What John said makes sense when you think about it, there are 8 positions for split up offensive statistics and no more than two players at each position per team. In the period we are looking at there are four starters and a few other pitchers hanging around. Maybe we should look at something like top four for pitchers as the equivalent of leading one's position for an outfield player.

Maybe we should look at something like top four for pitchers as the equivalent of leading one's position for an outfield player.

I don't have a set quota, but I have been averaging about four pitchers on my ballot since I noticed the difference between the pitchers and the position players.

I probably was slighting pitchers the most before that.

1938 preliminary ballot:

1. Mickey Welch
2. Jake Beckley
3. Max Carey
4. Ben Taylor
5. Tommy Leach
6. Edd Roush
7. Carl Mays
8. Jose Mendez
9. Jim McCormick
10. Vic Willis
11. Dick Redding
12. Stan Coveleski
13. Dave Bancroft
14. Oliver Marcelle
15. Bruce Petway

New eligibles:
19. George Kelly--I'm going to re-evaluate the defensive ratings I'm using--they seem to favor first basemen unduly.
32. Lu Blue
37. Eddie Rommel

No one else joins the top 80 eligibles.

Here's how I rank the 1938-40 pitching candidates. My system relies mainly on ERA+ and IP, with slight bonuses for good OBP, TB, and RnF by pitchers.

1. Burleigh Grimes--this surprised me, but it looks like Grimes will debut on the top third of my ballot.
2. Wilber Rogan
3. Eppa Rixey--these two are personal favorites.

Those three will be firmly on my ballot. Then there are the borderline candidates that will be on the edge of my ballot:

4. Red Faber
5. Jack Quinn
6. Nip Winters

And the final group are still very good candidates, but will likely languish in the 25-40 slots of my queue.

7. Dolf Luque
8. George Uhle
9. Herb Pennock
10. Eddie Rommel

I currently have 7 pitchers on my ballot--so this figures to be an interesting few elections as these good newcomers are added. Needless to say, I'd still like to see a few more 19th-century hurlers inducted.

19. George Kelly--I'm going to re-evaluate the defensive ratings I'm using--they seem to favor first basemen unduly.

If your system likes first basemen this much, why aren't Sisler and Chance in your top 15?

Coveleski deserves a more thorough defense than I have time to compose today, but a few thoughts in response to Kelly in SD's comparisons.

First, as John Murphy has rightly pointed out, direct comparisons of Coveleski to pre-lively-ball pitchers are going to be unfair and misleading, almost as much as comparisons of post-1893 pitchers to pre-1893 pitchers without adjustments for playing conditions would be.

Second, most of Kelly's complaints about how Coveleski stacks up against elected pitchers and pitchers to come are irrelevant to the current election, because _none_ of the eligible players look all that good when compared to the standards of players already elected and of top players on the horizon. Heinie Groh doesn't hold a candle to Hornsby or to Frisch. Better not elect him then? Beckley doesn't look so good with Gehrig on the horizon? The HoM doesn't work that way. (If you want to give more support to players who have a claim to be among the all-time greats in some way, support Hughie Jennings. His 5-year peak is the only achievement by any of the eligibles that reaches significantly above the bottom third of the HoM. I like Jennings, but among the current eligibles, I like Coveleski's 11 good-to-great seasons a little more.)

Third, using Win Shares uncritically to evaluate pitchers in this era is a mistake. It doesn't do a good enough job of separating pitching and fielding contributions. Look at it, sure, but also look at WARP. Even if you don't want to use WARP's comprehensive metrics, look at its assessment of fielding-indendent earned run average, DERA. All the work I've done studying the impact of defensive efficiency on team defensive value suggests to me that WARP's DERA is on target here. Coveleski's DERA and ERA+ are almost the same, which supports my findings (and the findings of win shares), that his defensive support was about league average. Finally, look at how many times Coveleski placed among league leaders in innings pitched and ERA+, the two most important simple measures of pitching quality. He's better than any other eligible candidate, placing in the top 10 in each of these categories 8 times, 7 times placing in both in the same season. That's a pretty good basic indicator of the quality of his peak.

If your system likes first basemen this much, why aren't Sisler and Chance in your top 15?

Right now I'm using a defensive rating, between 0 and 100, based on RnF compared to the league. I'm using raw numbers instead of ratios, so since first basemen have more chances, they get a bigger bonus. It's only one of several factors I use to come up with my rankings, so it's not making much of a difference, but I'd still like to get it right.

Because Sisler's RnF was close to the league, he doesn't get much of a bonus. Plus he had a relatively short career. Chance had a pretty good career, but he falls short in the rate stats. Beckley and Taylor come closest to putting together the complete package, in my opinion.

Wouldn't a first baseman's range factor tell you more about the ground-ball tendencies of his pitching staff than it would tell you about the first baseman's own quality?

What "rate stats" does Chance fall short in? From where I sit, Chance's rate stats (especially on offense) are spectacularly good. The reason only a few people are voting for him is that he only had a 1200 game career.

Chris,

Thank you for taking the time to respond to my post. Even though things are settling down, they are still not up to normal.

Your points about what statistics Coveleski shows up highly in is something that I had forgotten to look at. I haven't had the opportunity to work up the little player profiles that I had posted back in the early 30s because of the activities of the last 6 weeks. Also, I still need to check out Chris J.'s run support information.

As to your comments regarding the comparison of current players with the elected HoMers or other active players: I understand we vote for the best eligible each year. To me to determine the best, I compare the candidates to the eligibles, to their place in history, and to where they rank among their own cohort group.

However, your post has made me go back to look at some additional information - run support, fielding support, in what categories did each of coming "glut" of pitchers dominate, etc.

Only possible newbie is Rommel:

RSAA - 217
ERA+ - 121
Neutral Fibonacci Win Points - 153
IP - 2,557

looks very similar, but inferior, to this guy:

RSAA - 239
ERA+ - 124
Neutral Fibonacci Win Points - 172
IP - 2,681

who is Urban Shocker, and since Shocker doesn't make my ballot, Rommel won't.

Now that I am examining the pitchers, I find I actually did a write up for Coveleski. Dang, I can't wait to get back to a regular routine...

While doing some research I came upon this bit of, I assume, coincidence: Many of the upcoming pitching candidates had horrendous run support. Of the 191 pitchers looked at on runsupportindex.blogspot, Eppa Rixey received the 160th best, Jack Quinn 163rd, Ed Rommel 175th, Dolf Luque 180th, and Dazzy Vance 184th. Some of these pitchers hit the daily double by having comparitively poor defensive support. Out of the 193 looked at on the site, Rommel had the 118th best, Jack Quinn 127th, and Vance 165th. No wonder Vance struck out so many.

This cluster of poorly supported pitchers - any reason or just coincidence? I ask because this will be the first group of candidates who did not receive average or better support. Just about every pitcher enshrined or discussed has been within 1 or 2 points of 100 or way over. Waddell is the main exception so far.

Chris Cobb-
I looked at the leadership boards on bb-ref for the pitchers from my earlier post. I think Vance does the best by far in terms of first place finishes. The number of times Coveleski appears in the top 10 for ERA and aERA+ is impressive. Overall, I don't see a great difference between Coveleski and the rest of the pitchers soon eligible, though I do see him with an edge.

Open question:
Once Coveleski is in, who of the following group do you also see as HoM worthy: Faber, Grimes, Hoyt, Luque, Mays, Pennock, Quinn, Rixey, Rommel, Shawkey, Shocker, or Vance?

HoM worthy: Faber, Grimes, Hoyt, Luque, Mays, Pennock, Quinn, Rixey, Rommel, Shawkey, Shocker, or Vance?

Once Coveleski is in, who of the following group do you also see as HoM worthy: Faber, Grimes, Hoyt, Luque, Mays, Pennock, Quinn, Rixey, Rommel, Shawkey, Shocker, or Vance?

My analysis of pitchers not yet eligible is not yet complete, but here's the order I have at present:

Vance
Rixey
Coveleski
Faber/Grimes
-----Probable in/out line
Luque
Shocker
Mays
Rommel
Quinn
Pennock
Shawkey

Faber, Grimes, Luque, Shocker, and Mays are the real border group for me. I'm still waiting to see Chris J.'s full analysis of the opponents for Grimes, Faber, and Luque. Grimes and Luque were paired with Vance and Rixey, and I want to see how much each member of these pairs benefited or suffered in w-l records and ERA+ because they tended to face the stronger teams/pitchers more. Rixey has enough innings that I think he goes in even if he faced weaker opposition. Luque and Grimes are on the fence (and Faber also could be affected by this factor). Shocker and Mays were excellent pitchers -- with them, it's just a question of how deep we will eventually go into this group. Given how tough the competition will be from 1941 through the 1950s, I don't see either of them standing a chance in the next 20 years, but after that, it's possible.

This cluster of poorly supported pitchers - any reason or just coincidence?

Fascinating question. I suspect three reasons for it. 1) Chance. 2) Improving competitive balance. More teams competently run and aiming to win means good players more widely distributed. 3) Growing importance of pitching as a component of defense means that pitchers have a better chance of standing out even when they are backed by poor fielders.

Haven't finished looking at the upcoming pitchers yet, but here's what I've got so far.

Vance
Coveleski
Grimes
Rixey
---- In/Out line about here -----
Luque
Faber
Mays
Cooper
Shocker

What "rate stats" does Chance fall short in? From where I sit, Chance's rate stats (especially on offense) are spectacularly good. The reason only a few people are voting for him is that he only had a 1200 game career.


Yeah, that didn't make sense, did it? I was typing on limited sleep yesterday, sorry. What I meant to say was that Chance's good rate stats weren't enough to overcome the length of his career--and I put a premium on career length.

Thanks for keeping me honest, OCF, and I'm going to continue my scrutiny of range factor. :)

Much Thanks for the response, Chris!

Does there happen to be another way that I can get info on the top players by position by year other than scrutinizing them manually?

Fun fact:

19 of Eppa Rixey's last 40 starts were against Pittsburgh.

Fun fact:

19 of Eppa Rixey's last 40 starts were against Pittsburgh.

Weird wild stuff. Just a coincidence, or did he like Forbes Field? How were is Pitt/non-Pitt splits?

Weird wild stuff. Just a coincidence, or did he like Forbes Field? How were is Pitt/non-Pitt splits?

No idea. Haven't checked. Did notice that both Waners, Vaughn, and Grantham were all lefties though.

He apparently spent his last three years as some sort of uberOrosco.

Does there happen to be another way that I can get info on the top players by position by year other than scrutinizing them manually?

Depends how you want to measure "top"? I have the top players per year in Win Shares and in TPR (the old one) from Total Baseball...

Here's my guess as to how this will play out for the eligibles before Grove, Hubbell, and Lyons arrive in 1947-1949.

EppaRixey
DazzyVance (Rixey's career, Vance's peak look like HoM material to me)

RedFaber (these two look about equal to me, which would imply Faber in as well)
StanCoveleski

<in/out line? if so Waddell and Griffith, probably Welch are likely to stay in the mix for a while.>

EddieCicotte
CarlMays
BabeAdams
WilburCooper
DolfLuque (ML only, Cuban achievements might bump him above the line)
HippoVaughn
Burleigh Grimes (maybe just me, but I think the 107 ERA+ isn't going to be enough)
EddieRommel
EdReulbach
UrbanShocker
WaiteHoyt
JackQuinn

Shocker lost some prime time to death. Rixey and Faber lost some prime time to WWI --an advocate might argue for two years.

--
John Murphy #99, October 30, 2004
among all nine positions, pitchers have the most difficulty achieving a high percentage of league leading performances. IOW, throughout their careers, Ruth and Wagner averaged more than 80% of their position's major league superiority, while I have Johnson just barely hitting 60%.

comments
- the main cause must be what everyone says or implies: the different numbers of regular pitchers and regular batter-baserunners in the league
- snap judgment is risky without considering the in/dependence of the leader categories at different positions.
- how does the pitcher data compare with that for fielding leaders at one position, eg third base?
- I guess that absolute pitcher performances are less stable, eg less consistent from first half to second half of season. Why? mainly because injuries are more frequently important for pitchers than for other players.

--
112. Kelly from SD, November 01, 2004
Many of the upcoming pitching candidates had horrendous run support. Of the 191 pitchers looked at on runsupportindex.blogspot, Eppa Rixey received the 160th best, Jack Quinn 163rd, Ed Rommel 175th, Dolf Luque 180th, and Dazzy Vance 184th.
. . . this will be the first group of candidates who did not receive average or better support.

Maybe there was a change in use of star pitchers. Was ace-against-ace newly in vogue for a decade or three?

comments
- the main cause must be what everyone says or implies: the different numbers of regular pitchers and regular batter-baserunners in the league

It has nothing to do with that, Paul. In fact, the results would be exactly the same regardless of how many pitchers there were. What I'm doing is taking what a pitcher did in a season and then divide it by the major league best figure to see how far he deviated from the major league leader. Pitchers consistently fall short throughout history compared to the other eight positions using this method. My theory is that the demands of pitching (or abuse) makes it extremely difficult for consistency of quality (which underrates them, IMO).

John,

I'm not sure I follow your method.

Would your results for position players necessarily be unchanged if you looked at larger groups of position players together: say all outfielders or all 2b/3b/ss?

Would your results for position players necessarily be unchanged if you looked at larger groups of position players together: say all outfielders or all 2b/3b/ss?

Absolutely not.

Here's an example:

Player A has a .300 BA; Player B (the major league's best hitter) has a BA of .350. Dividing .300 by .350 gets you 86% (his "percentage of perfection"). Merrit Clifton created this technique years ago (and can be found in The Hidden Game of Baseball.

The results wil be the same if you have two people in your pool or 2,000.

There are many great position players in baseball history who easily averaged over 60% of major league leading performance throughout their careers, while a Walter Johnson can barely get over 60%. Few can best 50%.

OK. Now I understand more of what you're doing, but I'm still not convinced that pool size isn't a factor.

If you're studying batting average, to get your "percentage of perfection," are you comparing the player only to players within his own position, or are you comparing him to all position players? If the the former, then by enlarging the pool, won't the typical average of the "best hitter" in that group rise, thus dropping players' typical "percentage of perfection"?

Pitchers consistently fall short throughout history compared to the other eight positions using this method. My theory is that the demands of pitching (or abuse) makes it extremely difficult for consistency of quality (which underrates them, IMO).

Not sure I buy this. If you look at the "low end" or worst 'regular' pitchers vs. the worst regular position players, I think you'll find that by your method pitchers will consistently EXCEED the "% of perfection" vs. the other 8 positions.

OK. Now I understand more of what you're doing, but I'm still not convinced that pool size isn't a factor.

If you're studying batting average, to get your "percentage of perfection," are you comparing the player only to players within his own position, or are you comparing him to all position players? If the the former, then by enlarging the pool, won't the typical average of the "best hitter" in that group rise, thus dropping players' typical "percentage of perfection"?

Pool size is a HUGE factor. Just think if you had a league of just 2 position players, with roughly equal abilities. What would their career "% of perfection" look like? Something like this:

Year1 100%
Year2 95%
Year3 100%
Year4 100%
Year5 90%
Year6 97%
etc.

Increase the pool size, and a player will almost NEVER get 100%, will rarely be a 90% player, etc.

If you're studying batting average, to get your "percentage of perfection," are you comparing the player only to players within his own position, or are you comparing him to all position players?

Within his own position.

If the the former, then by enlarging the pool, won't the typical average of the "best hitter" in that group rise, thus dropping players' typical "percentage of perfection"?

I'm not following you here, Chris. If I'm going by position, the pool would be smaller since the there are fewer players at each position. Again, it doesn't matter because this is not a standard deviation excercise (where the size of the pool would definitely matter).

But let's say you are correct for arguments sake. Wouldn't the results even out in the long haul? I'm not noticing a discrepancy between the position players, but only between the pitchers and position players. This problem occurs in the same way in 1880 (when the number of pitchers were much smaller) as in 1980.

BTW, for my experiment, I was comparing a player's WS (with my adjustments) to the best WS major league total for his position in each particular season. I was only suggesting BA for the example above.

Pool size is a HUGE factor. Just think if you had a league of just 2 position players, with roughly equal abilities. What would their career "% of perfection" look like? Something like this:

Year1 100%
Year2 95%
Year3 100%
Year4 100%
Year5 90%
Year6 97%
etc.

Increase the pool size, and a player will almost NEVER get 100%, will rarely be a 90% player, etc.

This is incorrect. Your two players would still have the same % of perfection regardless of the pool size.

KJOK, let's do it this way for an example. In 1927, Ruth had the most homers in the majors with 60. If we wanted to find out Gehrig's homer "% of perfection" for that season, we would divide his 47 by 60 to give us 78%. Lazzeri's 18 homers would give him only 30% (18/60). It doesn't matter what the other players are doing in the leagues or the quantity of players.

Now, hypothetically, if they expanded the leagues in '27 to forty teams, even though the talent pool would be diluted (and cause a higher SD between the best and worst players), the percentage between Ruth and Gehrig or Ruth and Lazzeri would still be the same.

John,

Let's shift this discussion to win shares in 1927. If you are comparing Lazzeri to other second basemen in the major leagues, the leader is Hornsby with 40, so Lazzeri scores 60%. But if you compare him to all position players, he gets compared to Ruth as well, which drops his percentage to 53%. The larger the pool, the higher the peak value is likely to be. Since, after 1880, there are always more pitchers than players at any single position, they have a larger pool and hence are more likely to have a very high peak score in any given season than any _single_ position is.

Another pair of factors if you are indeed working with win shares:

1) the zero point for hitters is lower than the zero point for pitchers, as jimd has shown, so position players get more win shares between zero and average than pitchers do, thus making an average position player's percentage of top scores higher, on average, than pitcher's.

2) (a more speculative point) Because the better a pitcher is, the more innings he tends to throw, pitchers tend to peak more highly, relative to average, than position players do.

Anyway, unless I am still misunderstanding the way you are setting up your calculations, I'm pretty sure now that the differences between pitchers and position players are not, for the most part, intrinsic to the way pitchers are used (except for point 2 above) but are artifacts of the comparison set and the win-share system itself.

Since, after 1880, there are always more pitchers than players at any single position, they have a larger pool and hence are more likely to have a very high peak score in any given season than any _single_ position is.

There are a couple problems with this, Chris. The first one is there is no difference between the pitchers' results from the 1880s to now. The second problem is we're talking about the trend throught baseball history, not just for one season here or there.

I'll see if I can expand upon this tomorrow since I'm going to bed now.

Depends how you want to measure "top"? I have the top players per year in Win Shares and in TPR (the old one) from Total Baseball...

Measuring "top" by Win Shares, TPR, WARP, OPS+, or even an amalgam of unadjusted stats is perfectly fine by me. If there is organized statistical info regarding year by year performance at each position by which I can form my own opinion then I'm happy. That means I don't have to dig through Retrosheet and organize all of that myself!

Anyways, I've finally figured out how I'm going to cast my ballot for 1938. I'll post it here for now, and if there are no undue concerns I'll put it in the Ballot thread tomorrow!

1. Heinie Groh—Of all the candidates, Groh is the one I feel most comfortable about. Only Jennings had a higher Win Share total in his five best consecutive seasons.
2. Stan Coveleski—See comments on Rube Waddell.
3. Hughie Jennings—Jennings had the highest peak of any of the available candidates. His peak also comprised of 73.3% of his warp1 and 70.1% of his WS. (Decided to rate Jennings higher after Cobb’s persuasive stumping.)
4. Max Carey—Sticks out like a sore thumb among a similar group of center fielders with very good career and good peak numbers. His 351 Win Shares is the highest among HoM candidates (non-19th century pitchers division).
5. Lip Pike—After Negro Leaguers, Pike and Cravath are the hardest for me to judge. In the end, I gotta give credit to a man who reputedly outran a horse, though. . . .
6. Larry Doyle—Siding with Win Shares' interpretation of his defense, combined with an adjustment for Childs’s 1890 AA competition, gives Doyle the edge between the two second basemen.
7. Cupid Childs—See comments on Larry Doyle.
8. Rube Waddell—Coveleski played 14 seasons; Waddell played 13. Coveleski accumulated 245 WS and averaged 29.85 per season; Waddell accumulated 240 and averaged 30.59. They have the same dERA: 3.58. The difference between the two, other than strikeouts, is the comparison between their peers, but on “The Ballot of the Very Good” this is the spot that I think Waddell deserves.
9. Jose Mendez—Has been compared to Mr. Rube Waddell.
10. Tommy Leach—Very good career, good peak, a part of the underrepresented third baseman class. . . .
11. Dick Redding—Best Negro Leaguer pitcher from 1917 – 1919, according to Bill James, great nickname to boot. . . .
12. George Sisler—First half his career established his HoM candidacy, second half pushed him past the all-peak dudes (exempting Hughie "Ee-Yah, I have an insane peak" Jennings, of course).
13. Spotswood Poles—Referred to by some as “The Black Ty Cobb.”
14. Gavy Cravath—“He played ball, and lived his life, with a minimum amount of effort and nervous energy.” Cravath gets extra credit for his minor league performance (obviously), especially as the star of the Minneapolis Millers. And it’s Gavy, right?
15. Bill Monroe—Negro Leaguers are definitely the hardest to figure out. I’m reasonably confident on where I have Mendez, Redding, and Poles (although it’s certainly not set in stone), but am less so concerning Monroe. For now he gets the token 15 spot ahead of “The Centerfielders.” (An interesting class in and of themselves. . . . )

BTW, I apologize for the dada name. I'm hoping it won't be necessary soon. . . .

Much Thanks,
Tiboreau FKA The Erskine Thomason of UBW

Quick question: How do various people create a "season" for pitchers? For hitters, people seem to go with per 162 games or 648 PA. The one I've been using is per 40 starts with a start counting as 1 start and a relief appearance as .6 starts. I am worried about how to weight the relief appearences as pitchers are having more of them while throwing fewer complete games. Any advice?
Thank you.

2) (a more speculative point) Because the better a pitcher is, the more innings he tends to throw, pitchers tend to peak more highly, relative to average, than position players do.

Maybe. I have never said my theory was in any way 100% correct, only that something was causing the pitchers to deviate from the position players. This certainly makes more sense to me than the larger pool theory.

The ballot looks good, ?ïß?rêŵ. Welcome!

Tiboreau FKA The Erskine Thomason of UBW

Ah.... Tiboreau... that's what that is supposed to be. I guess it does kinda look like that. I think I was trying to translate each letter and came up with "Gissfreom".

Yeah, I like colorful names, but I don't know how to type some of those characters with my keyboard. :-)

As John Murphy inferred, I had missed the gist of his (Merritt Clifton's?) statistics, which is share of league maximum. I agree with Chris Cobb and KJOK that there is a statistical fallacy in John's reasoning re the effect of population size but . . .

Ruth and Wagner averaged more than 80% of their position's major league superiority, while I have Johnson just barely hitting 60%.

I don't believe the relatively large pitcher population can be the main cause of that great difference in magnitude (as I said in #123).

So, why do the relative measures --plural, or the one summary relative measure?-- differ significantly by fielding position?

Likely statistical causes include differences by fielding position in the mutual dependence, the unit size or typical magnitude, and the meaning of zero magnitude for the leader categories.

Yet, I guess the principal cause is that raw statistics are less stable for pitchers. The raw pitcher stats differ more from season to season, from first half to second half of any season, etc. Why? Partly because injuries are more frequently important for pitchers than for other players, so that pitcher playing time (or playing time at any ability level) is less stable.

Should I elaborate after John returns to the subject? If so, which points?

I agree with Chris Cobb and KJOK that there is a statistical fallacy in John's reasoning re the effect of population size but . . .

Let me stress again, if I created a project using standard deviation, population size would definitely matter. I would have to set up some type of criteria so as not to create a statistical illusion. I just don't see it with the measure I have reintroduced here.

Yet, I guess the principal cause is that raw statistics are less stable for pitchers. The raw pitcher stats differ more from season to season, from first half to second half of any season, etc. Why? Partly because injuries are more frequently important for pitchers than for other players, so that pitcher playing time (or playing time at any ability level) is less stable.

Paul, it appears we're pretty much on the same page here now. My point was that injuries may be causing this problem, which in turn may underrate certain pitchers with this project. IOW, if Walter Johnson and Stan Coveleski were position players (work with me here :-), I think the difference between the two would be smaller because injuries and other factors are less of a concern.

This is incorrect. Your two players would still have the same % of perfection regardless of the pool size.

Perhaps I'm being dense here, but I don't see how you could possibly say this. I just showed you an example where if the pool size was 2 relatively equal players that about 50% of the time they'll have a "% of perfection" of 100%, as they'll be the top player in about 50% of the seasons.

If you had a pool of 100 relatively equal players, then a player would have a "% of perfection" of 100%, realizing there would sometimes be ties for 1st, down around 1% of the time! POOL SIZE is tremendously important for any statistic of this type....

If you had a pool of 100 relatively equal players, then a player would have a "% of perfection" of 100%, realizing there would sometimes be ties for 1st, down around 1% of the time! POOL SIZE is tremendously important for any statistic of this type....

KJOK, have you actually tried the formula? If you had 100 relatively equally players, then there percentage of perfection would still be around 100%.

If I had 100 Babe Ruths who led the majors in '27 with 60 homers, each Bambino would have a percentage of 100%. 60 (which each Behemoth of Bust would have) divided by 60 (the major league total) would equal 1.00 (or 100%). The results would be exactly the same if there were only two Bambinos. There's no way of getting around this, Kevin.

For some reason, you must be thinking that the Merrit Clifton formula is average-based. It's not.

But even if what you were saying was true, the problem with pitchers compared to position players is the same pre-1890 as it is now. Your theory would state that, since there are more pitchers in the pool now, the percentage of perfection must be lower now than it was during the 19th century. This is not the case.

Yet, I guess the principal cause is that raw statistics are less stable for pitchers. The raw pitcher stats differ more from season to season, from first half to second half of any season, etc.

As best I can estimate, the typical season-to-season fluctation in pitcher performance is about 35% larger than the typical fluctuations for position player performance.

As best I can estimate, the typical season-to-season fluctation in pitcher performance is about 35% larger than the typical fluctuations for position player performance.

That sounds about right, Eric.

I think what KJOK is saying is that even though the players are of equal quality, there is random variation in their performance. With a sample of about 40 starting pitchers in the league every season, at least one pitcher every year is going to be about 2 standard deviations above average almost every year. Because the position player sample size is 1/5 of that, those fluke seasons happen only every 5 years or so. IOW, the position players score will be much closer to 100 most seasons, simply because fluke seasons happen less often realtime even though they occur at the same rate wrt the base population.

IOW, the position players score will be much closer to 100 most seasons, simply because fluke seasons happen less often realtime even though they occur at the same rate wrt the base population.

If this were the case, I would be noticing a difference between the pre-1890s pitchers and todays pitchers because of the huge difference in the size of the pitching pool, but I'm not seeing this. If anything, the percentage of perfection mught be higher now.

I should state that KJOK might have a point if I had been using individual counting stats (triples, doubles, earned runs, etc.), but my analysis is based on the sum total of each player's worth compared to best major league pitcher.

It should be trivial to prove or disprove the claim that pool size doesn't matter with reference to actual examples. Calculate the index of perfection for a group of players, say 2 lf, rf, and cf comparing them only to their position, then calculate their index of perfection comparing them to all outfielders. Compare the results. It seems a bit silly to have a theoretical argument on this particular point. If I weren't crazy busy at present, I'd do it myself.

That said, I suspect that, if such a study were done, it would show that position players in larger pools have lower indices of perfection than they do in single-position pools, but even grouped in 3-to-5 position pools, they still have higher indices, by and large, than pitchers do. The question is: how much higher?

It should be trivial to prove or disprove the claim that pool size doesn't matter with reference to actual examples. Calculate the index of perfection for a group of players, say 2 lf, rf, and cf comparing them only to their position, then calculate their index of perfection comparing them to all outfielders. Compare the results. It seems a bit silly to have a theoretical argument on this particular point. If I weren't crazy busy at present, I'd do it myself.

Chris, I have no doubt that you're right about this and wasn't arguing against it in that instance, but it doesn't really address my initial point.

That said, I suspect that, if such a study were done, it would show that position players in larger pools have lower indices of perfection than they do in single-position pools, but even grouped in 3-to-5 position pools, they still have higher indices, by and large, than pitchers do. The question is: how much higher?

That's my question, too. But whatever the reason, I think it's underrating pitchers for this project. That's why I brought it up here so that everyome was aware of it.

I think what KJOK is saying is that even though the players are of equal quality, there is random variation in their performance. With a sample of about 40 starting pitchers in the league every season, at least one pitcher every year is going to be about 2 standard deviations above average almost every year. Because the position player sample size is 1/5 of that, those fluke seasons happen only every 5 years or so. IOW, the position players score will be much closer to 100 most seasons, simply because fluke seasons happen less often realtime even though they occur at the same rate wrt the base population. Now why couldn't I have said that! That IS the point I was trying so feebly to make.

Now why couldn't I have said that! That IS the point I was trying so feebly to make.

I've been in the same position myself on occasion. :-D

That's my question, too. But whatever the reason, I think it's underrating pitchers for this project. That's why I brought it up here so that everyome was aware of it.

Given that 5 of the top 7 players on my ballot this year are pitchers, I certainly agree that they are in danger of being underrated, and I think you are pointing towards one reason.

How do you take account of this difference in making your rankings?

How do you take account of this difference in making your rankings?

The very unscientific thing to do of doubling my results from my analysis. :-) This brings the Johnsons, Youngs and Mathewsons in line with the rest of the top position players on my list and brings the rest of the pitchers up considerably. My unadjusted system was very unfriendly to pitchers, so some type of adjustment was needed.

I'm also averaging about four pitchers on my ballot since I have made the change (I have five for '38), so that feels right to me.

If I have time, I would like to make the adjustment on a more sound basis, but for now, it works for me.

1.
John Murphy #149, regarding the 80%/60% effect, whereby the best career pitchers uniquely achieve only about 60% of best, season by season.

I think it's underrating pitchers for this project. That's why I brought it up here so that everyone was aware of it.

OK. I doubt that many people here are using "percentage of perfection" (%Perf) or any sibling implicitly, not to mention explicitly. On the other hand, I concede that a few such voters might distort the results. (Suppose 80% of the voters follow a strict quota such as one pitcher in the top 3 and two in the top 6 on every ballot, and 20% implicitly use a %Perf sibling that is biased against pitchers. Then no marginal pitchers will ever be elected.)

2.
Percentage of Perfection

%Perf doesn't have any general validity --ie, validity across fielding positions or leader categories. It does have validity across positions given a single category with special features, and single-season Win Shares is such a category, I'm sure. (John Murphy evidently applies %perf to that category --among what others, I don't know.)

A crucial general problem with %Perf is that it measures the share of distance from zero to perfection, but zero has no general meaning. Zero batting average, zero winning percentage, zero K/BB, zero assists, zero complete games, infinite ERA.

John,
I think comparing to the top performer puts a lot of dependence on year to year random variablity in your rankings, if you compared instead to the #5 or #10 performer each year, you would get a lot more year to year consistency in your comparison level.

John Murphy #140 in reply to me:
>>[quoting Paul Wendt:]
Yet, I guess the principal cause [of the 80%/60% effect] is that raw statistics are less stable for pitchers. The raw pitcher stats differ more from season to season, from first half to second half of any season, etc. Why? Partly because injuries are more frequently important for pitchers than for other players, so that pitcher playing time (or playing time at any ability level) is less stable.
<<

Paul, it appears we're pretty much on the same page here now. My point was that injuries may be causing this problem, which in turn may underrate certain pitchers with this project.

OK.
I don't believe that the 80%/60% effect concerning %Perf is the main reason why the Hall of Merit is (for sake of argument) inducting too many leftfielders and too few pitchers, and I am skeptical that it is an important reason. Rather, I think the main reason, maybe the overwhelming reason is that few pitchers have comparably long, quality careers. Although the number of HOM pitchers seems a bit low, many have voiced the concern that Spalding, Caruthers, Rusie, at least, don't belong because their careers were too short. Non-pitchers careers of comparable quality and length are still in the Foyer, if that, but no shortage has been inducted.

Rather, I think the main reason, maybe the overwhelming reason is that few pitchers have comparably long, quality careers.

And why is this? Because maybe pitchers get chewed up by the game more than the other eight positions? Because it's much easier to have a long, quality career as a position player?

FWIW:

Eppa Rixey.

Joe Dimino wrote atop 1934 Ballot Discussion.
Pretty amazing class, 5 easy HoMers, 2 more very strong candidates and a tweener. Discuss.

I thought that Coveleski might be the tweener, but that appears to be Ben Taylor (NeL) with Coveleski and Torriente the strong candidates.

And why is this? Because maybe pitchers get chewed up by the game more than the other eight positions? Because it's much easier to have a long, quality career as a position player?

Probably, except Catcher is probably more like Pitcher than the other position players in this regard, which is why Bresnahan and Schang need more love....

Especially Schang! Wimping out to the outfield half the time doesn't do it :-))

Probably, except Catcher is probably more like Pitcher than the other position players in this regard, which is why Bresnahan and Schang need more love....

You're right. All positions are not exactly the same.

Because I apparently feel like doing this sh1t tonight:

Red Faber.

Faber, Willis and Rixey are all close for me, with Covaleski, Mays and Cicotte a little better and Grimes, Waddell and Vance a little worse (Grimes being the odd man out in the third group, obviously.) There's certainly a dime's worth of difference between them, but I'm not sure I'd go as high as a quarter. There will be wholly unjustified bumps and un-bumps in my scoring from where the bottom of the ballot happens to come in the long list.

Leever's in the Covaleski/Mays/Cicotte group, Welch is well above it and Griffith a little above it.