Regarding the AA discount, the number I see the most is 4%. How is that enough to push Stovey and Browning down so far on so many ballots? When James introduced MLE in his 1985 Abstract, he used 18% as the difference between the majors and AAA minor league ball. Using that number, if you assume the average value of the 8 NL teams in 1961 to be 1 and add 2 expansion teams with an average value of .82, you get a 1962 league average of .964, or a 3.6% discount. So when it comes time to compare the stars of the 1960?s with the stars of the 1950?s, will they be viewed as negatively as we are viewing the AA now?

A quick check of the Davenport adjustment for NL expansion of the early 60's.

1961 -.10
1962 .06

This metric is runs off of a 4.50 base. 62/61 is 4.56/4.40 == 1.036

dan b, you hit it dead center.

Converting the above AA/NL comparison numbers into percents. The base for 1881 is 4.92. The resulting numbers in chrono order are: 30,20,13,9,7,9,6,8,15,11. The AA may have ranged from AAA to AAAA in relative quality.

Instead of "the resulting numbers", I should have typed "the resulting percents".

My understanding of the project is not to name the top 200 players of all-time but the 200 players who had most value to their teams, keeping in mind the relative league strength in a particular era. With this in mind I don?t see how you can put 6 or 7 AA players on a ballot at the expense of NA players. Sure there may be more 80?s and 90?s players HoM worthy as baseball expanded but I truly believe each era should get fair representation. Therefore, I would be upset if players like McVey and Pike would not make it in the HoM at the end of the road, because some people are using too steep a timeline adjustment.

Apparently, Mark is on the same page as Bill James in the NHBA. James entirely dismisses NA performance, while seemingly not discounting AA performance at all.

Wright, Barnes, Spalding, Start, etc do not make his lists of top 100 at each position. Looking now at the highest ranked players in the NHBA on our current ballot:

CF-38th Browning
LF-39th Stovey
2B-39th Richardson
SS-43rd Glasscock
3B-45th Williamson
LF-48th O'Neill
C--49th Bennett

Pitchers
45th Radbourn
54th Keefe
82nd Mullane
88th Caruthers
97th Bond

I don't AT ALL agree with this treatment of players who starred only a decade apart, so take it with several pounds of salt.

jimd and dan b, you are confusing two different things. One thing is discounting within a season for a weaker vs. a stronger league. Another thing is discounting over time (1871 vs. 1881, 1950s vs. 1960s, 1960-1961, whatever).

To dan b, 4% is not enough for the reasons Tom H explains. And to those who have the AA guys on their ballots and no NA guys, your timeline adjustment is a cliff. The NA was just as good as the AA; the AA was not as good as the NL.

MattB: "BTotherW: I have posted my ballot and am immediately antsy. I've got Dan Brouthers and Buck Ewing entering my ballot next week. Am I missing anyone I should be considering?"

RobC's query gave us the list below. To that you should add negro George Stovey. And it also didn't capture Ed Stein, no big deal.

1902
DAN BROUTHERS,1896,1904 -> 2
BUCK EWING,1896,1897 -> 1
SHORTY FULLER,1896
CONNIE MACK,1896
TOMMY MCCARTHY,1896
CHIPPY MCGARR,1896
DOGGIE MILLER,1896
SAM THOMPSON,1896,1897 -> 3,1898 -> 14,1906 -> 8
AD GUMBERT,1896
ADONIS TERRY,1896,1897 -> 1

This is also the point we have to try and nail down the allowance for token appearances. Sam Thompson is eligible in 1902--or is he? I say yes.

I would say 1904 for Thompson. 14 seems like more than token to me, especially without any off years before it. It looks like he was still trying to play and just couldnt. 1906 is more obviously token. Since no specific committee was ever set up, I guess we are a committee of the whole. Anyone else have an opionion? Thompson will be on my ballot, so this is an important decision. Looking at it quickly, he will be *high* on my ballot, so it is very important what year he becomes eligible.

Apparently, Mark is on the same page as Bill James in the NHBA. James entirely dismisses NA performance, while seemingly not discounting AA performance at all.

I agree, but, except for Ross Barnes, I still think he would have some NA guys on his ballot. At least that is the impression I have of him from some of his books.

Mark, I take your point that the difference between 2B and 1B is not what it is today. But surely the typical 1Bman in 1885 hit better than the typical 2Bman!?! I haven't submitted my ballot yet, so maybe it's back to more research this evening.....

First base was the easiest position. It also had the lowest attrition rate, too. I'm puzzled also, Tom. Maybe I'm missing something.

At any rate, Stovey wasn't Anson, Brouthers or Connors.

Raise your hand if you think way too much attention is being paid to me, and my ballots...

LOL

The Constitution currently reads:

"To discount token appearances, a player becomes eligible 5 years after the first time he plays fewer than 10 games in the field or pitches in fewer than 5 games, assuming he never plays in 10/pitches in 5 games again. If he does play in 10/pitch in 5 games later in his career, the HoM ballot committee will determine in which year the player?s HoM eligibility begins."

The intent of this rule seems to me to make the "retirement year" the last year in which a player played 10 games unless special circumstances warrant an exception (e.g., he played exactly 10 games 15 years after he "really" retired). Absent a consensus to the contrary, Thompson last played 10 games or more in 1898, and should be eligible in 1904.

I agree with RobC's analysis. This is a pretty clear example of "non-token"-ness. More than ten games. Only one intervening year. If Thompson had played 10 games in 1906 instead of 8, I'd be voting the other way and saying that his 1906 performance was "token" even though it was ten games. But, as it is, he should not be eligible until 1904.

I am also happy to space out the new eligibles, and will already be inserting Brouthers, George Stovey, and Buck Ewing somewhere on my ballot.

I moved this discussion to the "Token Appearances" thread.

Mark, I think the right amount of attention is being paid to issues like timeline adjustments and discounting for weaker leagues within seasons. These are important issues and will be important during WWII and in evaluating Negro Leaguers etc. It ain't about you ;-)

And I say Brouthers in '02 and Thompson in '04.

And Tommy McCarthy...lemme see, he's in the real HoF right? Musta been a hell of a player.

Guess I'm looking for some empirical evidence that AA was weaker than the NL, from which to derive a discount that has some merit. The Cramer chart in The Hidden Game doesn't indicate NL superiority, for all of the aforementioned reasons (including Cramer's own disclaimer). The standard deviation analysis (i.e., smaller stdev is better) doesn't indicate NL superiority because AA often has a smaller stdev of OPS than the NL in the same year. I guess the raw OPS numbers could be used, but that could indicate strong pitching rather than weak hitting.

In my ballots, I have been applying an AA discount to a career, not on a year-by-year basis. If someone played all of his career in the AA, I would apply a 10% discount. I pro rate the discount for players who have partial careers in the AA. I pro rate based on Win Shares -- that is, if 50% of total Win Shares come from AA seasons, that's an overall 5% discount.

Why am I doing this? I don't know. It seems okay (and it's simple), but I'm looking for some harder evidence b/c I'm not quite comfortable. I'd like to go to a season-by-season adjustment, because in 1886 the AA looks stronger than the NL but in 1884 the AA looks weaker than the UA (based on OBA, SLG and OPS). But without something to back it up, a season-by-season adjustment might just make things worse.

Is there a mathematical statistician among us who could point us in the right direction for measuring groups of data like NL hitting statistics and AA hitting statistics. Is there merit to a lower standard deviation of OPS indicating league strength?

Thanks, Mark. That's what I've been thinking about stdev -- that it doesn't mean anything the way it has been discussed here with respect to league quality.

Looking at BP cards, the discounts can be severe. Pete Browning's career EqA is only down 10%, but his WARP2 is 40+% worse than his WARP1.

Also noticed on the BP cards that the league strength appears to affect fielding runs. Does that make sense?

The following percents are derived from the Davenport DERA offset.

NL dominance over AA (1882-1891)
30, 20, 13, 9, 7, 9, 6, 8, 15, 11

AL dominance over NL (1901-1930):
-6, 8, 7, 5, 5, 4, 6, 3, 5, 3 (oughts)
3, 0, -1, 0, 1, 4, 5, 3, 5, 4 (teens)
5, 7, 6, 7, 7, 5, 3, 1, 3, 0 (twenties)

NL dominance over AL (1931-1990):
5, 6, 4, 4, 2, 4, 2, 0, 2, 4 (thirties)
-2, 4, 3, 4, 4, -2, 0, -1, 1, -1 (forties)
-3, 0, 0, 3, 3, 6, 5, 8, 5, 4 (fifties)
6, 3, 4, 6, 6, 5, 4, 3, 2, 4 (sixties)
3, 3, 2, 4, 1, 0, 3, 2, 3, 5 (seventies)
3, 3, 3, 0, 0, 1, 1, 4, 3, 1 (eighties)

AL dominance over NL (1991-to date)
2, 6, 3, 2, 5, 6, 6, 8, 1, 3 (nineties)
1, 1

NL over UA 1884: 25

PL over NL 1890: 3

AL over FL 1914-1915: 13

This puts the AA vs NL numbers into a larger perspective.

The stratified NL of the dead-ball era, the NL of the early 20's, the AL of the late 50's/early 60's, the NL of the mid-90's; in each of these eras the gap between the leagues has approached the magnitude of the gap between NL and AA.

<Yes, because they're presenting Fielding runs above replacement/average. If they're adjusting the baseline based on the difficulty of the league, then the marginal numbers will change accordingly. >

Yeah, but not the same adjustment, right? I mean, the fact that an AA average or replacement hitter is 10% worse than his NL average or replacement hitter counterpart doesn't mean the average or replacement player is consequently 10% worse in the field. It's a hell of a lot harder to be a good hitter than to be a good fielder.

>The stratified NL of the dead-ball era, the NL of the early 20's, the AL of the late 50's/early 60's, the
NL of the mid-90's; in each of these eras the gap between the leagues has approached the magnitude of
the gap between NL and AA.

jimd, I don't understand this statement. The NL exceeded the AA +10% or more 5 of 10 years, plus there's a 9 and two 8s. Only three times in 100 years do I see an 8% difference between the AL and NL. What am I missing?

Pete Rose has more Win Shares than all but 11 MLers (Ruth Cobb Wagner Aaron Mays Young Speaker Musial Collins Mantle WJohnson). He is behind all of them in WARP(3) as well (how's that for consistency between tools), but in WARP3 he also trails Morgan, Clemens, Schmidt, Williams, Henderson, and Lajoie, and possibly others but that's all I found.

That's why I take into account Win Shares per Game. Morgan, Clemens, Schmidt, Williams, Henderson, and Lajoie kick Rose's butt with that category. When I combine them together, Rose falls back down the list behind all the players that you mentioned.

I'm still obsessed with the league quality analysis regarding the AA and would like your thoughts on the following:

In an attempt to compare apples to apples, I made a list of every AA player that had at least one 140+ at-bat season in the AA and at least one 140+ at-bat season in another league. Once I had the list, I compiled all of those players' seasons in which they had 140+ ABs in the AA and 140+ ABs in another league. This produced a list of 1,622 seasons -- AA (541), NL/AL (970), PL (57) and NA (54).

I then calculated the OBA, SLG and OPS figures for these seasons, grouped by league. The results should tell us something about how AA hitters fared in the AA vs. how they fared in other leagues. Here are the results for the AA vs. NL/AL with standard deviations in parentheses. Sorry I haven't figured out how to format this nicely.

Category ---------AA--------------------NL/AL
Avg. Age ---------27.1 (3.8)------------28.8 (4.3)
# Seasons --------541 ------------------970
OBA -------------.324 (.057) ----------.329 (.064)
SLG ------------ .360 (.079)-----------.363 (.080)
OPS -------------.684 (.129) ----------.692 (.138)

The players fared slightly better in the NL/AL than the AA. I was concerned about (at least) two things from this analysis: (1) the NL/AL data included seasons that were fairly remote from the beginning and end of the AA, and therefore since quality of play was rapidly changing, those outlying seasons might not be indicative of anything and (2) the average age of the NL/AL was almost 2 years older -- which could have pushed the numbers one way or another.

So, I excluded pre-1877 seasons and post-1896 seasons (i.e., five years on either side of the AA) to eliminate the outlier seasons, and got the following:

Category ---------AA--------------------NL/AL
Avg. Age ---------27.1 (3.8) -----------27.7 (3.6)
# Seasons --------541 ------------------781
OBA -------------.324 (.057) ----------.328 (.066)
SLG -------------.360 (.079) ----------.366 (.084)
OPS -------------.684 (.129) ----------.694 (.143)

The NL hitting got slightly better, but the standard deviation increased. That filter took care of my first concern and alleviated the second by reducing the average age of the NL/AL. Still, there are 240 more NL seasons in there. So, I tried the same thing by excluding pre-1879 seasons and post-1894 seasons (i.e., three years on either side of the AA), and got the following:

Category ---------AA--------------------NL/AL
Avg. Age ---------27.1 (3.8) -----------27.5 (3.5)
# Seasons --------541 ------------------653
OBA -------------.324 (.057) ----------.322 (.065)
SLG -------------.360 (.079) ----------.361 (.085)
OPS -------------.684 (.129) ----------.684 (.143)

Now there are really no outlying seasons, the age difference is insignificant and the number of seasons is closer.

Would like to know your thoughts on what this means, if anything, about league quality.

I've compared players (good and not-so-good) who hit in both leagues fairly regularly and found that they hit about the same in the NL/AL than in the AA. I'm not arguing it is conclusive, but isn't this some evidence that I shouldn't be discounting AA players as much as 10% (like I've been doing), since the discount is presumably trying to ratchet down the AA numbers to what the players would have done in the NL?

In fact, if the numbers above mean anything, these AA players would have done about the same in the NL.

I have been looking at the AA vs. other leagues in a more qualitative way, and have basically come to the conclusion that it does not make any sense to discount players who were successful in the AA between 1884 and 1889.

The first two years of AA play were fairly weak, and should be discounted somewhat. The last two years were very weak, but as I look at the rosters, almost none of the "AA stars" were in the AA in 1890 or 1891. It was a bizarre compilation of Jocko Milligans and Mox McQueerys and (bizarrely) Dan Brouthers. It was sort of like the last season of "A Different World", when all the regulars had graduated, and it was all new kids who had never even heard of Denise Huxtable.

So, even though those years need discounting, there aren't that many players who it applies to.

1884 every league was watered down, not just the UA, and the other 4 years any difference between the NL and AA is within the margin of error. Especially with whole teams switching leagues, it is not clear exactly what is being compared to what.

I will certainly be re-considering at least Harry Stovey next year, and probably moving him up.

Matt:

Your qualitative analysis is in line with what I've been thinking too.

Maybe this weekend I'll segregate my numbers for seasons in the AA before 1884 and after 1889 (on the one hand) and the middle years (on the other) and compare them to those same players' seasons in the NL to see what the numbers look like.

Jeff, I think you need to control for the league averages. IIRC the AA was a 'hitter's league' i.e. league averages were higher. I don't think a study can be valid unless you control for this. So I'd compare AVG/lgAVG or something, as opposed to just AVG.

Also, I wouldn't count ANY seasons after 1892 because in 1893 they moved the pitcher's mound back to 60' 6" and that caused wholesale changes. I'm sure it was a lot of work, and don't mean to sound too critical, but I don't think your study is valid at this point.

I will certainly be re-considering at least Harry Stovey next year, and probably moving him up.

I don't know if this is correct, but I deducted 4% off of Stovey's numbers (which is less harsh than some of the other voters here).

It's still a little too short of a career to make my ballot (though I could make room for him at the bottom), though I can see a stronger case for him solely on peak.

One thing I would suggest is to compare the player in question with the other players at the same position for each season. I think you would get a better grasp of where each player belongs on your ballot.

Joe:

Not claiming the study is spot on, but I wanted to start the discussion, which it has.

It is easy enough to run the numbers ignoring everything after 1892, and I'll do that.

Are you looking for something like OPS+ here -- that is, these players' AA OPS versus the entire AA's OPS over the same time period, and these players' NL OPS versus the entire NL's OPS over the same time period? I'll see what that produces, but I think we'll need to examine some things closely. For instance, if the group has OPS+ of 110 in the AA with a small standard deviation and an OPS+ of 105+ in the NL with a big standard deviation, does this mean the NL was stronger? Also, doing this may introduce age skewing, in that the total AA and total NL may have different mean ages than the selected group of crossover players. Finally, because there is a smattering of talent in the study, this approach isn't much better than just comparing the AA league average OPS to the NL league average OPS, which doesn't tell us anything and ignores the benefit of the analysis that we are comparing the same players' performances in both leagues.

By the way, I'm not a big OPS fan, but it's a simple place to start.

Maybe you can clarify your hitters' league comment. I don't know what you mean by "IIRC" -- maybe that the AA was a higher run scoring environment. That's interesting and may point out some limitations of OPS, because the AA OPS figures are much lower than the NL OPS figures during the years the two leagues co-existed, but the runs per game are higher in the AA. Maybe the disparity between high runs per game and low OPS is attributable to poor defense? That would make it run-friendly, but not necessarily hitter-friendly from a stats point of view.

Sorry about the double post

Question: How did the AA compare to the NL during the AA's "Peak Years" (1885-1889, inclusive).

Hypothesis: There was no statistically significant distinction between the two, as could be meaningfully measured by available statistics.

Method: Look at every hitter who played in at least 70 games in either the AA or NL between 1885 and 1889, inclusive, and then played at least 70 games in the other league in the next year (assuming that the next year was also between 1885 and 1889, inclusive). I then compared their unadjusted OPS and OPS+ in each year.

Why only 1885-1889? There was lots of player movement between 1884 and 1885 due to the UA and also between 1889 and 1890 (due to the PL and whole franchises switching leagues). These confounding factors make a direct comparison less valuable.

Raw data:

The study has only 20 data points (13 until I lowered the threshhold from 100 games to 70 games). Paul Radford and Sam Barkley each constitute 2 data points, as they both changed leagues and then changed back. There was very little movement between the AA and NL during these five years among players who could be considered at least "semi-regular." That, in itself, is relevant to any comparison.

Of these 20 data points, 12 moved from the AA to the NL, and 8 moved from the NL to the AA.

Of the 12 that moved from the AA to the NL, 8 moved between 1886 and 1887 and 4 moved between 1888 and 1889. None changed leagues between 1885 and 1886 or between 1887 and 1888.

Of the 8 that moved from the NL to the AA, 1 moved between 1885 and 1886, 3 moved between 1886 and 1887, 0 moved between 1887 and 1888, and 2 moved between 1888 and 1889.

Here are the numbers:

AA ----> NL

NAME AAYEAR OPS OPS+ NLYEAR OPS OPS+
Tom Brown 1886 .728 129 1887 .541 53
Bill Keuhne 1886 .551 73 1887 .749 112
Pop Smith 1886 .597 88 1887 .668 82
Fred Carroll1886 .783 146 1887 .882 149
Art Whitney 1886 .595 88 1887 .650 87
Sam Barkley 1886 .715 125 1887 .579 66
GeorgeMiller1886 .668 110 1887 .641 83
OttoSchomberg1886 .755 144 1887 .860 142

Cub Stricker 1888 .602 95 1889 .611 73
Ed McKean 1888 .765 147 1889 .793 123
Paul Radford 1888 .591 90 1889 .673 90
Jay Faatz 1888 .606 97 1889 .569 60

NL -------> AA

NAME NLYEAR OPS OPS+ AAYEAR OPS OPS+

Jack Manning 1885 .662 115 1886 .577 83

Paul Radford 1886 .594 77 1887 .745 111
Joe Gerhardt 1886 .479 45 1887 .557 57
Tom Poorman 1886 .659 103 1887 .699 93

Sam Barkley 1887 .579 66 1888 .571 78
Jack Farrell 1887 .582 64 1888 .562 82

Ed Daily 1888 .545 77 1889 .640 86
Joe Hornung 1889 .587 85 1889 .561 88

Conclusions:

1. Generally, players who switched leagues between 1885 and 1889, inclusive, had short, three syllable names.

2. Of the 12 players who switched from the AA to the NL, only 4 had their OPS go down, but 9 had their OPS+ go down.

3. Of the 8 players who switched from the NL to the AA, 4 had their OPS go up, and 6 had their OPS+ go up.

4. Looking only at players who were semi-regular hitters in both the AA and the NL in consecutive years between 1885 and 1889, inclusive, it is impossible to determine whether one league was consistently stronger than the other. There was simply insufficient player movement among regulars or semi-regulars to draw a conclusion.

Marc, my bad. I should have written something clearer like:

"in each of these eras the gap between the leagues has approached the magnitude of the gap between NL and the AA at its prime."

*********

I think that considering 1884 as a part of the AA's golden period is a mistake. In 1884, the AA weakened itself by the addition of 5 minor league teams (initially Washington, Indianapolis, Brooklyn, and Toledo; Richmond played out Washington's remaining schedule after they dropped out). The 5 teams had a combined 139-281 (.331) record; even if you believe that the original 8 were NL quality, you have to admit that these new teams collectively were pretty bad and provided an excellent opportunity to fatten up hitting/pitching stats. After the UA collapsed, the AA kept Brooklyn, but cut the others and small-market Columbus.

1884 Team W L WL% GB
Toledo..... TOL 46 58 .442 27.5
Brooklyn.. BRO 40 64 .385 33.5
Richmond RIC 12 30 .286 30.5
Indaplis.... IND 29 78 .271 46.0
Washngtn WAS 12 51 .190 41.0

MattB, it's a small sample. FWIW, the data indicates that the average player going from AA to NL (12 of them) declined by 18 OPS+ points (111 to 93), and the average player going from from NL to AA rose by 6 OPS+ points (79 to 85). Combined, 100 in the AA maps to 88 in the NL plus an additional 6 point penalty for switching leagues imposed no matter which way you switch (attributable to unfamiliarity/adjustment-to-new-league/whatever).

All posts have been reconstructed up to #102.

The thread is now fully restored.