42

Medwick moves from 4th at 28 to 3rd, passing Veach.

When I post the Oms-Brown study (I hope tomorrow morning), I will include the revised Medwick numbers again.

Here is a little preview of my moneyball study, yet to come. But apropos of the discussion of Bob Johnson's WS, here is Charlie Keller vs. Ralph Kiner.

Best 3 Seasons in Aggregate (best 3 by OPS+)

Keller 1885 AB+BB .287/.409/.539/164 102 WS total
Yankees win 302 games (1941-42-43)

Kiner 1997 AB + BB .310/.434/.642/179 102 WS total
Pirates win 197 games (1947, 1949, 1951)

Kiner's OBA was 25 points better, his SA more than 100 points, he had more than 100 extra PA, led Keller 179-164 on cumulative OPS+ and comes up with the same number of WS. And coming 6-8 years later, I don't think Keller was playing in a better league, especially considering one of his 3 great years was in the WWII year of 1943. Any discount for that, of course, would not appear in WS anyway.

In Kiner's 10 years in the bigs his teams went 664-875 (.431). But that includes one year in Cleveland at 93-61. His 9 NL seasons were 571-814 (.412) and his 7 Pirates seasons were 442-635 (.410). So anybody who thinks Bob Johnson got shorted by WS, I give you Ralph Kiner. Or, did Charlie Keller just get bonused for being a Yankee?

My point, BTW, is not to prove that Kiner was better than Keller, since Keller is nowhere on the ballots, and besides that is self-evident. For his career Kiner gets an extra 1000 PAs even over Keller's WWII XC MLE projected total. Kiner was BA title eligible 9 times to Keller's 5, projected to 7. And Kiner's OPS+ for their 5 peak years averages about 10 points better. Granted Keller catches up, but at his best Kiner seems clearly to have been better.

Yet, despite the 1000 extra PAs, Keller (adjusted) beats Kiner on career WS also 251.5-242. So there are those Yankees again.

My point is that anyone who believes WS shorts players on bad teams, Kiner is more proof.

I agree with Sunnyday2's point. I expect that the Keller/Kiner split is made up of equal parts Keller being on good teams as Kiner being on bad ones.

I would add, though, that win shares does show Kiner as the more valuable hitter in the three-season comparison Sunnyday2 is looking at. Keller catchers Kiner on fielding value:

Keller 91.7 bws, 10.4 fws
Kiner 94.4 bws, 7.3 fws

But giving a team an additional 15 points of OPS+ and an additional 100 games played should probably be worth more than one additional batting win. If Keller's bws rate is projected into Kiner's playing time, that alone brings Keller up to a 97.1 ws total.

Here’s a study bringing Willard Brown and Alejandro Oms into the outfielder study. For comparative purposes, I’m going to re-list data for Averill, Medwick, and Roush from the preceding studies alongside Oms and Brown’s data. I’ll include all 10 post-1910 outfielders studied so far in the comprehensive listings at the end of each section, and I’ll also reference the totals of other players when they seem relevant.

Five-year peak

Oms         Brown       Roush       Medwick    Averill       
1921-25     1935-39     1917-21     1935-39    1930-34
3154 PA     3050 PA     2901 PA*    3328 PA    3378 PA
751 g       726 g       679 g*      761 g      752 g
139 OPS+    134 OPS+    149 OPS+    152 OPS+   138 OPS+
--          --          .312 EQA    .324 EQA   .307 EQA
--          --          -4 FRAA     -5 FRAA    20 FRAA
--          --          160 FRAR    98 FRAR    170 FRAR
--          --          276 BRAR    332 BRAR   279 BRAR
113.5 bws   104.5 bws   121.8 bws   137.4 bws  113.4 bws
26.1 fws    22.7 fws    21.7 fws    17.4 fws   29.2 fws

In this group, Brown clearly has the weakest 5-year peak, while Oms appears a step behind Averill in the field, virtually even with Averill offensively. Medwick dominates in win shares, trails in WARP. Roush is slightly behind Averill in WARP, slightly ahead in win shares. He has the better rate, though he clearly benefits on defense from a lower centerfield replacement level in his era.

By Win shares in this group
Medwick 154.8
Roush 143.5
Averill 142.6
Oms 139.5
Brown 127.2

Including all ten players
Medwick 154.8
Burns 145.2
Veach 143.6
Roush 143.6
Cravath 142.9
Averill 142.6
Sisler 141.5
Oms 139.5
Brown 127.2
Johnson 108.4

WARP’s offensive analysis tends to favor high walk rates more that does OPS+ and, I think, more than win shares does, and it places more weight on defense. Oms, therefore, like Averill, would rate more highly in WARP1 than in WS, while Brown would probably rank about the same. His loss due to plate discipline issues would probably be offset by gains from his time at shortstop early in his career.

My guess as to total WARP1 rankings
Sisler 492
Averill/Burns 449
Veach 446
Roush 436
Oms 435
Medwick 430
Brown 410
Cravath 384
Johnson 368

Neither Oms nor Brown has peak value at the heart of their HoM case, but neither is weak by this measure. Oms is part of the very closely matched middle group, while Brown trails a bit more.

Now, ten-year prime

Oms         Brown       Roush       Medwick    Averill
1921-30     1934-43, 46 1917-26     1933-42    1929-38
6065 PA     6157 PA     5422 PA*    6379 PA    6647 PA
1444 g      1466 g      1265 g*     1476 g     1485 g
134 OPS+    130 OPS+    141 OPS+    139 OPS+   136 OPS+
-- EQA      --          .306 EQA    .305 EQA   .307 EQA
-- FRAA     --          -11 FRAA    -12 FRAA   15 FRAA
-- FRAR     --          287 FRAR    195 FRAR   309 FRAR
-- BRAR     --          520 BRAR    536 BRAR   548 BRAR
205.3 bws   208.0 bws   201.9 bws   226.8 bws  215.0 bws
46.6 fws    41.1 fws    42.3 fws    38.2 fws   51.9 fws

On ten-year prime, Brown and Oms close the gap with their major-league counterparts. There’s something a bit odd with my bws estimates for Oms and Brown here: By OPS+ (and better OBP within OPS+) Oms should be higher, despite a little less playing time, but Brown is above. I think the error is no more than 5-6 win shares, but I think Oms should be a bit farther ahead of Brown here than he appears to be. Let’s look at the totals:

By WS Totals
Averill 266.9
Medwick 265.0
Oms 251.9
Brown 249.1
Roush 244.2

By WARP1 totals
Averill 857
Oms 810?
Roush 807
Medwick 731
Brown 705?

WS treats the corner outfielders better, while WARP likes the centerfielders, so I’ve tried, in my WARP1 guestimates, to reflect the preferences of the two systems.

Here’s the overall list by WS and WARP totals:

WS Totals
Averill 266.9
Medwick 265.0
Burns 262.0
Cravath 255.5
Oms 251.9
Brown 249.1
Roush 244.2
Veach 239.9
Sisler 236.4
Johnson 217.7

WARP1 Totals
Averill 857
Oms 810?
Veach 808
Roush 807
Burns 786
Cravath 750?
Sisler 745
Medwick 731
Johnson 711
Brown 705?

By 10-year prime, Oms is probably in the top half of the group; Brown is probably in the bottom half. It’s in career value that the two NeL candidates will do the best. That probably reflects the biases of the translation system, which tend to lengthen careers and suppress peak. Hopefully voters’ awareness of these biases will enable them to correct for them. To a certain extent, in a balanced ranking system the two biases will cancel each other out.

Now, career

Oms         Brown       Roush       Medwick    Averill
9056 PA     8407 PA     8156 PA     8142 PA    7215 PA
125 OPS+    129 OPS+    126 OPS+    134 OPS+   133 OPS+
2178 g      2036 g      1967 g      1984 g     1668 g
-- EQA      -- EQA      .294 EQA    .300 EQA   .303 EQA
-- FRAA     -- FRAA     7 FRAA      1 FRAA     7 FRAR
-- FRAR     -- FRAR     430 FRAR    254 FRAR   317 FRAR
-- BRAR     -- BRAR     580 BRAR    624 BRAR   560 BRAR
276.3 bws   262.6 bws   255.7 bws   267.4 bws  223.9 bws
63.7 fws    53.4 fws    58.3 fws    44.4 fws   54.5 fws
2178 g      2316 g#     1931 g#     1899 g#    1761 g#
340.2 ws    361.0 aws   321.7 aws   311.8 ws   308.4 aws
                        (+24 RAR)              (+100 RAR)

g# are adjusted games. Brown’s adj. games include two seasons WWII credit. Roush’s adj. games include expansion of his 1918-1919 seasons to 154 games and a pinch-hitting conversion. Medwick and Averill also have the ph conversion, and Averill has one season of MLE credit for PCL play added. Oms and Brown do not need PH adjustments becauses a decline in PA per game is built into their late-career projections. The career ws totals of Brown, Roush, and Averill are adjusted in response to changes in their career games. The PH adjustment does not affect career win shares.

Let’s start with totals, just listing all ten:

Career Win Shares
Brown 361.0
Oms 340.2
Cravath 326.2
Roush 321.7
Medwick 311.8
Averill 308.4
Sisler 300
Burns 296.1
Johsnon 287.4
Veach 273.0

WARP1
Brown 1080?
Roush 1034
Oms 1020?
Averill 977
Johnson 933
Veach 909
Burns 898
Medwick 878
Sisler 869
Cravath 848

Analysis. Clearly Brown and Oms have the superior careers in this group. Brown’s career rate stats slip very little in comparison to his ten-year rates: he appears to have a career shape much like Bob Johnson’s, but longer. Oms drops some in his longer career: his career length is less meaningful than Brown’s in that it is a more Sisler-like tail, though his level of play was somewhat better than Sisler’s.

Here’s how the ten outfielders stack up when their places in each of the rankings are added up:

Averill 20
Roush 25
Oms 26
Burns 27
Medwick 31
Brown 33
Veach 34
Cravath 37
Sisler 40
Johnson 53

This way of ranking treats all differences between places in a rank order as equal, doesn’t include WWII league-quality adjustments or Federal League adjustments, doesn’t include league strength adjustments at all, accepts the WS assessment of Johnson at face value, and doesn’t adjust first-base fielding win shares for Sisler, so it’s hardly a definitive ordering of these players. But it’s something to work with.

If we were creating a HoM based only on best 3 nonconsecutive seasons, sure I could see an argument for Kiner. But, giving Keller fair war credit and then noting that (1) his "shoulder" seasons are better than Kiner's, (2) he was more consistent than Kiner, and (3) his run production compared with Kiner is slightly better than indicated by OPS+ because of his balance of OBP and SLUG, I have to conclude that Kiner was, at best, Keller lite. I, for one, think that the rapid election of Kiner would be a mistake.

Al Peterson
5 year consecutive peak (with WWI prorated totals)

Name    Years  [u]Total WS[/u]   WinPct of teams
Medwick  33-37  155.3   .574
Burns    16-20  145.2   .591
Veach    15-19  143.6   .545
Roush    17-21  143.6   .544
Cravath  13-17  142.9   .564
Averill  30-34  142.6   .531
Sisler   18-22  141.5   .516
Johnson  36-40  108.4   .353 

Anybody else consider this a possible valid reason why Bob Johnson WS totals lag behind?

Beside the issue at hand, concerning Win Shares, the other three columns make an impressive table.

Chris Cobb:
Thanks for working this out. I certainly agree that this is a major reason why Johnson's WS totals lag behind.

Bill James says "no" and gives some evidence. It isn't clear to me whether he studied it systematically. Wally Berger's great 1935 seasons provides the most spectacular example of "yes" that I know.

P.S. My note on "Double Duty Radcliffe" is more than four days old and still fourth on the Hot Topics board.

EricC, it is unfair to Ralph Kiner to suggest his case is based solely on 3 non-consecutive peak seasons, and it is unfair to me to suggest that that is what I was saying. I have presented his fuller case elsewhere and if he ever gets a thread I'll do it again. I was only pointing out that if WS shorts guys on weak teams, that Bob Johnson is not alone in that category.

I could just as easily point out that Keller's case depends on giving 2 years of WWII credit to a notoriously injury prone player, that he was more consistent than Kiner (consistently at a lower level for 4 of the 5 years he was actually BA eligible), that he had one shoulder on each side of his peak while Kiner had 3 on each side, etc.

Keller 3 yr nc peak 167-64-63 shoulders 58-42
Kiner 3 yr nc peak 183-82-72 shoulders 54-45-40-31-21-16

IIf Kiner was "Keller lite" it was for four years that Keller spent largely in the training room. Keller was consistent during those shoulder seasons, it's true:

.275, .238, .267, .250, .314, .258 in 45-84-60-50-54 games over 5 seasons. I think you have to treat those 300 games as if they were 2 real seasons to get Charlie into Kiner's league.

If the knock on Kiner is he had too short of a career, then it's goodbye Charlie, too. If the knock is wasn't as good as Keller (with his 152-148 career OPS+ edge) for 5 years, I don't see it.

I was only pointing out that if WS shorts guys on weak teams, that Bob Johnson is not alone in that category.

Marc- This one I actually agree with, and is indeed a point in Kiner's favor. Not facing your own teammates helps players on good teams and hurts players on bad teams. I estimate for example a 2 percent bonus in win shares for a player on a .600 team.

I could just as easily point out that Keller's case depends on giving 2 years of WWII credit to a notoriously injury prone player

Those disinclined to give WWII credit to Keller surely won't be voting for him. Is there any evidence that he was injury prone before 1947? Seems to be a 140+ game a year player up till then.

Giving plausible 1944-1945 credit to Keller as non-war-discounted 167 OPS+ seasons (they were his age 27-28 seasons), and treating the 750 PA from 1947 on as one season, the OPS+ comparison would be (from bbref):

Keller  168-67-67-63-62-59-44-41-26
Kiner   184-84-73-56-46-40-32-21-17-16

Keller's advantage beyond 3rd best year is clear, as his career OPS+ would have been 155, a more significant lead on Kiner's 149.

It's interesting that you use BA to diss Keller- surely the greater emphasis on BA in the past had something to do with Keller's poor showing in MVP votes, though sabermetrically he was the best position player not named Ted at his peak.

In writing this up, I discovered something else I hadn't appreciated: the degree of HR suppression during WWII. During Kiner's best 5-year stretch, 1947-1951, the NL averaged 958 HR a year. During Keller's best 5-season stretch, 1941-1946 sans 1944, the AL averaged only 565 HR a year. If nothing else, Keller had terrible timing for a slugger.

While their names make Keller and Kiner a natural pair for arguments, we'll soon have Doby and Rosen to add to the fun. It's also interesting to consider who are the more recent best comps. Dick Allen? Albert Belle? Jason Giambi?