Baseball for the Thinking Fan

Login | Register | Feedback

You are here > Home > Hall of Merit > Discussion
Hall of Merit
— A Look at Baseball's All-Time Best

Wednesday, January 23, 2019

2020 Hall of Merit Ballot Discussion

2020 (December 2019)—elect 4

Top 10 Returning Players
Luis Tiant (263), Todd Helton (251), Kenny Lofton (217), Andruw Jones (201), Ben Taylor (196), Jeff Kent (188), Johan Santana (176), Wally Schang (153), Sammy Sosa (148), Lance Berkman (135)

Newly eligible players

Player Name	HOFm	HOFs	WAR	WAR7	JAWS	Jpos
Derek Jeter	337	67	72.4	42.4	57.4	55
Bobby Abreu	95	54	60	41.6	50.8	56.8
Jason Giambi	108	44	50.5	42.2	46.4	54.7
Cliff Lee	72	30	43.5	39.8	41.7	61.7
Rafael Furcal	54	32	39.4	30.7	35.1	55
Eric Chavez	29	25	37.5	31.1	34.3	55.7
Josh Beckett	43	23	35.7	31.2	33.4	61.7
Brian Roberts	34	24	30.4	28.1	29.2	56.9
Alfonso Soriano	105	31	28.2	27.3	27.8	53.6
Paul Konerko	80	36	27.7	21.5	24.6	54.7
Carlos Pena	25	18	25.1	24.1	24.6	54.7
Chone Figgins	18	19	22.2	22.5	22.3	55.7
Marco Scutaro	11	19	22.1	20.9	21.5	55
Raul Ibanez	38	27	20.4	20.1	20.2	53.6
Brad Penny	23	11	19.1	21.5	20.3	61.7
Jason Bartlett	15	5	18.3	19.6	18.9	55
Adam Dunn	75	32	17.4	17.7	17.6	53.6
Lyle Overbay	12	13	16.8	16.7	16.7	54.7
J.J. Putz	25	17	13.1	12.9	13	32.7
Jose Valverde	51	13	11.5	12	11.7	32.7
Ryan Ludwick	13	14	11.2	13.5	12.4	56.8
Alex Gonzalez	11	19	9.2	12.8	11	55
Jamey Wright	10	2	9.1	10.1	9.6	32.7
Joe Saunders	10	3	8.6	10.1	9.3	61.7
Heath Bell	31	13	7.1	8.9	8	32.7
Nate McLouth	10	12	6.4	10.2	8.3	57.8
Kyle Farnsworth	22	4	6.2	9.3	7.8	32.7

we’re alternating “elect 3” and “elect 4” years

2020, 2022, 2024, 2026, 2028, 2030, 2032, 2034, 2036-37, 2039, 2041 are elect 4

DL from MN Posted: January 23, 2019 at 01:21 PM | 184 comment(s) Login to Bookmark
  Related News:

Reader Comments and Retorts

Go to end of page

Statements posted here are those of our readers and do not represent the BaseballThinkFactory. Names are provided by the poster and are not verified. We ask that posters follow our submission policy. Please report any inappropriate comments.

Page 2 of 2 pages  < 1 2
   101. bachslunch Posted: February 06, 2019 at 07:33 AM (#5812643)
@98: I don't think there's a moderator here. And I can be as contentious on arguing a point as anyone, but I've also learned to my chagrin that sometimes it's not worth pursuing. Depends on how feisty I happen to feel, but I've got my limits. Especially when it comes to folks like Sugar Bear Blanks and his various aliases. Not saying that's kwarren's circumstance, of course. So I guess what I'm saying is: don't worry about it. Your blood pressure will thank you. :-)
   102. bachslunch Posted: February 06, 2019 at 07:40 AM (#5812644)
And why not? This is yet another area where the HoM can do better than that Hall in Cooperstown. As our knowledge and understanding increase, we should not feel beholden to past analysis; we did the best we could at the time.

If certain players' merit can now be seen as less than the standards, the voters should be able to weigh in and make that correction to the membership rolls. Any players voted out go back to the candidate pool and are eligible for election.

You could go the "kick out the worst guys" route after the fact -- but my guess is that in some cases the marginal players elected early on were among the best available folks out there. For example, any 3B elected prior to the 1950s may not look so good now compared to later options at the position, but they were reasonable choices given who was available at the time.
   103. kwarren Posted: February 06, 2019 at 10:34 AM (#5812695)
Don't know why he does it. Don't know why he picked me.
You offer opinions with no empirical evidence, that measures a players career value, peak value or merit. With regard to Abreu being better than Lofton, you basically say that Abreu's offensive superiority is more than Lofton's defensive superiority, in spite of lots of empirical evidence that says the exact opposite.

The evidence for your position is 150+ game seasons, 100 run seasons, 100 BB seasons, 100 RBI seasons, a vague reference to Abreu's base stealing being "not that much worse then Lofton's", and statements such as "And when I look at Abreu I see very good range and arm on defense".

The evidence that Lofton is better is over whelming & convincing. And your arguments to the contrary are weak, and missing the point of measuring a player's value in terms of adding wins to his team.

I've got a few positions that I know are at variance with the mainstream of thought here.
You don't like WAR, as was quite apparent in your Rivera & Abreu narratives. That seems to be the basis for your "variance with the mainstream of thought here". WAR is the best measure of a player's value that we have. It is the only realistic way we have of fairly comparing defense, offense, & pitching with each other. And for comparing relief pitching with starting pitching. Many people here have developed their own variations, but all of them seem to be based on some variation of WAR and people generally use "value based metrics" when explaining their various positions. You don't do this at all, and it's frustrating as hell to attempt to refute your arguments which you are asking people to do. To evaluate a player's value as we are trying to do here, we need to use metrics that actually measure player value.

Please show the evidence that you have that shows that Abreu's offensive superiority is greater than Lofton's defensive superiority. If you want to use the argument that the defensive metrics do not properly measure Jones' & Lofton's defense properly that's fine. If you want to give evidence that shows the metrics are actually incorrect, that would be even better. It's been attempted by some of the others posters here also. Jones and Lofton are both ranked lower here than WAR would suggest they should be, and that is generally because of some skepticism regarding their defensive prowess. Many people believe that nobody could ever be as good defensively as Willie Mays was, and therefore the defensive metrics must be doing something wrong. The baseball writers in their Hall of Fame balloting, seemingly, totally ignored the defensive contributions from these players and only looked at their offense.
   104. . . . . . . Posted: February 06, 2019 at 10:42 AM (#5812702)
Please show the evidence that you have that shows that Abreu's offensive superiority is greater than Lofton's defensive superiority.

There's no requirement he use empirical evidence, let alone empirical evidence that meets kwarren's standards. He only needs to be fair to all eras and explain his reasoning. #### off.
   105. DL from MN Posted: February 06, 2019 at 11:09 AM (#5812716)
he also doesn't strike me as a ridiculous choice

We haven't made any ridiculous choices. Most of our weakest choices are in the Hall of Fame.
   106. Carl Goetz Posted: February 06, 2019 at 11:17 AM (#5812719)
"WAR is the best measure of a player's value that we have."
I personally agree with this statement, but that doesn't make it empirical fact. It is an opinion. There are other methods of valuing players, including Win Shares, Kiko's Player won-loss records, the more traditional stats, and even subjective eye-test opinions on defense.
I also generally agree with kwarren on the relative value of Kenny Lofton vs. Bobby Abreu. Again; not a fact, but my opinion based on my preferred methods of analysis.
progrockfan is certainly entitled to his own opinions on the matter. You are certainly welcome (and encouraged) to question his line of reasoning. In fact, the whole point of requiring voters to state their reasoning is to allow others to question it respectfully. Respectfully being the key word.
For example, "With all due respect these categories that you have cherry picked for you comparative analysis seem silly and arbitrary." First off, putting "with all due respect" at the start of the sentence does not turn it into a respectful sentence. This point could easily have been made in a respectful way, such as "The categories you've selected for your comparative analysis seem somewhat arbitrary to me" followed by some reasoning as to why.
I'm not going to go through every example here as I think the last one covers it.
My point is this. As someone who agrees with you (kwarren) on the analysis in general and on these specific players, your method of argumentation is both turning others off to our arguments and making this process less enjoyable for all. I'd much rather be responding to comments about a player right now than this.

   107. progrockfan Posted: February 06, 2019 at 02:01 PM (#5812809)
You offer opinions with no empirical evidence... People generally use "value based metrics" when explaining their various positions. You don't do this at all... You don't like WAR...
You continually misquote me and twist my words. You offer bogus 'facts'. You insult other posters.

Even in insulting me you can't get your facts straight. If you had actually read my stuff, which clearly you haven't except to troll for insut bait, you'd know that I routinely use "value based metrics". But I guess it's more fun to put words in my mouth.

I've cited numerous examples of you insulting me and misquoting me. Your only rsponse has been to insult and misquote me some more.

And it's not just me you fail to respond to. You were called out twice for failing to consider older players in your ballot, yet somehow your ballot was still counted. Not sure why.

Do yourselves a favor, guys, and kick this clown out of here. His style is disruptive, he doesn't check his facts, and he loves to insult people.

Have fun with kwarren, guys. So long as he's here, I'm gone.

What a shame... It was a cool place, until he came along.
   108. Kiko Sakata Posted: February 06, 2019 at 02:11 PM (#5812816)
Have fun with kwarren, guys. So long as he's here, I'm gone.

What a shame... It was a cool place, until he came along.

progrockfan, I'm sorry you feel this way and hope you stick around. Unfortunately, this website is not really moderated. If somebody gets extra, super-offensive, the website's owner (Jim Furtado) will ban somebody, but that's extremely rare, and, frankly, it should be. You can ignore other members (as in formally have their posts not appear on the website when you're logged in), although that's not the greatest solution specifically at the Hall of Merit. But unless you're a ballot counter for HOM elections, it's an option you might want to consider.

I have enjoyed your contributions to the Hall of Merit a great deal and hope you stick around.
   109. Bleed the Freak Posted: February 06, 2019 at 06:09 PM (#5812914)
Have fun with kwarren, guys. So long as he's here, I'm gone.

What a shame... It was a cool place, until he came along.

You'll be missed tremendously progrockfan, hopefully the rest of us can keep you interested in staying here.

This reminds me of a situation at the Baseball Fever where newer folks have left the site due to being insulted or verbally abusee.

It's a shame, not many passionate and knowledgeable baseball fans that enjoy a good dialogue, our discussions allowed me to increase my knowledge base, and I am thankful for that.

Wishing you the best.
   110. Howie Menckel Posted: February 06, 2019 at 06:17 PM (#5812916)
I'm just here so I don't get fined.
   111. Dr. Chaleeko Posted: February 06, 2019 at 06:23 PM (#5812919)
Prog, please keep coming back. It's nice to have enthusiastic rookies!

I owe KCG2 an explanation, but first I want to go to the point about how many Negro Leaguers should be elected. Well, the first answer is, I dunno, and neither does the electorate. It's been a feel-as-you-go thing. But here's some facts. The Hall has elected 29 Negro Leagues players, the HOM 31 (I'm not counting any time that Doby, Jackie, Campy spent in the Negro Leagues nor any time that Brown, Paige, Banks, Aaron, Mays, etc spent in the big leagues.) That's pretty close! And one or two guys on the margins isn't all that off the beaten path. We've got a bigger imbalance in the 1800s than here, IMO.

Now, I don't consider myself a Negro Leagues expert, but I am more well informed than the average bear. But I went looking for the answer to this very question on my own recently, and I'll be posting an article on my blog on March 8th about how I arrived at an answer. I'll give everyone a hint of it in this way. Take as assumptions that:
a) The Negro Leagues era spans about 1885 to 1960
b) The Hall of Fame, despite its flaws has about the right number of white MLB guys in it for this entire period---especially if we control for Frischian errors
c) The rate of induction should be the same for the white MLBers as it is for the Negro Leaguers.

There's quite a bit of detail work that goes on in between those letters and the final answer. You'll have to read the article. But if these assumptions are reasonable, and I think they certainly are defensible, then the "right" number could be higher than what we've elected at the HOM. In fact, I suspect it's about one "ring" out from where the Hall is at (1 per position and three pitchers). I'm pretty sure that a lot of folks will vehemently disagree with my results and my math, and that's OK. Makes for rich debate. But I am working from data and not only from suppositions.

I also want to address the idea that there's latent bias in my numbers. OF COURSE THERE IS! Look, no one would go to the trouble of running up a batch of numbers if they didn't care and didn't think the induction of more Negro Leaguers was a potentially important consideration. At the very least, there's a bias toward wanting to get as close to truth as I'm capable of. And I may not be as capable as others, I'll grant you. ;)

OK, next post will get into KCG2's questions more directly.
   112. Rob_Wood Posted: February 06, 2019 at 06:41 PM (#5812924)
The HoM journey throughout baseball history will be just as important as the final destination. Lively, spirited discussion will help shape voters’ beliefs regarding the relative merits of baseball’s best players. All members are expected to be considerate of others’ opinions/arguments and be willing to consider alternative points of view. Disagreements will inevitably arise, but we will strive to maintain civility at all times.

The above is from the Statement of Purpose from the Hall of Merit's Constitution.

We have always been a very collegial group over the years. We have always welcomed new members, especially if they take the Hall of Merit "seriously" (which, of course, does not preclude anyone from having fun here). While there have been a few acerbic comments in various threads, they were quickly forgotten and everybody moved on.

I suggest we all strive to embody the Statement of Purpose. Arguments and disagreements are to be expected. They are inherent to this exercise (a case could be made that we have had too few arguments and disagreements). As we discuss and engage with others, let's everyone try to be civil.

   113. Dr. Chaleeko Posted: February 06, 2019 at 07:23 PM (#5812933)
I have to catch an early flight tomorrow morning, so I am not able to get to all the point you raised, KC. But I got to a few. I hope this doesn't come across as defensive. I want to do what's best.

But that assumes that Taylor's batting lines would translate to MLB competition *and* to seasons about twice as long as Taylor was playing.

The season-length issue was, IIRC debated a long time ago. The gist is this: Negro Leaguers probably played MORE games per season than MLBers because they barnstormed all the heck over, played league games, and then went and played winter ball somewhere. These guys were likely playing “schedules” along the lines of 250 to 300 games. Furthermore, their bodies had it rougher because they didn’t travel by rail or plane (by bus), and they didn’t have the benefit of even the lower end of white hotels. God knows what medical treatment they might have gotten. It’s been generally agreed here that the league schedules they played and the schedules against top teams are not representative of their durability.

This is another underappreciated source of attrition/regression that is not only ignored in the MLEs, but likely exacerbated by pro-rating playing time.
I’m unclear whether you’re saying proration shouldn’t happen here, so I’m going to answer as if that’s your position. It may clear up some other issues as well (actually, it’s the Negro Leagues, and their chaotic nature just opens more cans of worms all the time).

We could translate these guys to the exact number of games they played, but even that is problematic when seasons have a little different meaning and vary in scope between a handful of games and 105 games. And when also, we’re talking about guys who during the same season or calendar year jumped among teams who played differing numbers of games in the same league, between competing leagues where each team played a different number of games, and then winter leagues with more stable yet differently long schedules. Not to mention the minors and majors for later players. So we must translate the players into some common unit of measurement of what a season is. The common unit of measurement that’s most easily understood and agreed upon is the length of an MLB season for the year in question.

My choice to avoid simple proration has been to group all play during a given age of a player’s career into one “season.” I then attempt to determine over the course of their career the average number of games played per scheduled game, and I get a percentage. For seasons that require proration (most seasons), that career-long percentage becomes the proration factor. I’m trying to avoid the issues with simple proration that you are raising.

Being in the 90%-ile is easier when the talent distribution of a league is highly right skewed as the NgL distribution was in comparison with MLB. This is one of my biggest sticking points with the MLEs.
I don’t really disagree with you here, but I think this is the best solution to a bad problem. And I’m not statistically trained, so I’m limited in what tricks of the trade I know, so I’d be glad to talk through some alternatives. That said, I apply the Quality of Play debit after this step to address this very issue. Also, it appears to me that, in general, Negro Leagues regulars played more than the scrubs versus the same situation in MLB. Negro Leagues teams are known to have very small traveling rosters. Often 11 or 13 players, so the rightward skew may be both a function of league quality and also of financial realities that limited the teams to smaller traveling squads.

I'm skeptical of NgLs being that high, again given the talent level of players filling out rosters.
So here’s my operating theory on this. The cream of the Negro Leagues are Hall of Fame players. I think that’s an absolutely non-controversial comment given the performances of Larry Doby, Roy Campenella, Minnie Minoso, Monte Irvin, and Jackie Robinson among others. They played at the top of the league in the Negro Leagues, and they played at the top of the league in MLB. In fact, Satchel Paige as a 40+ year old player was a valuable swingman. Now, we aren’t electing anyone below the top levels of the Negro Leagues, so we are looking for the right end of the curve anyway.

Here’s where you can find more info on the league-quality calculations: In essence, I pulled other people’s work together, relied on some of the conversations that have taken place about league quality here at the HOM, and also used the minor-league hierarchy’s changing nature to put this together. You can find some additional theoretical thinking about all of this and more here:

   114. Mike Webber Posted: February 06, 2019 at 10:56 PM (#5813019)
Doc C/Eric I think those links are busted

MLE for Negro League Pitchers


MLE for Negro League Hitters

Maybe that is where you were going?
   115. DL from MN Posted: February 07, 2019 at 09:32 AM (#5813078)
Thanks for posting that Rob. This only lasts as long as we have a critical mass of members. If people are rude the whole thing dies.
   116. Dr. Chaleeko Posted: February 07, 2019 at 01:47 PM (#5813274)
Here’s something fun! A pal of mine is teaching a course this summer in decision-making. He is using the idea of constructing one’s own Hall as the metaphor through which to examine and apply decision making ideas. It’s a six or eight week course. What would YOUR syllabus look like? (PS: My pal is Miller from the Hall iof Miller and Eric.)
   117. Carl Goetz Posted: February 07, 2019 at 02:45 PM (#5813350)
"(PS: My pal is Miller from the Hall iof Miller and Eric.)"
Already assumed that :)
   118. Esteban Rivera Posted: February 07, 2019 at 05:02 PM (#5813464)
I’ve been looking at the positional adjustment values that Baseball-Reference uses and there are some changes in values that I find odd with the way they occur. I believe that they calculate and arrive at the positional value adjustments based on the offense of the players that play the position and by looking at the changes in fielding performance of players that move between different positions (someone please correct me if I'm wrong). I don’t have an issue with this method, because it’s a good method for establishing what the baseline adjustment should be.

Where I have some pause is in some of the changes the values of the adjustments have for the same positions across the years. I’m not sure if these changes in the values are only based on the offense and fielding performance method or if there is something else that influences the value change. But I think if a positional adjustment is used to represent how much harder or easier it is to field that position relative to the others, then there should be some sanity check based on what the position actually demands from the player and if there has been a change in that regard. From my viewpoint, once you’ve established a baseline value for the fielding position, the factors that should impact changes to the fielding position values are changes in the physical environment context (if it is something uniform that affects all teams, if not than I view it more of a park factor adjustment than a positional adjustment), to the rules that impact the context the position is being played in, and/or changes in the responsibilities you would expect from the position.

One other thing I’ve noticed about the positional adjustments is that the values of all eight fielding positions (leaving P and DH out of this discussion) is that the sum of all their values usually is 0. So, it appears the values tend to operate on the supposition that an increase in in difficulty for one position means another becomes easier. This is not always the case, as there are several years where the sum ranges from 1.5 to 1.5. Here are the years where the sum does not equal 0:

1875-1876 = -0.5
1879 = 0.5
1880-1882 = 1
1883-1892 = 1.5
1893 = 1
1894-1895 = 1.5
1896-1897 = 1
1898 = 0.5
1900-1902 = 1
1917 = -1
1918-1919 = -0.5
1922 = 0.5
1932 = -0.5
1951-1952 = -1
1954-1955 = -1
1956-1957 = -0.5
1961-1968 = -0.5
1972-1976 = 0.5
1979-1981 = -0.5
1982 = -1
1983-1986 = -1.5
1987-1991 = -1
1992 = -0.5
1995-1997 = 0.5
2000 = 0.5

Also, here’s a summary of how the values in runs for each position change (per 1350 innings played according to the B-R explanation page and up to 2017 since 2018 has not yet been included in the table):

1871-1916 = 10
1917 = 9
1918 = 8.5
1919 = 7.5
1920 = 7
1921 = 6.5
1922 = 5.5
1923-1952 = 5
1953 = 5.5
1954 = 6
1955 = 6.5
1956 = 7
1957 = 7.5
1958 = 8
1959-1968 = 8.5
1969-1982 = 9
1983-2002 = 8.5
2003-2017 = 9

First Base
1871-1896 = 0
1897 = -0.5
1898 = -1.5
1899 = -2
1900 = -2.5
1901 = -3.5
1902 = -4
1903-1916 = -5
1917-1918 = -5.5
1919-1920 = -6
1921-1922 = -6.5
1923-1953 = -7
1954-1955 = -7.5
1956 = -8
1957-1958 = -8.5
1959-1976 = -9
1977-2017 = -9.5

Second Base
1871-1874 = 3
1875-1879 = 2.5
1880-1896 = 3
1897 = 2.5
1898 = 2
1899-1900 = 1.5
1901 = 1
1902 = 0.5
1903-1916 = 0
1917 = 0.5
1918 = 1.5
1919 = 2
1920 = 3
1921 = 3.5
1922 = 4.5
1923-1936 = 5
1937-1938 = 5.5
1939-1940 = 6
1941-1942 = 6.5
1943-1950 = 7
1951-1953 = 6.5
1954 = 6
1955-1956 = 5.5
1957-1958 = 5
1959-1960 = 4.5
1961-1988 = 4
1989 = 3.5
1990-2017 = 3

Third Base
1871-1926 = 5
1927-1928 = 4.5
1929-1930 = 4
1931 = 3.5
1932-1936 = 3
1937 = 2.5
1938 = 2
1939-1940 = 1.5
1941 = 1
1942 = 0.5
1943-1952 = 0
1953-1955 = 0.5
1956 = 1
1957 = 1.5
1958 = 2
1959 = 2.5
1960-1978 = 3
1979 = 2.5
1980-1981 = 2
1982 = 1.5
1983-1994 = 1
1995-1999 = 1.5
2000-2017 = 2

1871-1950 = 10
1951-1953 = 9.5
1954-1980 = 9
1981-1997 = 8.5
1998-2000 = 8
2001-2002 = 7.5
2003-2017 = 7

Left Field
1871-1878 = -10
1879-1892 = -9.5
1893 = -10
1894-1895 = -9.5
1896 = -10
1897 = -9.5
1898-1899 = -9
1900 = -8.5
1901-1917 = -8
1918-1920 = -7.5
1921-1957 = -7
1958-1959 = -7.5
1960-1971 = -8
1972-1979 = -7.5
1980-2017 = -7

Center Field
1871-1896 = -8
1897 = -7.5
1898 = -7
1899 = -6.5
1900 = -6
1901 = -5
1902 = -4.5
1903-1926 = -4
1927-1928 = -3.5
1929-1930 = -3
1931-1932 = -2.5
1933-1937 = -2
1938-1941 = -1.5
1942-1980 = -1
1981-1982 = -0.5
1983-1986 = 0
1987-1988 = 0.5
1989 = 1
1990-1991 = 1.5
1992 = 2
1993-2017 = 2.5

Right Field
1871-1876 = -10
1877-1882 = -9.5
1883-1899 = -9
1900-1901 = -8.5
1902-1917 = -8
1918-1921 = -7.5
1922-2017 = -7

I'll comment later on some of the questions and thoughts I have about the change in values, but I wanted to share this in case anyone wanted to comment as well.
   119. Carl Goetz Posted: February 07, 2019 at 09:50 PM (#5813536)
"I believe that they calculate and arrive at the positional value adjustments based on the offense of the players that play the position and by looking at the changes in fielding performance of players that move between different positions (someone please correct me if I'm wrong)."
My understanding of Baseball-reference's methodology is that its all by fielding for multi-position players. They don't really give an in-depth treatment of the methodology but they don't mention using offensive numbers in the process.
   120. Esteban Rivera Posted: February 08, 2019 at 09:18 AM (#5813602)
Thanks Carl! I went back to the page where the positional adjustment values were listed ( and checked if they had any explanation there. You’re right that it is vague on how exactly it’s done. The initial paragraphs in the Rpos, Positional Adjustment Runs section are what led me to assume that it’s a combination of the offense level of the position players and the fielding performance among those that switch:

“If you take a quick look at the batting performance by defensive position, you'll quickly see that teams are willing to sacrifice offense at "defensive" positions (stats are prorated to 650 plate appearances).

‘Table with offensive numbers by position is listed’

When one quantifies these differences and also looks at the changes in fielding performance when players move to different positions, we can estimate the average differences between positions.”
   121. kcgard2 Posted: February 08, 2019 at 06:57 PM (#5813837)
@Dr Chaleeko

Thank you for the response. I was unaware of previous debates about the season-length issue, but I'm glad you brought them up here. If those estimates of barnstorming, etc. games played are even semi-accurate, then I would retract my complaint about season length. That said, did the top NgL players participate to that degree in the lower level competitions? Not a rhetorical question, I'm sure you are better equipped to answer this question than most.

About pro-rating, my complaint is definitely NOT that pro-rating is done, I agree that pro-rating should be done. In fact it needs to be done to be fair to NgL players. My complaint was pro-rating to MLB length schedules without adjusting performance level, given that playing time is typically doubling (or more) from the high-level NgL playing record. Your comments about barnstorming, which I hadn't considered before, alleviate this concern that I had to some degree. So I am glad this dialogue is happening. Still, a full pro-ration of playing time with no adjustment to performance seems strong to me. The variance factor that you mention for length-of-schedule adjustment could be a/the factor to alleviate this concern if applied correctly. I am very interested to get to the bottom of that factor in particular, so I await you having the opportunity to check into it.

So we must translate the players into some common unit of measurement of what a season is. The common unit of measurement that’s most easily understood and agreed upon is the length of an MLB season for the year in question.

Agreed, I have absolutely no problem with that approach.

Being in the 90%-ile is easier when the talent distribution of a league is highly right skewed as the NgL distribution was in comparison with MLB. This is one of my biggest sticking points with the MLEs.
I don’t really disagree with you here, but I think this is the best solution to a bad problem. And I’m not statistically trained, so I’m limited in what tricks of the trade I know, so I’d be glad to talk through some alternatives. That said, I apply the Quality of Play debit after this step to address this very issue. Also, it appears to me that, in general, Negro Leagues regulars played more than the scrubs versus the same situation in MLB. Negro Leagues teams are known to have very small traveling rosters. Often 11 or 13 players, so the rightward skew may be both a function of league quality and also of financial realities that limited the teams to smaller traveling squads.

I appreciate your honesty. Trust me that I have a full appreciation of the difficulty of the task you're trying to accomplish (I'm a professional statistician). I've wrestled with alternative answers to the issues you're trying to confront, but have admitted that the nature of the problem makes preferable methods unfeasible (at least the ones I've been able to come up with). I will continue to think about this particular issue because I think it's one of the bigger ones for the MLEs. Yes, quality of play adjustments will regress NgL performances downward, but that is a necessary step in addition to finding out how NgL percentile performance would translate to the MLB distribution of talent.

If NgL regulars took a disproportionate amount of playing time, then perhaps that alleviates the rightward skew issue somewhat, but it reintroduces the representation issue in a stronger form. I.e, that NgL players start to look less like 4-5% of the eligible pool and more like 2-3% if half the players were only part time rostered. But let's let that rest for now.

So here’s my operating theory on this. The cream of the Negro Leagues are Hall of Fame players. I think that’s an absolutely non-controversial comment given the performances of Larry Doby, Roy Campenella, Minnie Minoso, Monte Irvin, and Jackie Robinson among others. They played at the top of the league in the Negro Leagues, and they played at the top of the league in MLB. In fact, Satchel Paige as a 40+ year old player was a valuable swingman. Now, we aren’t electing anyone below the top levels of the Negro Leagues, so we are looking for the right end of the curve anyway.

100% agreed, and have made the exact same point myself in previous HOM discussions. But it is still a relevant question in terms of how to calculate "quality of play" or league strength compared to MLB, because even though we're only interested in the extreme right tail of NgL talent, they still accumulated their playing records against the full population of NgL talent.

I don't think anything you wrote came across as defensive, I think all the points you raised were entirely fair, and some I hadn't considered before. Actually, if I had put as much work into something as you have into the MLEs, I'd probably be crazy defensive about them :) I'm glad others are finding value in this discussion as well.
   122. kcgard2 Posted: February 08, 2019 at 10:12 PM (#5813863)
I'm not able to access the URLs you provided, even if I remove the action=edit suffix, which gives the following:

You are logged in as [name] and do not have the necessary privileges to access the dashboard for "the Hall of Miller and Eric".

I'm certainly interested in the methodology, so if that's fixable, I'll give it another try.
   123. Esteban Rivera Posted: February 08, 2019 at 11:19 PM (#5813878)
After looking over the positional adjustment values used by Baseball-Reference, here are some observations and questions I have about the changes in values. I'll go by position, first doing a quick summary of the changes throughout the years and then making some comments about the values.


Catcher starts with a value of 10, the same as shortstop. The value starts going down at the tail end of the Deadball era (starting in 1917) until it reaches a 5 in 1923. It remains there for the next three decades until gradually increasing throughout the 50s. The value goes up to 8.5 in 1959 and remains there until expansion in 1969 where it goes up to 9. It goes down to 8.5 in 1983 and remains there for the next two decades. Then in 2003 it goes back to 9 where it remains like that for the present day.

A couple of thoughts about the catcher positional adjustment value. In my opinion, catcher should always be the position with the highest positional adjustment relative to the other positions. It is the one that demands the most from the person playing it (compared to the rest of the position players) and is the most physically taxing. So, why the dip in value starting at the tail end of the Deadball era that lasts from there to until the end of the 1950s (placing it below shortstop and second base at that point in time)? And why the increase from that point forward when the responsibilities and demands of the catcher position haven’t really changed since the start of the Liveball era. And I know the first response to this would be the downturn in stealing during those years, but I don’t really see that as a good argument for adjusting the positional value downward. Because even though there is a downturn in terms of stealing, that responsibility is never eliminated from the catcher position. Less opportunities to have to thwart stealing attempts does not mean it’s still not the catchers responsibility when it does happen. I would think that the lower number of stealing attempts would lead to less fielding value for the catchers in that context but should not result in a lower positional adjustment. Although, you could also argue that the decrease in steal attempts does put more responsibility on other positions to manage the runners on base and increase the value of turning a double play.

I also find it ‘off’ that the decrease in the adjustment starts in 1917, still in the Deadball era and the era of damaged, discolored and soiled baseballs still being in play. In my opinion, this decrease shouldn’t be considered to begin until after 1920.

Finally, I consider that the initial value of 10 may be too low of an adjustment for catchers (particularly if that is where you are starting shortstops at). If an offense only position like DH has a -15 adjustment, why shouldn’t a position as catcher, where anything you get on offense is almost gravy in terms of who ends up manning the position, be at +15. Or at least the Deadball era catchers should have a much higher adjustment than the present-day catchers.

First Base

For first base, the value throughout most of the 19th century is 0. At the turn of the century there is a shift in the adjustment, starting in 1897 with-0.5 and dropping to -5 by 1903. It stays there until 1917, where it again decreases in value throughout the tail end of the Deadball era and the start of the Liveball era (ending at -7 starting in 1923). It stays at -7 for three decades, until it starts again deceasing starting in 1954 and stabilizing by 1959 at -9. It remains at -9 until the expansion year of 1977, where it drops to -9.5 and remains there until the present day.

It’s interesting that playing first base has no penalty throughout most of the 19th century, although catching balls without gloves and all of the running after balls in play in this era would contribute to this. The one thing I’d say likely needs adjusting for the 19th century
is the value for 1871-1876 due to fair-foul hits being legal during that time (which must be quite the workout for any of the players at first base, third base and catcher).

It’s a steep change from 1897 to 1903, but don’t really have any comment as to if the change should be earlier or if that change in value is too much, too little or is just right.

What I will comment on is the value being a -7 (the same value as the corner outfield positions) throughout the first three decades of the live ball era and then during a span of six years decreasing to a -9. Is there a change in that span in the demands of the position that can explain this? And what about the last adjustment that occurs in 1977, what can explain the need for that?

Second Base

At second base the value starts at 3 in 1871, drops to 2.5 in 1875 only to go back to a 3 in 1880. From there it remains at 3 for the next decade and a half, then decreasing at the turn of the century starting in 1897 (similar to first base). It ends up stabilizing at 0 in 1903 on through 1916. Then (again like first base) it shifts around the tail end of the Deadball era into the transition to Live Ball, ending up at a value of 5 in 1923. It remains at 5 until 1936, where it then increases sharply stating in 1937 and settles at a value of 7 by 1943. It remains at 7 through the end of the 40s and the begins to decrease in value throughout the 1950’s, settling at a 4 by 1961. The value of 4 remains from there until 1988, where it decreases to 3.5 in 1989 and then to 3 in 1990, remaining at a value of 3 until the present day.

At second base there’s that blip for 1875-1879 for some reason. Then there’s the decrease to 0 for most of the Deadball era. I’m guessing this is tied to the lower incidence of double plays and no need for the pivot, but does a 0 value really make sense? Particularly since you still have a lot of a lot of runner movement on the base paths

Then there’s the dramatic increase once the Liveball era kicks in. Which value makes more sense for the position once the demands and requirements have stabilized? Is it the 5 or 7 or somewhere in between? You could say the increase in the 40s could be explained by the responsibility of runner advancement prevention falling more on the right side of the infield due to lower base stealing attempts. This also could mean that the value of turning double plays on defense is much higher in that context. But if that’s the case, the shouldn’t the value not drop during the 50s? And what of the later drops in the 80s and 90s, do these make sense based on the demands of the position?

Third Base

At third base we start with a value of 5 from 1871-1926. Then it slowly transitions downward from 1927 to 1932 where it stops at a value of 3. This value only lasts for five years before slowly decreasing again from 1937 to 1943, where the value is now 0. It remains at 0 for ten years, then begins a slow increase in 1953, ending at 3 by 1960. The next two decades sees a value of 3, then a decrease starting in 1979 that sees the value become 1 by 1983. It stays that way through 1994, then increases to 1.5 for the latter half of the 90s before increasing to 2 in 200, where it stays to the present day.

The change in values for third base are vexing to me. First no change from 1871 to 1926. I would think that 1871-11876 should be adjusted for the fair-foul role likely resulting in more running and chasing of batted balls for the position.

I get the transitional period post Deadball era, but why the decrease all the way to 0 in the 40s and then subsequent increase? Is there any change in the position’s demands and responsibilities that makes sense with this? And why the higher positional adjustment for the 60s and 70s third basemen? Again, what obvious change to the demands of the position that would require this? And again, it decreases for the 780s and 90s third basemen but I don’t see an obvious difference to justify this.


Shortstop begins at a value of 10 and it remains as such for a stretch of eighty years. It decreases to 9.5 from 1951-1953 and then goes to 9 from 1954 all the way to 1980. Starting in 1981 the value drops to 8.5 and remains that way through the 80s and most of the 90s. In 1997 it drops to 8, then to 7.5 in 2001 before settling at 7 in 2003 and staying at that value in the present day.

The responsibilities of shortstop really haven’t changed much despite the different contexts of the 19th century, Deadball and Liveball eras. So why the dips in the mid-50s, early 80s, and turn of the 21st century? Or, is it more of a case that the initial value of 10 may be slightly high?

Left Field

Left field starts at a -10 in 1871, before increasing to a -9.5 in 1879. Starting in 1893 it ping-pongs back and forth for a few years between -10 and -9.5, before starting an increasing trend in 1898 that sees it settle at -8 in 1901. The value remains at -8 through 1917, then increases to -7.5 for a three year stretch before going to -7 in 1921. It remains at that value for the almost four decades before decreasing again, first to -7.5 in 1958, then to -8 in 1961, before rebounding back to -7.5 in 1972. It final returns to -7 in 1980 where it remains to the present day.

The only comment I have here is why the decrease for the 60s and 70s when right field experiences no changes.

Center Field

Center field has a value of -8 throughout most of the 19th century. It begins steadily increasing in 1897 and settles at -4 by 1903. It stays at -4 through 1926, then it begins a slow decrease from 1928 to 1941, where it settles at -1 by 1942. It stays that way for four decades then slowly transitions throughout the 80s and before expansion in 1993 to an increasing value adjustment (-.05 in 1981, 0 in 1983, 0.5in 1987, 1 in 1989, 1.5 in 1990, and 2 in 1992), settling at 2.5 in 1993 (staying there until the present day).

Why is it deemed harder to play center field now in the sillyball and the launch angle/three true outcomes era than in in the other eras? The centerfielder has virtually the same responsibilities and ground to cover since the beginning of the Liveball era and you would think that modern technology (such as improvements in gloves and shielding from the sun’s glare) should make it easier to handle. I have a hard time believing that the difference in the centerfielder’s responsibilities is such that it’s basically a six-and-a-half-point swing from the mid-20s to the present day.

Right Field

In right field, the value goes from -10 in 1871 to -9.5 in 1877, then -9 in 1883. By 1900 it becomes -8.5 increasing two years alter to -8 where it remains that way through 1917. Then four years at -7.5 before settling in 1922 a t -7 where it remains unchanged to the present day.

Don’t have any comments with the values for right field.

Designated Hitter

And just to mention the DH, its adjustment is -15. There is some thinking that that may be too harsh of an adjustment and it may not be considering the DH penalty. I believe there was an article on Fangraphs addressing this topic.

Pitchers are not discussed at this moment since their fielding is already included in their pitching value and the adjustment made is on the batting side (although that may be something to go over since the highest adjustment values are made in recent times with a full-time DH in one league and the NL having one during interleague play).
   124. kcgard2 Posted: February 09, 2019 at 05:25 PM (#5813991)
Perhaps 60s and 70s were rife with bunt attempts, causing the uptick in 3B positional adjustment for that era? As for SS, I would guess that the constant trend of fewer BIP has allowed teams to gradually play for slightly more offense at the position as time has gone on. Also, keep in mind that positional adjustments are created by comparing offensive output. Perhaps SS has seen a downward adjustment because teams simply no longer believe in defense-at-all-costs at SS anymore. If that was conventional wisdom in MLB for many decades, it would result in higher position adjustment because teams were universally playing all-field no-hit types there. Simply becoming "smarter" about the wisdom of that can result in the positional adjustment moving downward, because SS as a position will not be so depressed offensively compared to other positions anymore. Aggregate offensive output is used as a proxy for the defensive demands of the position.
   125. Mike Webber Posted: February 10, 2019 at 11:37 AM (#5814082)
@123 - Esteban

So are these adjustments that BB ref makes intuitive guesses or are they based on a formula?
For instance
Right Field
1871-1876 = -10

Were RF on the average 10 runs better offensively than the league, so the adjustment is 10 runs or is someone just making an educated guess what the adjustment should be.
   126. Bleed the Freak Posted: February 10, 2019 at 05:02 PM (#5814161)
Interested to get the electorates feedback, should we be electing closer to a 57% position player, 43% pitcher split.

Or is it more, as Doc has pointed out, relief pitching is quite valuable, but not many individual relief pitchers are:
Hitter and Pitcher WAR split
   127. Dr. Chaleeko Posted: February 10, 2019 at 05:08 PM (#5814166)
KCgard2, more commentary. Apparently, I had some trouble with linking last time....

Key quality of play link.

Runs Created
I looked at a more recently created MLE to be sure I was using the right formulas. Here's what I'm doing in Excel language

A = On-base term
B = Advancement term
C = Opportunity term

A = H + BB + HPB - SF - IBB
B = (1.125 * Singles) + (1.69 * 2B) + (3.02 * 3B) + (3.37 * HR) + (0.29 * (BB - IBB + HPB)) + (0.492 * SF) - (0.04 * K if known)
C = PA

Formula per James is

RC = { [ ( ( ( 2.4 * C ) + A ) * ( ( 3 * PA ) + B ) ] / ( 9 * PA ) } - (0.9 * PA)

Therefore: RC/PA = ( RC / PA )

Schedule Adjustment
But there's a comment about length of schedule accounting for 1/3 of the variation in Charleston's league.
At one time I did a bunch of work about this, and I can't locate it. Ugh. I may need to go back and recreate it. Basically I plotted a STDEV rating against either schedule length or year (which is something of a proxy for schedule length), and what I got back suggested that the schedule explained about a third of the variance.

An adjustment for this league variation should cause further regression of Charleston's stats, but their ultimate results show that they used this variance to make the league quality adjustment even weaker, from a league that's 85% MLB quality to a league that's 90% MLB quality.
Here's how I reasoned this out. I figured that if a third of the variance was explained by schedule, and I'd be adjusting separately for the schedule, I should knock out a third of the difference between the league's QoP and MLB so that I didn't double-dip the player.

By the way I tried making the QoP adjustment before the Z-score adjustment at one point, and that created MLEs where the best players topped out at 70 WAR. We know that's inaccurate because the play of dark-skinned players who debuted within ten years of Jackie Robinson (Aaron, Mays, Clemente, Frank Robinson) indicates that blacks had the same capacities and limitations that whites had. So I recognized that adjusting first for QoP was incorrect. It turns out you have to maintain the player's relationship to the league before QoP adjustments, or you end up hitting him with a double whammy.

Instead of regressing the player against his league, I'm regressing him against his own durability levels and against nearby performance. I'm not literally doing regression. I'm not that statistically able, but it's a similar idea that I am easily capable of executing. I've already mentioned how I use the player's career durability. For his in-season performance, I use a five-year rolling average calculated this way:
n = season in question
(.05 * n-2) + (.15 * n-1) + (.6 * n) + (.15 * n+1) + (.05 * n+2)

These weights are kinda like how some of the Marcel-type calculations are done except that because the future seasons are known, I split the weighting for prior/future seasons. Also, Marcels regress against the league in order to create a prediction. I'm trying to create a regressed-like estimate using the real data that we know about the player. That real data is put into an above/below average form as well, so there is an element of the league to it.

I chose to use this rolling method because it would give a little more peakiness to the peaks and create a nice, gentle slope for early and late careers. If I'd used the career average in some fashion, it probably would be more defensible, but when I tried it, it really ground down the ends...too much in my estimation.

BTW: For the first, second, penultimate, and ultimate seasons, I use the next/prior two seasons instead of the ones that I can't use. For example in a player's rookie year:

(.1 * n+2) + (.3 * n+1) + (.6 * n)

Generally, re results
So here's what I got for results both in WAR terms and in terms of my own JAWS-like concoction, CHEWS+ (CHalek's Equivalent WAR System, indexed). This could get lengthy, so bear with me.
-30 hitters above my in/out line among Negro Leaguers. My article on the eighth will suggest that this is a very reasonable number based on what we know about induction rates for white players.
-10 pitchers above my in/out line. My article suggests that 10 to 12 is an apt number for pitchers as well.

Now, I will swear up and down on any book you want that I didn't cook my results to reach those marks. I'm glad they line up, but I wrote the article a long time after I did the MLEs, and I'd never actually bothered to check this out until today.

Unfortunately I don't have time this evening to provide a more detailed look at representation in the 1885-1960 era I've described as the Negro Leagues epoch, maybe next time if it's useful to others.
   128. Esteban Rivera Posted: February 10, 2019 at 06:25 PM (#5814182)
Perhaps 60s and 70s were rife with bunt attempts, causing the uptick in 3B positional adjustment for that era? As for SS, I would guess that the constant trend of fewer BIP has allowed teams to gradually play for slightly more offense at the position as time has gone on. Also, keep in mind that positional adjustments are created by comparing offensive output. Perhaps SS has seen a downward adjustment because teams simply no longer believe in defense-at-all-costs at SS anymore. If that was conventional wisdom in MLB for many decades, it would result in higher position adjustment because teams were universally playing all-field no-hit types there. Simply becoming "smarter" about the wisdom of that can result in the positional adjustment moving downward, because SS as a position will not be so depressed offensively compared to other positions anymore. Aggregate offensive output is used as a proxy for the defensive demands of the position.

That may be a possibility for the increase in third base value in the 60s and 70s, but should that result int hat much of an increase in the positional adjustment? And is there a way to confirm that there is a higher incidence of bunt attempts during this time? The responsibility for fielding bunts on the left side of the infield has always been a part of the responsibilities of a third basemen (as well as the pitcher and catcher depending on how good or bad the bunt is), what changes in this scenario is the frequency with which they occur. So is it a case where the increase in bunting should be reflected in the positional adjustment in the fielding due to the higher chances that result?

I've thought about that same line of reasoning for the decrease in the shortstop value, it's one that makes sense. But there are a couple of questions that the reasoning brings to mind. If the context is one of less BIP, the same should apply for other positions as well. The way the values are constructed, they tend to sum up to 0 and shifts in one position that make it 'easier' mean that another one shifts to being 'harder' to maintain the sum of 0. But the BIP context should apply to all positions, so you would expect something that affects all positions to maintain the relative difference in value between the positions because they all experience the same contextual issue. However, you can say that value can change because shortstop now is is 'easier' than shortstop in the past, but if that is the case, then why tie all values together by having them sum to 0? I think you either tie the values relative to the other positions if it must sum to 0 or you do it comparing the change across time for the position but in this case you can't keep in tying it to a sum of 0 (you'd then be forcing adjustments to positions that haven't really changed just to keep the sum balanced). And a last thought, there are limits to how much offense serves as a proxy for defense, at some point some of that may just be improvement in player quality at the middle and bottom of the player talent pool.

@123 - Esteban

So are these adjustments that BB ref makes intuitive guesses or are they based on a formula?
For instance
Right Field
1871-1876 = -10

Were RF on the average 10 runs better offensively than the league, so the adjustment is 10 runs or is someone just making an educated guess what the adjustment should be.

I don't really know what the formula they use is, but the explanation on the page indicates that it's likely a combination of offensive performance by those that play the position and the fielding performance of the players that switch positions. But if I had to guess, I think that the RF value for 1871-1876 is based on just matching it to LF. RF during this time tended to be used more as spot where you would have your backup pitcher and also rotate players you wanted to rest from more demanding positions. I doubt the offensive performance is such as to be worth that much of a penalty. The other issue is the insistence of the values having to sum to 1. It's not something that is always kept to (for example the values don't add to 0 in the 1880s and 1980s), but having that requirement means that you'll have to adjust some positions more negatively in order to maintain the balance.

I guess the question I have here is what is one looking for with the positional adjustment value, is it the relative value of the positions among themselves to the value of the position across the years? I don't think you can answer both by forcing the values to add to 0.
   129. progrockfan Posted: February 11, 2019 at 09:57 AM (#5814261)
My sincere thanks to all who asked me to return, both in the forum and in private messages.

Many thanks also to those who have called for civility in these forums.

Had I known I could simply block kwarren’s stuff (which I've now done), I would’ve done so at least a month ago, and none of this nonsense would have transpired.

Allow me one parting shot on the imbroglio – a quote from the father of “value-based metrics”, Bill James:

The bad statistical analysis will assume that what the statistical record tells him must be true and complete – and by making that assumption, will forfeit his ability to add anything significant to the record.

<takes a deep breath>

Okay... back to baseball. :)
   130. Kiko Sakata Posted: February 11, 2019 at 02:38 PM (#5814371)
is it the relative value of the positions among themselves to the value of the position across the years? I don't think you can answer both by forcing the values to add to 0.

I think positional adjustments are means of comparing across positions within a given year. It's not, "How do second basemen in 1920 compare to second basemen in 1980?", it's "How do second basemen in 1920 compare to third basemen in 1920; and how do second basemen in 1980 compare to third basemen in 1980?" And, as such, I think the average positional adjustment has to be 0 every year (granting that due to rounding error, it may average out to 0.05 or -0.07 or something). Otherwise, you're going to have seasons where the total wins above average (WAA) for the season as a whole may be positive - i.e., the average player that year was above average - or negative - i.e,. the average player that year was below average - neither of which make sense to me - unless you're doing some sort of timelining, but, if you're doing some kind of timelining, I feel like you would want to make that very explicit and do it in a distinct step separate from everything else.
   131. Kiko Sakata Posted: February 13, 2019 at 11:56 AM (#5814899)
Okay, I have completed a project that I hinted at a couple of times and which segues perfectly from the conversation started by Esteban Rivera in #118 here. I have modified my website so that you can choose your own positional averages in evaluating my Player won-lost records. Not literally, typing in a different number for every position for every year, but there are four options: 0.500, 1-year avgs, 9-year avgs, and long-run avgs, and you can do a weighted average of any combination of those. I also allow three options for pitching: 0.500, starters and relievers different based on comparing all starters to all relievers, and starters and relievers different based at looking only at pitchers who did both. A very brief explanation is here. Options for picking your weights should be near the top of any page which displays or uses positional averages.

I think I've updated everything except for some of the leader pages and I tried to set everything up so that once you choose a set of weights, they'll transfer with you between pages (i.e., if you pick a set of weights for a player, those weights will stay in effect if you click on one of his teams or if you compare him to another player; the weights will revert back to normal if you go back to my homepage). Please let me know if you see any problems, though; it was a lot to update and it's possible that I missed something.

I've also written a 50-page essay(*) that looks at positional averages and talks about some other factors in using my Player won-lost records to compare players. This is a PDF (I thought that might be easier to download and read off-line; also, it has a bunch of graphs and I'm not sure how best to do graphs in HTML) and can be found here.

(*) - I don't know if "essay" is the right word. It's shorter than a book but longer than an article. The term "treatise" seemed pretentious. In terms of length/content, it's basically a chapter of a book, but that seems like a weird thing to call it since it's not part of a larger book (yet).

Finally, here's a special version of my "Uber-Stats" page that can be used for Hall-of-Merit voting - the main advantage here is that it excludes anybody who's ineligible (either because they're already in the Hall of Merit or because they haven't been retired long enough). I would suggest reading the 50-page PDF to better understand how you might want to use this, but feel free to just type in some numbers and see what pops out.
   132. Bleed the Freak Posted: February 13, 2019 at 12:36 PM (#5814924)
Thanks for sharing Kiko, I will try to take a dive into this soon, your continued efforts are greatly appreciated on my end.
   133. cookiedabookie Posted: February 13, 2019 at 01:08 PM (#5814950)
So I was looking into giving credit to players mentioned by others, for PCL/minors, War, blackball, etc.

When I do this, Bob Johnson jumps up to the top of my rankings. I'm wondering if I'm giving too much credit. I'm basically looking at what they did before and/or after the time I'm giving credit for, and giving similar credit based on rough aging curves. Curious how others are doing it...
   134. Esteban Rivera Posted: February 13, 2019 at 02:13 PM (#5815006)
Looking forward to reading that PDF Kiko. And I'll definitely be checking out the link where you can play around with the weights. I do agree with you in that if the purpose of the positional adjustment is the difficulty relative to other positions within a year then it should sum to 0 (or within rounding distance to 0). That still leaves what the 'correct' value should be for each position for that given year, but there's no easy or definitively right answer here.
   135. DL from MN Posted: February 13, 2019 at 02:34 PM (#5815028)
I am one of Bob Johnson's biggest supporters here. I'm only giving him credit for 1932 at the level of 1933. The Athletics purchased his contract after the 1930 season so it is difficult to give him credit before that. Plenty of good players were forced out in 1946 when everyone came back from the war. If I gave him credit for 1931 he would move up to 5th on my ballot. After looking at this in more detail I think he deserves some more credit. There is evidence he was held back not just because of Al Simmons blocking him but also because of his Cherokee heritage.

SABR bio
   136. Esteban Rivera Posted: February 13, 2019 at 02:53 PM (#5815037)
Interested to get the electorates feedback, should we be electing closer to a 57% position player, 43% pitcher split.

Or is it more, as Doc has pointed out, relief pitching is quite valuable, but not many individual relief pitchers are:
Hitter and Pitcher WAR split

I wanted to comment on Bleed's post #126. From my view I don't necessarily think it should be a 57/43 split between position players and pitchers per se, or at least in the strict quota sense. Setting aside that the different metrics use different splits* for the value they give between position players and pitchers, there's also the case of how many pitchers the innings are divided among. I agree with Dr. Chaleeko in that regard (and paraphrasing him a bit), pitching is valuable, but not necessarily pitchers for our purposes if the innings are divided among too many arms. So while the value for pitching may be 43% (or 41% or whatever other value gets thrown about), the number of pitchers that are viable candidates will depend on how fractured the distribution of those innings are (and the quality of them).

As an aside, does anyone know what the split for position players in terms of offense and fielding for WAR?

*For reference, they are as follows (and by all means please correct me if I'm presenting erroneous splits for any of them):

Fangraphs WAR is 57% Position player / 43% Pitchers

Baseball-Reference WAR is 59% Position player / 41% Pitchers (I believe Baseball Gauge uses the same split)

Win shares is around 64% Position players / 36%% Pitchers (although it is presented in the book as 48% Offense and 52% Defense with around 35-36% going to pitchers and other fielders getting 16-17%)
   137. Esteban Rivera Posted: February 13, 2019 at 03:04 PM (#5815039)
So I was looking into giving credit to players mentioned by others, for PCL/minors, War, blackball, etc.

When I do this, Bob Johnson jumps up to the top of my rankings. I'm wondering if I'm giving too much credit. I'm basically looking at what they did before and/or after the time I'm giving credit for, and giving similar credit based on rough aging curves. Curious how others are doing it...

In my case, and it may be a bit lazy and I may adjust this for the coming ballot, for Bob Johnson I basically take the approach that the PCL credit cancels out any further discount of the WW2 years that Baseball-Reference already bakes into their WAR values. So I just use the numbers as is. May not be the best approach, but that's how I've been handling Bob Johnson's case.

Incidentally, I wanted to ask the electorate if they apply further discounts to the ones already included in the WAR values. Are those discounts in addition to the ones already baked in to the WAR calculation, or is it a case of not being aware of the quality adjustments baked in and applying an accidental double discount? And, do you think the discounts being applied already in the WAR value calculation are appropriate?
   138. Kiko Sakata Posted: February 13, 2019 at 03:23 PM (#5815046)
As an aside, does anyone know what the split for position players in terms of offense and fielding for WAR?

This is hard to answer because WAR is built up from WAA and offensive and fielding WAA are both centered at zero. So, literally, if you multiplied everybody's fielding number (Rfield) by two (or six or ten or 0.02), the total WAR for position players would stay the same. For individual players, the numbers would change a lot - Gary Sheffield and Manny Ramirez and Derek Jeter would look much worse if you multiplied their Rfield by two; Ozzie Smith, Brooks Robinson, and Andruw Jones would look much better. Bill James talked about this a fair bit in his Win Shares book (with respect to Pete Palmer's linear weights, but it's the same concept) - there's no internal check on the reasonableness of the fielding numbers if you center on zero (I'd quote James, but (a) I'm at work right now, and (b) I'm moving in a few months and I think I already packed up that book and it's probably in storage).

I think the way you would measure the implicit split of offense and fielding is by comparing the standard deviations of the two distributions - what's the spread of offense values versus what's the spread of fielding values. And I think that if you do this, you find that fielding is being over-weighted in WAA-centered WAR systems (which is all three you mentioned). Specifically, if you think that offense should be 50% of total WAR and position players should account for 57% of total WAR, then fielding should account for (7% out of 57% = 12.3%) of position-player WAR. Which I think means that the spread on offense should be 8 times larger than the spread on fielding. But it's not; I think the spread is closer to 4 than to 8 which would imply that fielding is being valued twice as much as it should be. But I can't swear that those numbers are accurate.

My Player won-lost records (and Win Shares) avoid this problem by building up from zero with a result that defense splits about two-thirds pitching and one-third fielding. Adding in offense, then, the split is about (3/6) offense, (2/6) pitching, (1/6) fielding. Pitching's share then increases if you shift to a comparison to replacement level - I touch on this in the PDF that I shared above - but the offense/fielding split would presumably continue to hold (at what works out to about (3/4) offense, (1/4) fielding). Note that all of these types of splits in my system are empirical results. I don't constrain anything to make up a certain percentage of the total; these are just the numbers that fall out of the system.
   139. progrockfan Posted: February 13, 2019 at 05:09 PM (#5815084)
There are, I think, two players amongst the ‘old-timers’ backlog who deserve a close scrutiny from me. I’ll start with Gavvy Cravath.

(Undoubtedly a lot of what follows will be old hat to some of you, but I needed to run through the argument for myself. The results genuinely surprised me.)

At first glance, Gavvy’s case seems to center around his prodigious home run hitting. Only six players have led a major league in homers six times or more: Ruth (12), Schmidt (8), Kiner (7), Cravath (6), Ott (6), and Killebrew (6). That’s rarefied company. All of these players other than Gavvy are in the Halls of Fame and Merit. This fact alone demands that we treat Cravath as a serious candidate.

Gavvy played in the Baker Bowl, which featured a ridiculous 280-foot right-field porch. This was probably the homer-friendliest park in the history of the Majors – but for left-handed batters far more than for righties. Chuck Klein’s left-handed swing was tailor-made for the Baker Bowl, Gavvy’s right-handed slash seemingly less so.

Gavvy’s extensive minor-league experience included three years with the Minneapolis Millers, playing at Nicollet Park, which had a 279-foot, 10-inch right-field porch – identical to the Baker Bowl within two inches. The porch was topped with a 30-foot wall, 10 feet shorter than the Baker Bowl’s wall in Gavvy’s day; both walls prevented low-angle line drives from leaving the park, but neither were an impediment to high-trajectory fly balls. With the Millers Gavvy taught himself to pull the ball sharply and hit long flies to right, with dramatic results, including a then-professional record 29 home runs in 1911. When he hit the bigs in 1912, Gavvy had already perfected the skill that would make him famous, and was playing in the perfect park to show it off.

Gavvy certainly deserves credit for tailoring his swing to his park. Many hundreds of players came to bat in the Baker Bowl from 1887 to 1938, but despite his hitting from the ‘wrong’ side of the plate, no player in those fifty-two years figured out how to exploit that right-field porch more consistently than Gavvy.

So, two vital questions:

* How extreme was the home/road effect on Gavvy’s gaudy home run totals?

* Was Gavvy a great hitter apart from his home runs, or was his power his only real calling card for the Hall of Merit?

Here are Gavvy’s home run totals from 1912 to 1919, which encompasses the whole of his full-season experience plus a final year when he managed to lead the NL in home runs while playing just 83 games.

Year  Home  Road
1912     6     5
1913    14     5
1914    19     0
1915    19     5
1916     8     3
1917     8     4
1918     8     0
1919    10     2

We have two seasons, 1913 and 1918, in which Gavvy led the NL in home runs while hitting zero homers on the road. His home run titles must therefore be significantly discounted, which renders his HoM case heavily reliant on his status as a great hitter apart from his home run totals.

Irrespective of his discounted home run totals, Gavvy was a truly great hitter, albeit in a very short Major League career. Had the award existed, he would have been the obvious NL MVP for 1913 and 1915, with 1914 and 1917 also being very strong campaigns. Here’s a listing of Gavvy’s top-5 finishes in various offensive categories for 1912-19:
.                                                                            Tot  Off       Off
Year    R    H   2B   3B   HR  RBI  XBH   TB   BB  TOB  Avg  Slg  OPS  OPS
+  WAR  WAR   RC  Win%
1912                        3         5
1913         1    4    4    1    1    1    1         1    2    2    1     1    1    1    1     1
1914              5         1    2    2    2    5    3         3    2     1    1    3    2     2
1915    1                   1    1    1    1    1    1         1    1     1    1    1    1     1
1916                        3                   5              1    5     4    5               5
1917              4    2    1    3    1    4    3                   2     2    3         4     3
1918              2         1         1         3
1919                        1

Here again, the Baker Bowl played an outsized role in Gavvy’s success. His calculated home/road splits for 1912-19:

Year   Split   Avg   OBP   Slg    OPS  OPS+
1912    Home  .260  .320  .425   .745   109
        Road  .302  .386  .502   .888   160
      League  .272  .340  .369   .710    94

1913    Home  .378  .439  .637  1.076   214
        Road  .302  .373  .496   .869   163
      League  .262  .325  .354   .679    94

1914    Home  .300  .423  .601  1.024   200
        Road  .295  .375  .384   .758   138
      League  .251  .317  .334   .651    94

1915    Home  .276  .401  .598   .999   199
        Road  .283  .381  .426   .807   153
      League  .248  .309  .331   .640    93

1916    Home  .304  .410  .493   .903   174
        Road  .266  .352  .394   .746   141
      League  .247  .303  .328   .632    94

1917    Home  .304  .391  .545   .936   188
        Road  .257  .349  .402   .750   141
      League  .249  .305  .328   .633    94

1918    Home  .246  .318  .469   .787   138
        Road  .224  .321  .310   .632   101
      League  .254  .311  .328   .638    95

1919    Home  .398  .482  .816  1.299   274
        Road  .291  .399  .487   .886   170
      League  .258  .311  .337   .648    95

Gavvy’s home/road discrepancies, especially in slugging, are astonishing. The Baker Bowl also magnified Gavvy’s on-base percentage, as opposing hurlers were forced to pitch him far more cautiously in a park where he could blast any poorly-selected offering clean over that close porch.

But this doesn’t negate the fact that Gavvy’s hitting was far above league average every year from 1913 to 1917, both at home and on the road. Even deprived of that porch, his road slugging is far above league average each year. 1913 and 1915 are MVP-level years no matter how you slice them.

1912 was a minors-to-majors transition year, but interestingly, while his home stats are indifferent, Gavvy’s road stats are well above league average. His 1918 home run title is shown up as the home-field fluke it is, but as a part-timer he once again put together an excellent campaign in 1919.

Irrespective of his extreme home/road splits, Gavvy was clearly an outstanding hitter from 1913 to 1917. That five-year window makes for a solid peak argument. He also had good years in 1912 and, as a part-timer, in 1919. We can also throw in a part-time 1908 (136 OPS+ in 94 games).

And then there are MLEs. Prior to 1912, Gavvy racked up 5156 minor league at bats, versus 3951 in his entire MLB career. I see Cravath’s worthiness as a candidate as dependent on how much career value a voter is willing to add through MLEs. Having read through the entire Cravath thread, I’ll take @Brent’s numbers and @Gadfly’s rationale, and offer credit starting with 1904 – and that’s easily enough for me.

Despite his extreme splits, I’m frankly puzzled as to exactly why Gavvy has fallen off voters’ radar. It looks like he’ll comfortably make my 2020 ballot.
   140. DL from MN Posted: February 13, 2019 at 05:27 PM (#5815089)
for Bob Johnson I basically take the approach that the PCL credit cancels out any further discount of the WW2 years

There are 3 WWII years and two PCL years to consider. That approach basically says to discount WWII at 60% of a non-WWII season. Seems pretty harsh.
   141. RJ in TO Posted: February 13, 2019 at 05:27 PM (#5815090)
Here's the very old link to Gavvy Cravath's Hall of Merit discussion page, which may be of some use in formulating your thoughts on him, and reviewing what others once thought of his case.
   142. progrockfan Posted: February 13, 2019 at 05:36 PM (#5815096)
I've read the entire Cravath thread already - but thanks in any case, @RJ. ;)
   143. DL from MN Posted: February 13, 2019 at 05:36 PM (#5815097)
if you multiplied everybody's fielding number (Rfield) by two (or six or ten or 0.02), the total WAR for position players would stay the same

The upper limit you could get is a fielder who catches everything and a fielder who catches nothing. The maximum spread is directly proportional to the number of fielding plays at that position. It makes sense to me that the fielders who handle the ball the most often would also get the largest percentage of difficult chances and would have the largest standard deviation in Rfield.

This is one reason why catcher defensive numbers in WAR don't pass the smell test for me. Catchers have 100x the chances to handle the ball as a regular fielder (150 pitches versus 27 outs spread across 9 fielders) but their spread in RField is less than centerfielders. They don't have a whole lot of ball-in-play fielding numbers since they don't even field bunts that often but some percentage of pitching value is actually catching value.
   144. Dr. Chaleeko Posted: February 13, 2019 at 07:43 PM (#5815131)
I concur with the comment above that just because pitching balance is north of 40% doesn’t mean the HOM should be comprised of the same percentage of pitchers. I’m of the kind that any hall should likely fall around 30% pitching. Certainly not lower than than 25-28% but also not higher than 35%. I tend to think about it as 30% pitching and then give or take for an individual’s preferences and the actual spectrum of candidates. Once you get above 30%, you start encountering the extreme cases like Dean and Rucker (all peak) on one side then TJ and Sutton (no peak) on another and then guys like Gooden on another (one big year overly influencing any system that has a peak component to it). Those guys don’t really make a big impression until you find yourself at the tail end of that 30%ish area fighting over little tiny edges. The same is true of hitters when you get near 70%, though I don’t see the fracturing there as splintery as with pitchers. YMMV. So that, to me, starts to make it feel like we are near the right balance.

One could, of course argue that this could be true of any arbitrary cut line. That’s probably true, but what’s also true is that baseball is a pyramid-shaped talent/value structure, which means that the further down the sides of the pyramid you go, the more guys become serious candidates because there are simply more players that are similar. And while the three classes of pitchers I mention above may not seem similar to one another, they are similar in that they are all borederline cases with significant flaws that cause the electorate to fracture its support.
   145. Esteban Rivera Posted: February 13, 2019 at 10:23 PM (#5815152)

There are 3 WWII years and two PCL years to consider. That approach basically says to discount WWII at 60% of a non-WWII season. Seems pretty harsh.

You're most likely right, so I'll definitely revisit Bob Johnson for this ballot.

This is one reason why catcher defensive numbers in WAR don't pass the smell test for me. Catchers have 100x the chances to handle the ball as a regular fielder (150 pitches versus 27 outs spread across 9 fielders) but their spread in RField is less than centerfielders. They don't have a whole lot of ball-in-play fielding numbers since they don't even field bunts that often but some percentage of pitching value is actually catching value.

I definitely agree with this. In the 2019 discussion thread I commented on how we tend to view offense and defense and how I've started to think that we don't have it entirely right in how we divide the responsibilities.

Offense can be broken down into the two major responsibilities of hitter and runner.

Defense is typically divided into the responsibilities of pitching and fielding. But if you really think about it, the defensive responsibilities should be viewed and studied as hitter deterrence and runner advancement deterrence. The only two positions on defense that have an active role in hitter deterrence are pitchers and catchers. They are the only two positions that interact with the hitter and can stop a hitter before he becomes a runner. All the other seven positions can only act to prevent the advancement of runners. They never face a hitter per se because the moment the bat makes contact with the ball, the role of hitter ceases to exist and instead the player is now a runner that needs to head to first.

When you think about it in those terms, I believe we are short changing catchers to some degree when you only ascribe the value in hitter deterrence to just the pitchers. Every ball thrown to a hitter has both the pitcher and the catcher involved and both have their responsibilities in this part of the game. No pitcher does their job in hitter determent without the catcher being there (and vice versa). I think we should start viewing catchers as part of hitter determent instead of as a regular fielding position. If you view it that way, then the reality is that some of the value designated as the pitching part of defense should be shared with or belongs to the catcher. How much that should be is the big mystery.

One interesting quirk of the hitter deterrent positions (pitcher and catcher) is that they are the only ones that can simultaneously also serve as runner advancement deterrents when there are runners on (in this case it would be pick-off attempts and throws on steal attempts). Also, most of their plays as fielders would likely fall under discretionary plays (pop-ups, bunt attempts), with the most notable exception being plays at the plate with runners going for home and come-backers hit right to the pitchers.

Because of this, I don’t think catchers should be lumped in with other position players, they are a bit of a unique animal.
   146. DL from MN Posted: February 14, 2019 at 10:17 AM (#5815232)
If we assume that WAR is doing a good job picking the good defensive catchers from the bad ones (probably not correct, it is likely the stuff we aren't measuring is why the variance is too small) and double the fielding contribution for catchers, who pops up as a candidate that wouldn't otherwise?

There are 76 catchers > 5000 plate appearances in Fangraphs database. Here is the top 20 before and after the adjustment

1) Bench 74.8
2) Carter 69.4
3) I Rodriguez 68.9
4) Fisk 68.3
5) Berra 63.7
6) Piazza 63.7
7) Torre 62.3
8) Dickey 56.1
9) Simmons 54.2
10) Hartnett 53.7
11) Cochrane 50.6
12) Mauer 49.2
13) Downing 48.4
14) Ewing 48.1
15) Kelly 45.1
16) Tenace 45.0
17) Freehan 44.8
18) Posada 50.4
19) Parrish 43.4
20) Lombardi 41.9

WAR + 0.1*DEF
1) I Rodriguez 100.6
2) Carter 91.6
3) Bench 90.9
4) Fisk 81.6
5) Berra 74.7
6) Piazza 67.0
7) Dickey 63.4
8) Torre 60.3
9) Hartnett 60.3
10) Sundberg 60.1
11) Y Molina 59.3
12) Simmons 57.0
13) Ewing 56.7
14) Parrish 56.3
15) Freehan 55.7
16) Boone 55.1
17) Cochrane 54.9
18) Kendall 52.3
19) Martin 51.1
20) Munson 50.7

Catcher WAR might not be perfect but Sundberg, Yadier Molina and Bob Boone have sterling defensive reputations. They likely were good at everything behind the dish.
   147. Kiko Sakata Posted: February 14, 2019 at 05:56 PM (#5815471)
In my continuing quest to make my Player won-lost records more accessible in the hopes that more of you will find them helpful in putting together your HOM ballot, the table below puts my numbers on the same scale as WAR. I talk about how to do this in the PDF that I shared, but briefly, eWAR = .0276*eWins + eWOPA + .9724*eWORL. I show "eWAR" using three different positional averages: one-year, nine-year, and long-run. I then show a pWAR (same formula but with pWins - which tie to team wins) just for long-run.

bWAR is from Baseball-Reference; fWAR is from Fangraphs; gWAR is from The Baseball Gauge (I think it's bWAR w/ Michael Humphreys' DRA substituted in for fielding, but I'm not 100% sure of that). oWAR is from Baseball-Reference: this includes positional adjustments, but not fielding values - i.e., this is what the guys' WAR would be if they were all average fielders at their positions. I include this because fielding is one of the things that differs between my numbers and the WAR folks and I thought this might highlight that.

I included all of the significant new guys and returnees as well as most of my pet favorites. I apologize if I left out any of somebody else's favorites. All of the numbers here make no adjustments for schedule length, season quality, no zeroing out negatives. These are just raw career WAR totals.

A few player caveats. I left out Ben Taylor because obviously he was no statistics at any of these websites (except for The Baseball Gauge - I think they have Negro League data). My source is Retrosheet play-by-play data, which only exist back to 1921. I therefore am missing most of Wally Schang's career. The data I have for Schang look like 1921 - 1924 would fit comfortably at the end of a Hall-of-Merit prime for a catcher. But, realistically, I don't think I have enough information to necessarily say anything about his career. Retrosheet is also missing some data within seasons from 1921 - 1936. This affects Tommy Bridges, Bob Johnson, and Kiki Cuyler here. For these three players, I extrapolated my numbers to fill in the games that I'm missing. This isn't ideal and should be taken with at least a grain of salt. I suspect this is one reason for the gap between eWAR and pWAR for Bridges and Johnson, although I'd be hesitant to say which is more likely to be correct. We should have some more data for both of these guys before election time rolls around next December.

So, without further ado. Here you go. Players are arrayed by position and alphabetically within position. Here's hoping that the "code" button works like I think it's supposed to.

.                 bWAR   fWAR   gWAR   oWAR   eWAR(1)   eWAR(9)   eWAR(L)   pWAR(L)
Tommy Bridges     50.9   47.1   54.5      -      39.0      38.9      41.9      58.1
Dwight Gooden     53.0   56.7   52.6      
-      49.8      49.9      47.9      63.7
Orel Hershiser    56.3   48.0   52.5      
-      54.2      54.4      51.9      57.1
Tommy John        62.0   79.4   55.7      
-      67.0      67.0      68.0      81.8
Don Newcombe      38.0   35.9   40.8      
-      33.5      33.8      35.3      45.9
Andy Pettitte     60.3   68.9   64.5      
-      64.8      64.6      61.8      71.1
Johan Santana     51.6   45.3   55.0      
-      48.9      48.9      46.8      52.3
Luis Tiant        66.5   54.8   63.4      
-      55.5      55.6      56.7      71.0

Thurman Munson    46.1   40.9   49.2   43.1      35.8      36.4      38.8      45.6
Jorge Posada      42.8   44.7   45.4   48.7      47.9      47.5      45.9      53.8
Wally Schang      45.0   41.0   48.1   47.6         
*         *         *         *

Lance Berkman     52.1   56.0   53.1   54.1      58.5      58.7      58.9      62.9
Jason Giambi      50.5   49.8   49.0   58.4      56.8      57.0      59.3      66.0
Todd Helton       61.2   55.0   51.4   54.5      54.0      54.0      56.2      43.3

Jeff Kent         55.4   56.1   59.2   60.1      66.3      66.5      69.3      75.9

Sal Bando         61.5   56.2   50.5   58.3      52.0      51.4      54.3      60.1
Buddy Bell        66.3   61.7   64.7   48.1      44.9      45.2      48.7      41.9
Ron Cey           53.8   55.6   55.8   50.6      55.1      55.2      58.4      61.6

Bert Campaneris   53.1   44.9   61.5   47.6      63.7      62.9      54.9      56.0
Dave Concepcion   40.1   39.7   47.2   33.1      54.2      53.8      47.2      61.3
Toby Harrah       51.4   45.7   43.6   62.4      66.0      66.8      65.3      57.5
Derek Jeter       72.4   72.7   46.7   96.3      80.2      80.2      86.2     113.0
Phil Rizzuto      40.8   41.3   41.8   28.5      35.1      35.9      36.5      56.6
Vern Stephens     45.5   48.6   47.7   45.7      66.0      66.7      68.1      74.2

Bob Johnson       57.3   57.0   57.7   55.7      73.0      72.6      70.2      44.3

Andruw Jones      62.8   66.9   62.0   39.9      60.9      61.2      58.1      65.0
Kenny Lofton      68.3   62.4   53.8   57.9      46.0      45.9      42.6      51.9

Bobby Abreu       60.0   59.6   53.4   61.6      60.4      60.6      60.8      59.6
Bobby Bonds       57.9   57.2   64.3   51.9      59.0      59.2      60.7      50.7
Kiki Cuyler       46.7   52.9   49.8   46.1      58.1      58.9      58.1      62.1
Tommy Henrich     35.7   37.1   38.6   33.0      45.8      45.3      43.8      57.4
Sammy Sosa        58.6   60.1   60.6   50.3      64.9      65.5      66.3      50.8 

   148. baudib Posted: February 14, 2019 at 06:05 PM (#5815472)
Where did Mauer go?
   149. Jaack Posted: February 14, 2019 at 07:41 PM (#5815494)
@143- I've come to a similar conclusion regarding catchers - if you look at just the fielding numbers for catchers and compare them to shortstops, you'll see that, according to fangraphs there are 26 shortstops with 100+ fielding runs in their careers, while there are only 5 catchers with that many. Now part of that is playing time - catchers aren't going to play as many games. But even if you straight up double the catchers' fielding stats, there would still be fewer 100+ fielding catchers than shortstops.

Jim Sundberg definitely looks like the strongest candidate if you do try to correct for this apparent fault of WAR - his defense was clearly excellent, and the bat isn't too bad. He perhaps deserves a bit of EC for the 1981 strike since it cut short his best season. He's just over the borderline for me.

I'm also a bit intrigued by Rick Ferrell. Baseball Prospectus' new DRC+ sees him as being a pretty good hitter and his defensive reputation seems solid - the numbers aren't great, but he was likely was better than them, at least in regards to his receiving - he was known for working with knuckleballers like Dutch Leonard. Probably not someone I'm going to be supporting anytime soon, but he's probably closer than I'm giving him credit for.
   150. Esteban Rivera Posted: February 14, 2019 at 09:17 PM (#5815508)
@147 - Thanks a lot for the data Kiko.

From what I've read about gWAR at the Baseball Gauge site ( the key difference between bWAR and gWAR are:

The PythagenPat method is used when converting runs to wins and also uses .285 as the exponent in the formula.

The Offensive component uses Base Runs to estimate a player's value (in runs) above/below average, with the individual player's Base Runs are calculated using the theoretical team method.

The Fielding component uses a multi-year average of Defensive Regression Analysis (DRA) to estimate a player's value (in runs saved) above/below average.

The default pitching component is based on runs allowed just like Baseball-Reference.

The replacement level, league quality and positional adjustments are the same as Baseball-Reference.
   151. kcgard2 Posted: February 15, 2019 at 09:23 AM (#5815578)
@Dr Chaleeko

I figured that if a third of the variance was explained by schedule, and I'd be adjusting separately for the schedule, I should knock out a third of the difference between the league's QoP and MLB so that I didn't double-dip the player.

Two things.
1) where do you do an adjustment for schedule? Adjusting for strength of schedule is not the same as adjusting for length of schedule, which is what is purportedly responsible for up to a third of the variation in NgL performance. You also mention that we don't have information available for strength of schedule so nothing is done there anyway.
2) An adjustment for variation due to schedule length and an adjustment for league quality are two separate things. There is no double dipping involved to adjust for both. I think it is confusing that you lumped schedule length in with your QoP adjustment, but at the end of the day consider. You want to adjust for quality of league. Correct, good thing to do. You want to adjust for the fact that season length was typically only half for NgL compared to MLB. This is another thing you should want to adjust for. It has nothing to do with the quality of league adjustment. You should not erase a third of the QoP adjustment in order to add a schedule length adjustment (which again, I don't think even happened, but correct me if I am wrong).
   152. cookiedabookie Posted: February 15, 2019 at 03:10 PM (#5815740)
@149 I agree with you on Sundberg being a sneaky good candidate with any sort of correction for catchers. In the same vein, those same adjustments have to kill the chances of someone like Posada. He was an awful defensive catcher, and if we double that awfulness, he should drop down most rankings
   153. Kiko Sakata Posted: February 15, 2019 at 03:40 PM (#5815751)
In the same vein, those same adjustments have to kill the chances of someone like Posada. He was an awful defensive catcher, and if we double that awfulness, he should drop down most rankings.

It depends on what you're trying to do with the adjustments. If the argument is that "some of the value designated as the pitching part of defense should be shared with or belongs to the catcher" (Esteban in #145), I read that to mean we should be giving catchers more value - i.e., increasing the number of catchers who are HOM candidates - which, in a WAR framework, would mean either boosting the positional adjustment (Rpos) or the replacement value (Rrep) for catchers. If all you're going to do is bump up good-fielding catchers and knock down bad-fielding catchers, that isn't going to increase the overall value of catchers, it's just going to shuffle the list.

As to Posada specifically, he caught 99 or more games for 11 Yankees teams (in 12 years). Baseball-Reference says those Yankees' teams rank in the AL in ERA+ was 1-2-5-3-4-4-9-10-7-7-[6 in 2008 when Posada missed most of the year]-3. They won 6 pennants in those 11 years and 4 World Series. If you think that Posada (and Jeter) was dragging those numbers down by his "awful defens[e]", you should take a really close look at Andy Pettitte as a Hall-of-Merit candidate (honestly, you should take a really close look at Andy Pettitte as a HOM candidate regardless of how important you think catcher fielding is).
   154. Kiko Sakata Posted: February 15, 2019 at 04:17 PM (#5815769)
Following up on 153, it looks like what DL is doing in #146 is both: he's adding what Fangraphs calls "DEF" (Fangraphs expresses it as runs, DL divides by 10 to convert it to wins) - which includes both the positional adjustment and the player's specific fielding. So, Posada has a "DEF" of 56.9, so he'd get a boost of 5.7 WAR - from 44.7 to 50.4 (the latter is mistakenly listed in the first list in #146, I think). So, he's still getting the benefit of having been a major-league catcher but he gets passed by Bob Boone and Jim Sundberg, because they were much better defensive catchers than Posada (which, no doubt, they were).
   155. cookiedabookie Posted: February 15, 2019 at 04:29 PM (#5815774)
Makes sense. And on Pettitte, I think I may have been the highest on him in this past year's election.

edit: Just checked, and I was :)
   156. DL from MN Posted: February 15, 2019 at 06:52 PM (#5815799)
That's exactly what I did. I'm pretty sure WAR is getting the offensive contribution correct. In order to do it better you would need to double the defensive contribution (which shouldn't move the average), recalculate the positional adjustment / replacement level and re-do WAR.
   157. Jaack Posted: February 15, 2019 at 07:28 PM (#5815804)
It's also important to only double (or modify or whatever) catcher defense. Shouldn't matter much for a guy like Jim Sundberg but for Gene Tenace you wont get accurate numbers unless you separate out the first base numbers.
   158. progrockfan Posted: February 16, 2019 at 11:31 AM (#5815857)
My (very) preliminary 2020 ballot.

1. Derek Jeter. 3400 hits from a middle infield position comes along about once every half-century. His 200 playoff hits reads like a misprint. Played a key role on all four of his World Series-winning teams. A remarkable collection of signature defensive plays for a significantly minus defender; were he the worst defender in history (which he’s not), he might still place first on this ballot.

2. Ben Taylor. Top defensive first baseman in NgL history. 146 OPS+. Ranked 3rd all-time by both James and Holway. Imho, a major HoM omission that needs to be rectified.

3. Bobby Abreu. A five-tool player with rock-solid durability and exceptional plate discipline. Tremendous all-around offense: 7x 40+ doubles, 9x 20+ home runs, 8x 100+ runs, 8x 100+ RBI, 8x 100+ walks, 6x .300+ average, 8x .400+ OBP, 6x 30+ steals. Averaged 156 games played from 1998 to 2010. Slightly minus defense is partly compensated by excellent range and a very good arm. My pick as the most underrated player of the new century.

4. Wally Schang. His .393 career OBP ranks second all-time at catcher. Good career longevity. World Series-winning defense in 1918. Until Mauer becomes eligible, I see Schang as comfortably the best catcher not in the HoM.

5. Luke Easter. Every external indicator seems to point to him being a superstar if he'd been given the chance. 141 OPS+ at age 36 despite major injuries. Like Paige, I see him as deserving of significant pre- and post-career MLEs. Though I acknowledge the uncertainty surrounding him, I see Easter as the HoM's single greatest omission.

6. Bob Johnson. A teriffic combination of steady offense, above-average defense and tremendous longevity. Had substantial careers both before and after his MLB playing time. II strongly suspect that given the right opportunities he would've been a Winfield-type 3000 hit man, with less basepath speed but more walks and possibly a bit more XBH power.

7. Johan Santana. The electorate has sold me on his greatness. Three firsts and a second in ERA, three firsts and two seconds in strikeouts, four firsts in WHIP. His all-too-short career was the very definition of ‘high-impact’.

8. Phil Rizzuto. Great defense up the middle, an MVP, and a solid chunk of WWII credit get him here.

9. Kenny Lofton. A monster center fielder with outstanding range and arm. A basepath blazer with 622 steals at a 79.5% rate and five stolen base titles. Not very durable, and couldn't maintain his initially very high levels of offense.

10. Bobby Bonds. Seven consecutive 20-20 seasons and great range on defense. Decent plate discipline despite the strikeouts, and drew plenty of walks to boost his offense.

11. Todd Helton. .300-.400-.500 is pretty difficult to compile, in or out of Colorado. Decent glove man as well; I think he'll be in the HoM sooner or later.

12. Hugh Duffy. The greatest defensive outfielder of the 1890s. His .440 average in 1894 is tops all-time. Compiled a .300+ career average in all four of his major leagues – an unparalleled accomplishment.

13. Dolf Luque. Cuba credit pushes him ahead of Tiant for me; otherwise I see their candidacies as quite similar.

14. Luis Tiant. Two ERA titles, but lacking the consistency of a truly great pitcher.

15. Gavvy Cravath. A truly great hitter for a relatively brief window, with MVP-quality seasons in 1913 and 1915. Deserving of substantial pre-career MLEs. The greatest home-run hitter of the pre-live ball era, though the Baker Bowl dampens this distinction.

16. Thurman Munson. I suspect at least a partial illusion of context on his superficially excellent counting stats. A decent player, but many steps behind Schang in my book.

17. George Van Haltren. A stubborn, idiosyncratic holdover for a long-time personal favorite. Even in a high-scoring era, he stands out as a monster run-scorer. Plus 40-31 as a pitcher!- God bless you, George.

18. Andy Pettite. Above-average consistency, with year after year of the same basic solid value, is a valuable asset for a pennant contender, and is the best thing Andy's got going for him.

19. Andruw Jones. The anti-Ichiro. Charting his weight-induced collapse from 2007 onwards is like watching a stock ticker in October 1929. Only raw defensive magnificence keeps him on my ballot.

20. Sammy Sosa. Despite the corked bat, despite the multi-year clubhouse disruption, despite abandoning his team on his final day, despite being persona non grata with the team for which he won his MVP, he did hit a bunch of home runs, and therefore grudgingly makes the very lowest rung of my ballot.
   159. Dr. Chaleeko Posted: February 16, 2019 at 09:17 PM (#5815931)

So basically, I'm double-dipping. That's a problem. I reran a handful of fellows from among various strata of the candidates: The cream, the solid HOMers, the borderline+, the borderline-, and also-rans.

J Gibson goes from 92.4 to 81.0 WAR, 12% loss (from Jimmie Foxx to Charlie Gehringer in BBREF's WAR rankings for the Negro Leagues era)
O Charleston 95.0 to 83.3, 12.5% loss (Foxx to G Davis)
W Wells 94.1 to 83.9, 11% loss (Foxx to Davis)

B Leonard 73.4 to 62.8, 14.5% loss (S Crawford to Beckley)
J Wilson 72.8 to 63.7, 12.5% loss (Waner to Goslin)
P Hill 81.7 to 70.9, 13.5% loss (Gehringer to Frisch)

B Taylor 69.5 to 59.1, 15% loss (A Simmons to S Magee)
R Dandridge 62.7 to 55.2,12% loss (HR Baker to E Slaughter)
R Garcia 35.2 to 31.7, 10% loss (Jim Bottomley to Ted Klu)

J Beckwith 53.9 to 46.5, 14% loss (J Sewell to Cuyler)
G Scales 55.1 to 45.8, 17% loss (Billy Herman to Del Pratt)
M Suttles 59.6 to 50.7, 15% loss (Z Wheat to Joe Kelley)

P Chacon 51.6 to 45.5, 12% loss (Stan Hack to Vern Stephens)
CP Bell 58.1 to 50.9, 12.5% loss (Sherry Magee to Joe Kelley)
B Pettus 50.8 to 43.5, 14.5% loss (Kelley to C Klein)

The average loss is 13.2% from my current estimates. What's this mean in context?Currently I've got about 30 guys over the in/out line as CHEWS+ (my sifter) sees it. CHEWS+ includes a measurement of the median career value among the top 40 candidates at each non-pitcher position. That's a decent proxy for how much career value you need to get into the HOM at a given position. Let's knock 13.2% off all my MLEs, and then say that we multiply each Negro Leaguer's career MLE WAR by 1.05 to account for the 154 schedule. How many will now fall below that median at their primary position?

7 Currently above the median of 50.4 WAR at position: Gibson, Mackey, Santop, Trouppe, Regino Garcia, Gervasio "Strike" Gonzalez, Bill Cash
After QOP update: 4 (Gibson, Mackey, Santop, Trouppe)

2 currently above the median of 60.1 WAR at position: Leonard and Taylor
After QOP update: Same 2

3 Currently above 2B median of 57.0 (M Williams, B Serrell, G Scales)
After QOP update: 1 (Williams, though Serrell is just shavings below the median after)

3 currently above 3B median of 54.3 (J Wilson, C Moran, R Dandridge)
After QOP update: Same 3

7 currently above SS median of 60.4 (J.H. Lloyd, W Wells, D Moore, D Lundy, S Garcia, S Bankhead, G Johnson)
After QOP update: 6 (Moore drops off)

None above median

11 currently above median of 56.7 (Charleston, Dihigo, Stearnes, Hill, Torriente, Oms, Irvin, W Brown, Salazar, Jethroe, Bell)
After QOP update: 9 (Bell, Jethroe fall away)

4 currently above median of 62.4 (Rogan, H Johnson, H McNair, B Wright)
After QOP update: 2 (Rogan and McNair)

So we go fro 37 to 27, which is a pretty substantial loss. However, five are below the in/out line by CHEWS+ anyway because they already don't have enough peak. We can't be sure what kind of slashing of peak we are looking at (though, to be fair the regression-to-the-mean aspect of things saws off the heights of peaks anyway when it comes to MLEs). In addition, there's some variation around that 13.2% figure that might skew one way or another. What I'm concerned about is that the cream is getting pushed down into the milk.

Anyway, this seems to be a problem worth fixing. Do we need to optimize anything else at this juncture for hitters?
   160. cookiedabookie Posted: February 19, 2019 at 03:15 PM (#5816547)
Don Newcombe passed today. He's not quite a HoM player for me, but a great player, and from all accounts a great guy.
   161. Bleed the Freak Posted: February 20, 2019 at 07:10 PM (#5816959)
Don Newcombe      38.0   35.9   40.8      -      33.5      33.8      35.3      45.9

Rest in peace Newk...

Kiko, how does he look updated for MLE and war credit for you?

And great read on the pdf you shared.
   162. Kiko Sakata Posted: February 20, 2019 at 10:01 PM (#5816976)
Kiko, how does he look updated for MLE and war credit for you?

And great read on the pdf you shared.

Thanks, Bleed!

Newcombe looks pretty good. Maybe three 1955-caliber seasons away from the bottom of my ballot before giving him bonus credit: 1952 and 1953 could get you two-thirds of the way there. I'm not entirely sure how much MLE credit he deserves. He certainly stepped into MLB ready from day one but he was only 23 in his rookie year. I'm also never sure what to do with guys who have down years in their first year back from war - Newcombe's 1954 was pretty easily his worst season between 1949 and 1955 or 1956. It seems plausible that it would take some time to re-adjust after two years off, but should he get bonus credit for that (this applies even more strongly to Phil Rizzuto who, in my opinion, needs absolutely every last bit of war credit you're willing to give him to be a credible candidate).

Newcombe's case reminds me a lot of Luke Easter but with a bit more certainty and higher-level MLB performance for Newcombe. If these two guys were born 10 or 20 years later, I suspect they'd both already be in both Halls. And the same might be true if they were born 20 years earlier, too.
   163. kcgard2 Posted: February 21, 2019 at 06:29 PM (#5817216)
@Dr Chaleeko (159)

Thanks for digging into that. Those numbers look much more palatable to me on first glance. That's a notable adjustment for sure! but it's in line with the adjustment I was intuiting for QoP as discussed before, and I enjoy when my intuition turns out to be luckily right ;)

I have been very busy the past couple weeks, but I think there are a handful more points I want to discuss about the MLE methodology, just continuing down through your steps. I will hopefully have some time to get to some of that this weekend, but if not, I guess I just have to ask to bear with me in that I'll get to them at some point. We've still got time before the next election, right...

Just to reiterate, in no way do I hope you will consider any of my opinions or critiques as a dismissal of your project. Ideally, I'm trying to help get to the best answer we can, and not denigrate your work by any means. I am somewhat in awe at the dedication of time and effort that must have gone into it and think does a valuable service for our understanding of baseball history and the Negro Leagues. I also appreciate your obvious willingness to take feedback and be intellectually honest and objective about a project that is obviously meaningful to you. It has obviously already impacted the HOM in a notable way. In short, I have a ton of respect for the foundation you've created.

Out of curiosity, how long did it take you to collect all the NgL data and run your MLEs once you had nailed down the methodology? And are you working all the data in Excel?
   164. Dr. Chaleeko Posted: February 23, 2019 at 11:10 AM (#5817564)
Out of curiosity, how long did it take you to collect all the NgL data and run your MLEs once you had nailed down the methodology? And are you working all the data in Excel?

A very long time. In fact, it's a big enough project that I chose to collect data as needed rather than all at once. For my level of sophistication with various softwares, that's what I was capable of doing. (Still am.)

I'm not proficient with database software, SQL, python, or anything that might help me collect/scrape and/or query data. Which means a lot of cut/paste and some hand entry (mistypes are my own only). I do use Excel, which I am proficient with, so lots of VLOOKUPs going on. In addition, however, there are numerous instances where a judgment call may be required, in which case there's no automation to be done.

Each player has their own Excel workbook with several tabs of information and analysis. In addition, I have background files for league/team/position contexts, STDEV contexts, and a few other goodies. As you say, it's a high-time/effort operation, which is cool because I don't have children. ;)

But also, I'm working to some degree in a vacuum, so it's helpful to have someone help me to improve the rigor involved.

KC, did you get the email I sent via the site? If it would help expedite things to talk by email and present the highlights to the group, we can do that too. I'm at Eric dot Chalek at gee mail dot com.
   165. Dr. Chaleeko Posted: February 23, 2019 at 11:11 AM (#5817566)
Oh, also, the reason I'm in a bit of a hurry is that Miller and I are going to start electing Negro Leaguers in March, and I'd like to have the latest and greatest MLEs to report as part of that process as well. So it's not just a December deadline for me. Not that you have to have the same urgency as I do, but I wanted to let you know why I got ants in my pants and I need to dance.
   166. kcgard2 Posted: February 23, 2019 at 04:20 PM (#5817609)
No, but sent you an email instead, DrC. Also updated my profile email on here.
   167. Esteban Rivera Posted: March 08, 2019 at 02:36 PM (#5821283)
As always a big thanks to Dr. Chaleeko for the work you continue to do with the MLEs and also to kcgard2 for providing the recent feedback in helping to fine tune them. Even with the decreases, I’d say it establishes a clear candidate pool for evaluation and a pecking order for sorting them into subgroups. The change also brings the values a bit more in line with the career values you would expect from players from the pre-1961 era. You still get a large amount of the NGL candidates having their MLE careers starting from ages 18-21 and lasting up to ages 37-41 (and beyond) which may not have been the case if they had been in the majors, but it’s part and parcel of the reality of placing the careers they did have (they did play at those ages) into some value measure to help with evaluations of what they accomplished.

The one area where there still is a trade-off of sorts is in terms of candidates who are are more peak than career in nature looking worse in the comparisons since the peaks are smoothed out (but this likely starts becoming an issue when you start trying to decide among the next tier of worthy candidates after the stand outs). Missing seasons data definitely plays a role here for a few of the candidates.

So even after this adjustment, I still see the cream remains the cream relative to the other candidates in the NgL player pool and should definitely help in organizing the candidates. Would the pitchers also require an adjustment or would this only affect the position players?
   168. kcgard2 Posted: March 12, 2019 at 08:02 PM (#5822261)

There was at least a small adjustment for pitchers that involved fixing league RAA/IP factors. Whether Dr. Chaleeko decided on any other changes to the pitching methodology I don't know, but I would guess that will probably be the biggest change, along with weighting season performance by IP when doing the multiple-season rolling average and a minimum IP threshold over multiple seasons, which in the end had an impressively small (net) effect for hitters, so I assume similar for pitchers even if he implements that change. Also, would you happen to be Miller?

When Dr. Chaleeko wraps up whatever changes he ultimately decides on, I plan to present some of the discussion that we've had here at BBTF for general discussion and/or education about the methodology.
   169. Dr. Chaleeko Posted: March 13, 2019 at 12:23 PM (#5822423)
KC, go ahead and release the hounds.

Esteban, one of the biggest things affecting all players that I’ve done is to move to using the NLDB’s park factors. They have different factors for batters and pitchers, which I didn’t, and they are computed in a real way from h/r stats, where mine weren’t, and they include strength of schedule. Generally this didn’t make a big difference, but in one case it certainly made a massive difference. That person is Willie Foster. I’ll be writing more about that at another time, but he dropped like a rock to well below the in/out line. More later.
   170. Esteban Rivera Posted: March 13, 2019 at 12:54 PM (#5822439)

I'm not Miller, although I do enjoy reading his and Dr. Chaleeko's writings on their blog. I'm an old HOM voter who was inactive for the last few years and recently have started participating here again. Thank you for the info and look forward to what is presented of the discussion.

@Dr. Chaleeko

Thanks for the update. One question, did you use the upgraded info the NLDB is doing to their ballpark and strength of schedule adjustments? Not sure if the upgrades/changes are live or not, but I've heard that they've been working on this. Don't know how these changes would affect your MLEs again if they are not already incorporated. I had done some comparisons of the candidates doing a rough -13.2% to the MLEs (as you had suggested in one of your previous posts), but I'll wait for the updated numbers before posting further thoughts. Thanks again for all your efforts.

   171. Dr. Chaleeko Posted: March 13, 2019 at 07:00 PM (#5822602)
I'll let KCGard2 tell you about the details, but here's the 30,000 foot view.
1) Previously, I regressed the Quality of Play back 33% due to shorter schedule. That regression is removed.
2) Previously, I used my own home-cooked park factors. Now I'm using the NLDB's latest PF/SOS figures.
3) Previously, I used a five-year rolling average weighted as season n = 60%; seasons n+1 and n-1 are 30%; seasons n+2 and n-2 are 10%. Now I'm using a technique that instead is a 200-PA minimum. Like this:
-If season n => 200 PA, then no additional seasons required
-If season n <200 PA, then add seasons n+1 and n-1 at same weight as season n
-If those three seasons <200 PA, then add season n+2 and n-2 at 60% of other seasons
-If still under 200 PA, then add the number of missing PAs at career average.

1) See number 2 above
2) Similar to number 3 above but using 30 innings.

KCGard's highlights will show why 200 and 30.

Now, here's the results expressed as career WAR.
NAME         PREV   NEW
R Garcia     35.2   35.8
J Gibson     92.4   84.0
B Mackey     58.2   53.7 
L Santop     48.9   52.6
Q Trouppe    52.5   46.9

B Leonard    73.4   70.7
B Taylor     69.5   62.9

M Williams   60.6   61.6

R Dandridge  62.7   56.5
J Wilson     72.8   66.7

S Bankhead   72.9   66.0
G Johnson    68.0   72.8
J Lloyd      86.8   91.5
D Lundy      71.1   60.8
D Moore      63.4   64.1
W Wells      94.1   91.4


O Charleston 95.0   88.0
M Dihigo     96.3   85.7
P Hill       81.7   67.5
M Irvin      63.5   65.2
A Oms        72.7   62.9
T Stearnes   72.0   75.1
C Torriente  72.0   69.3

H Johnson    65.8   66.8
H McNair     74.5   66.0

D Barnhill   57.5   59.8
R Brown      64.8   64.8
B Byrd       64.2   62.0
M Dihigo     87.4   89.8
W Foster     70.5   51.3
J Mendez     63.8   59.1
S Paige     121.5  107.1 
D Redding    92.4   77.8
B Rogan     104.8   96.2
R Welmaker   59.8   57.3
J Williams  122.6  102.7

The answer to your question is park/SOS factors.
   172. Dr. Chaleeko Posted: March 13, 2019 at 07:21 PM (#5822605)
Just thinking here about the future. We alternate threes and fours each year. Here's the players we know will be eligible through 2024:
2020 (4): Jeter, Abreu
2021 (3): Hudson, Buehrle
2022 (4): Ortiz
2023 (3): Beltran
2024 (4): Beltre, Mauer, Utley, Wright (maybe Ichiro)
2025 (3): Sabathia (maybe Ichiro)

21 guys over the next six elections. Among them, Jeter, Beltran, Beltre, Mauer, and Sabathia (and Ichiro) seem like locks for first ballot. That's five or six, which leaves about 15. Here's our top 20 backlog:

Most of these guys will be HoMErs pretty soon. Anyone got a passionate plea AGAINST any of them?
   173. DL from MN Posted: March 14, 2019 at 02:18 PM (#5822826)
Anyone got a passionate plea AGAINST any of them?

Only 20% pitchers as you go down the backlog list (one every 5). Only a few of those guys lower standards at the position they represent (Helton, Berkman, Bell, Posada, Bando) but not by that much. Most of the players are from the past 30 years of baseball and that's who those spots were designed to pick when the project was set up like this. Giles, Pettitte and Oswalt could get another look too.

15 Recent Players?
Jeter, Beltran, Beltre, Mauer, Sabathia, Ichiro, Utley (he's a lock), Helton, Lofton, Jones, Kent, Santana, Sosa, Abreu, Hudson

That leaves 5-6 slots for old timers at most:
Tiant, Taylor, Schang, Rizzuto (who was almost elected once) all seem to meet standards we have established previously.

Then the question boils down to do we elect two from the 1970s (Munson, Bonds, Bell, Bando, John, Campaneris or Concepcion), look somewhere else in the backlog (Newcombe, Gavy Cravath, Tommy Bridges, Bob Johnson, Urban Shocker) or elect two more new guys (Ortiz, Giles, Pettitte, Oswalt, Buehrle, Berkman, Posada, Wright)?
   174. Rob_Wood Posted: March 14, 2019 at 07:40 PM (#5822908)
As a reminder to anyone who uses WAR in their voting (any anyone else too) to be "fair to all eras" it is important to take into account different schedule lengths. Of course, before the 1961/1962 expansion, MLB seasons were 154 games (even fewer going way back in history) compared to 162 games in "modern" times.

Giving an across-the-board 5% bump to pre-1960's players is an easy way to accomplish this. Although 5% may seem small, a season-length adjustment has impacted my HOM ballot several times in the past and I expect it to be relevant for a long time going forward.
   175. DL from MN Posted: March 15, 2019 at 10:00 AM (#5822981)
Giving an across-the-board 5% bump to pre-1960's players is an easy way to accomplish this.

Good point, but this adjustment breaks down before 1900 when it comes to pitchers.
   176. caspian88 Posted: March 16, 2019 at 11:56 AM (#5823146)
Good point, but this adjustment breaks down before 1900 when it comes to pitchers.

I'm working on a system to allow me to submit a ballot - it's WAR-based and similar to Bill James/JAWS in structure (weighted career/peak/prime).

What I did for pitchers is, after adjusting for season length, I adjusted pitching WAR for the percentage of total league innings thrown by the top N pitchers (N being the number of teams in the league that season) compared to a rough historical average (right now I'm using 19%). The logic here is that a HOM-quality pitcher necessarily is ace-quality most of his career, so I take the percentage of innings pitched by "aces".

For 1870's and 1880's pitchers, it's easier to go season by season, because of how rapidly pitcher usage patterns changed. Later than that and I can generally take an average of a pitcher's career (really just their full-time years).

This tends to boost pitchers from the late 1980's onward and weaken pre-1930 pitchers. However, for most pitchers the effects are relatively small and they pass the sniff test. For 19th century pitchers, the effects are also pretty plausible, but there are a few guys I think are still a bit overrated (Tommy Bond).
   177. kcgard2 Posted: March 20, 2019 at 05:57 PM (#5824137)
OK, let me get to a few of the other points I wanted to touch on regarding the MLEs.

Estimate the player’s games played into an MLB schedule based on his in-season and/or career durability records
My thoughts: This is an adjustment which I might call an unfortunate but necessary evil. Of course, something along these lines has to be done because we need an estimate of playing time. However, I think the simple method you have used here can probably be improved. And it's one of the few times where I have a specific methodology in mind. I think you would admit yourself that playing time estimates via your method are probably too high, which you allude to with the comment about capping at 95% because otherwise there's too many seeming iron men from NgL. I have a hunch that even capping at 95% you have NgL players looking on the whole (notably?) more durable than MLB players.

Here is my proposal: Find a player's location on the distribution of durability compared to the rest of his own league. For a made-up example: a player plays 78/84 team games, and that puts him at the 85th percentile of durability for his league based on this metric. Find out what the 85th percentile of MLB durability is, and assign that as the expected number of games for the translated NgL player. That way, you are going to end up with a playing time translation that mirrors reasonable estimates in an MLB context. What do you think of this? I would also make a few other statements to be aware
of. In earlier discussions on BBTF, you've mentioned smaller traveling rosters for many NgL teams, so you want to make sure you capture all players' "durability" metric based on games where they had actual opportunities to appear, and not simply overall team games. You could tell me whether that would make a notable difference, but I think it would.

I like your suggestion here. I think it would increase the sense of reality. It’s got some tricky elements because of all the jumping that players did among teams during seasons, but I think it’s possible to do.
I would quote the rest but can't find it in my email chain anymore. Essentially, the league switching turned out to make this apparently much trickier than DrC was expecting, and he decided not to follow it through because of that difficulty.

Because of small traveling rosters and small non-exhibition seasons, I think the playing time estimates still represent a small issue for the MLEs, in that NgL players as a group probably look much more durable than their MLB counterparts according to the MLEs.

Apply the destination league’s PA/game to those estimated games
My thoughts: Yes, and I'm glad you incorporated this, because it would be an easy adjustment to miss. For players on either side of the translation who are playing in either a very high run scoring environment or a very low one, this could end up making a notable difference.

This is probably my favorite element of the MLEs, from a practical perspective.

Express as RC, figure his translated Rbat, then express the result per PA
I think this particular step in the adjustment generated the most discussion and exploration of any element we talked about.

My thoughts: I actually think this might regress slightly too much. Given that you are already incorporating a length of schedule adjustment, I would probably not regress single season performance to 60%. That's very strong, IMO. However, I am not arguing from a position of statistical/number theory. If you feel 60% is right, and I feel 75% is right, it's just a matter of preference which one to go with. I think the concept of a rolling average is very interesting though, and as of right now, inclined to think it's a good idea.

Again, I unfortunately can't find the earliest emails we exchanged on this, but I suggested switching from a pure rolling average to a method that would weight each season by the number of PAs in that season (ie, the amount of actual information contained in the data for that season), and set a minimum threshold on the acceptable number of PAs to have before considering the data from those seasons reliable. DrC really opened my eyes to some of the difficulty of this project when he mentioned how often NgL players have seasonal data on the order of a few dozen PAs or less ('s way too often).

I don’t remember if I responded to this email or not. But I really like the idea of weighting the rolling average toward the seasons with the most PA. Very wise, and probably pretty easy to do. I will work on implementing this soon. QUESTION: What do you think is the minimum number of PAs acceptable in a sample? There may be a few instances where even five seasons are summing to less than 100, certainly less than 200 PAs.

We examined some of the sabermetric work that's been one on the stabilization sample size for various metrics, and unfortunately the picture was kind of bleak. Things that can be picked up reliably from box scores take 300 to 1000+ PAs to stabilize (to a mere 50% signal/50% noise, but it's something). Many NgL players take half their careers or more to get to PA totals like that. But in any case, I suggested 200 PAs as an intuitive guess for a decent tradeoff between being able to reach that threshold, and the threshold being big enough to capture meaningful signal. DrC ran this new method on a test set of players.

In general, it turns out that the average degree of change of all 365 seasons of test data is about 0.02%. ... You’ll recall from our previous exchanges that I had compared the test runs to the overall current method (current = without a change in QOP method or the 200-PA minimum) to the test runs with QOP and the 200-minimum method, and I’d found that the test runs came in 10% lower. So dang all of the difference in runs is really at the QOP level. So the runs method probably doesn’t matter as much as I thought it might. But that doesn’t mean that the 200-minimum method isn’t well optimized anyway, and I probably should pursue it for all players.

After seeing these results (which I found highly surprising), I tested a few other cutoffs that could have been used for the minimum PA threshold to see how they would have changed the results. And in the end, it mattered very little *in the aggregate* whether the cutoff had been chosen as 120, 150, 200, 250, 300, or more, though player seasons start to become very rare after 300 or 350. Since DrC had already done a bunch of work using 200, he continued with that for all players. Ultimately, the fact that DrC's original method and this new approach generated remarkably similar aggregate results was something that made me feel better about both methods.

If I disagree systematically with the MLEs as a whole, it is not due to the underlying method for determining offensive rate performance. The (sometimes severe) lack of data is a difficulty that simply has to be lived with to some degree, but the method for translating RC/PA seems sound to me.
   178. kcgard2 Posted: March 22, 2019 at 05:43 PM (#5824745)
I don't like the formatting on the previous post, kin of hard to tell what's what. Let me continue and see if I can be more clear.

One other interesting idea I've had that I know you'd be able to do is an alternative to the "weight by PA" approach that you're investigating for low or no PA seasons. If you proceed with the weight-by-PAs and ultimately decide you don't like it, this could be an alternative. The thinking would go as follows:

Let's start from the assumption that for seasons where we have decent numbers of PAs, we are happy with your current methodology and think it's doing about as good as we can do, but for seasons with very few PAs we honestly have no idea what we should do for imputation of those because there's no data. In order to use the idea I'm about to propose, a player would have to have at least a handful of years where we have decent number of PAs, my intuition would say *at least* 4 seasons as a minimum. Use your current methodology on those seasons (an I *really* hope they are consecutive or nearly consecutive!). Now, you have an MLE for those seasons, and at this point I will do my usual and create an example. We have at least 150 PAs for player X for his age 31-36 seasons, but his early career ages 22-30 is full of seasons from 0-40 PAs for our data. Use current methods to create an MLE for his ages 31-36. Now find MLB comps who match player X closely specifically in ages 31-36, ideally not just on overall value but also on the trend/shape of value. We'd like to have maybe 20 or more. Since it's a shorter period of time to match, there should be more matches than looking at whole careers. Now, from our group of comps, we find out the median value of their stats. For example, the median number of PAs in age 22 season, age 23 season, and so on. The median Rbat for the group in age 22 season, age 23 season, and so on. We use these values to create the seasonal MLEs for player X.

I like this approach for several reasons. One, projecting MLEs from samples as small as some we've been discussing is foolhardy in a mathematical sense, and using the median of a closely matched comparison group is at least an unbiased approach. It will also incorporate things that we can't know like injuries and random fluctuation, and general aging trends naturally. Players in the comp
group will have all sorts of career trajectories, so some of the unpredictability for injuries, unexpected career paths and so forth will be lightly baked into the method. Using career averages as we might be doing now will create MLEs such that NgL players with little data will have careers that look much too "normal" on the average. Also, we're using informative data to create the MLE rather than...essentially nothing, sorry to say.

I understand the appeal of using the player's own data to the greatest extent possible to create the MLE, but this is maybe the best method I can think of when the data is far too sparse. Just another avenue to consider.

I spot a potential issue for this method. It is the discontinuity of zero/small PA seasons. There are just two semi-predictable locations for zero/super-low PA years: Prior to about 1901, 1926 (for NNL teams only), and 1929. You otherwise never know. So that’s a big thing because you may have broken up careers in terms of low PA seasons occurring in lots of places across a career.
   179. kcgard2 Posted: March 22, 2019 at 05:57 PM (#5824747)
Now we look for players with similar careers and styles to our man. Once we have a bunch of them identified...we look at their playing time [and] look for a rough number of plate appearances to shoot for
So, I'm not sure about this step. Didn't you translate playing time in an earlier step on a seasonal level? I'd like a little bit more explanation about your reasoning for this step.

Armed with the information in #11, we adjust...
Same question as above. I agree with altering playing time in very young and very old seasons, but I thought you'd already done it previously.

Now, find MLB players from the PBP era with similarly long careers and find similar percentages of SB above the league
Good, no problems with the methodology. I feel like we can probably do something better than pad at random for guys who are "speedy," but I'd have to think about it. I don't love just picking some numbers and padding with them, but I understand due to the missing data why you do it. I do have one question. Once you've gone this far through the steps, what do the Rbaser numbers look like for NgL as a whole? Pick a year or three and find out the distribution of Rbaser for one league. Does it resemble a comparable MLB distribution within reason? That might be a good guide as to how much padding might be appropriate.

One other thing, from your Charleston example, it looks like you used a single season of Charleston's steals v league (where he was 59% better than league, padded to a range of 70-90%). Then you found MLB comps whose CAREERS were 70-90% above their leagues. If you want to find comps, you should do careers vs careers, or seasonal patterns vs seasonal patterns, for example, say Charleston's career with padding looks like 25, 35, 55, 70, 80, 80, 80, 40, 0, 0, -15, -15 just to make something up for percentages of steals above league. Then you' look for MLB comps with similar patterns. Or maybe Charleston's padded career value was 20% steals above league, then you'd look for MLB comps based on that. I would prefer the first method, and I think you would too. Or maybe this is what you already did and I misinterpreted.

DrC indicated that the baserunning section is badly out of date, but I can't find anything on the website newer than the walkthrough I've been referring to throughout.

If the candidate has a pronounced decline in his net steals versus the league, sculpt the trajectory of his running runs appropriately
Could you explain how this works? If he has a marked decline in steals vs league, wouldn't that just show up as lower Rbaser scores for those seasons?

Again, maybe if I could find the updated baserunning explanation, this would be addressed.
   180. kcgard2 Posted: March 22, 2019 at 06:04 PM (#5824752)
One new thing occurs to me about padding the steals above league stat. All player should pretty uniformly have SB data missing. Thus, the rate above league should stay close to the same. For example, we pad Charleston from 59% above league to 80% above league because of the missing data. However, if we HAD that missing data, other players' SB rates would go up about proportionally to Charleston's, leaving his rate still at more or less 59% above league. Now, whether the speediest players would suffer by the missing data disproportionately compared to other players is something I'd have to think about more. But if SB data is missing in 25-50% of box scores, the rate of padding should certainly be less than 25-50%, because even the speediest player is not going to have a monopoly on stealing bases in those games among his league mates. Thoughts?
   181. bachslunch Posted: March 26, 2019 at 02:07 PM (#5825492)
Revised slightly. For some reason I thought Tony Perez was in. Also moving Downing up because he played a fair bit at catcher. That will put the latter in my top 40 now.

1. Derek Jeter. Excellent WAR and hit well at a premium position. Yeah, he was darned near Dick Stuart at SS and overrated by many. Still doesn't change things for me.
2. Jim McCormick. Best WAR for starters not in by a mile; even removing all his UA-earned WAR leaves him a point up on Tiant. Short career, but played in NL except for one UA season.
3. Luis Tiant. Best WAR for non-19th century starters.
4. Buddy Bell. Best WAR at 3B. Currently inclined to trust the metric for him.
5. Andruw Jones. Best CF WAR. Close between him and Bell for me.
6. Jeff Kent. Was best WAR at a middle infield position before Jeter came on the ballot and hit well, can't in good conscience rank him below Helton, Sosa, or Johnson.
7. Todd Helton. Excellent WAR and the best qualified non-NGL 1B.
8. Ben Taylor. Best NGL position player per Seamheads.
9. Bobby Abreu. Best WAR among available RFs, definitely better than Sosa.
10. Wally Schang. Among best C WAR, also hit well.
11. Bob Johnson. Best WAR among available LFs.
12. Vic Willis. Good WAR.
13. Sammy Sosa. Better WAR than I remembered. Happy to give him some benefit of the doubt given his treatment by the BBWAA.
14. Vern Stephens. I value hitting at a premium position highly, so I'm ranking him here.
15. Kenny Lofton. Not as much hitting as I'd like, but lots of WAR at a premium position.

16-40. Tommy John, Sal Bando, Ernie Lombardi, Thurman Munson, Mickey Welch, Urban Shocker, Tommy Bridges, Joe Tinker, Jim Fregosi, Bobby Bonds, Andy Pettitte, John Olerud, Luis Aparicio, Bert Campaneris, Johan Santana, Gavvy Cravath, Jorge Posada, Ron Cey, Tony Lazzeri, Jose Cruz, Jack Quinn, Harry Hooper, Brian Downing, Lance Berkman, Willie Davis.

1B. Helton, Taylor, Olerud, Perez, McGriff, Cash
2B. Kent, Lazzeri, Evers, Phillips, Myer, Pratt
SS. Jeter, Stephens, Tinker, Fregosi, Aparicio, Campaneris
3B. Bell, Bando, Cey, Ventura, Elliott, Harrah
LF. B. Johnson, J. Cruz, Downing, Berkman, J. Gonzalez, Veach
CF. A. Jones, Lofton, W. Davis, Lemon, Damon, Pinson
RF. Abreu, Sosa, Bonds, Cravath, Hooper, J. Clark
C. Schang, Lombardi, Munson, Posada, Tenace, Kendall
P. McCormick, Tiant, Willis, John, M. Welch, Shocker, Bridges, Pettitte, Santana, Quinn, Cicotte, Finley, Tanana, Powell, Hershiser.
   182. Bleed the Freak Posted: March 28, 2019 at 09:33 AM (#5825993)
Happy opening day, hope it's a fabulous and safe one for everybody!
   183. kcgard2 Posted: March 30, 2019 at 09:28 AM (#5826994)
There was much less discussion on pitchers. As already mentioned DrC fixed an issue where league RAA/IP was not equaling zero. Beyond that, I had only a few remarks about the pitching methodology.

[One issue is] the odd (by MLB standards) usage of NgL pitchers. You say that NgL starters were often used in shorter stints but in more games compared to their MLB starting brethren. I think this is actually an important factor to consider. And I think there are two ways it might impact our MLEs. Most importantly, the pitching aspect. This usage is a sort of hybrid between starting and relieving. Shorter stints allow for greater effectiveness. Now, if ALL NgL starters are essentially being used in this manner the same proportions of the time, then everything is OK, because every NgL pitcher is deriving the same performance benefit from it and comparing a particular pitcher against the NgL average does not introduce any bias. However, if some pitchers are following this usage pattern, and others are following a standard (MLB) starter usage pattern, then the hybrid types are deriving a systematic benefit to their rate of production compared to the league. Essentially, this is why reliever WAR has a downward adjustment (because relievers as a group have an easier job to do and therefore as a group perform more effectively on a rate basis). We may need to introduce a "reliever penalty" as WAR does for pitchers with that usage, to reflect the performance boost of shorter outings. The penalty would ideally be proportional to the average number of innings/start if it is below some threshold. That would be my advice. The second way this might impact the MLE is on the batting runs (since these pitchers may not be getting the PAs you assert), although perhaps this effect is small enough to ignore.

Next item: pitcher batting WAR needing adjustment by 35%. I don't have much to say about this in terms of methodology, but it is highly interesting to me in a more philosophical way. Namely, doesn't this tell us something about QoP in general in NgL? Perhaps that the pitching talent in general was not at MLB standard? If NgL pitchers as a group are 35% better than MLB counterparts in terms of relative performance at bat, does this tell us nothing about whether hitters' stats should be adjusted, or about the strength of pitching in NgL overall? I'm inclined to believe so, but I don't expect you or others to follow all my inclinations. But it's something to think about.
   184. kcgard2 Posted: March 30, 2019 at 10:04 AM (#5826996)
Continuing on my thoughts from #180, theoretically the point should be true, but this is also easy to test.

I created 3 fictional players who played 100 games, and used a simple indicator of whether the player had a SB in each game. Player A had steals in 10% of his games (randomly assigned the indicator for 10% of his 100 games), player B in 30%, player C in 60% of his games. Then I randomly removed 35% of the records for each player, to simulate these players having missing box score data for steals, so we have data for 65 games at random out of the 100 that each played.

Player A ended up with 7% steals in "recorded" games, player B 26%, player C 62%. If you repeat this many times, what you'll find is that player A will on average come out to 10% of course, player B 30% and player C 60%. So that if we measure a player by steals above league average, we should not adjust that percentage due to missing games, because the percentage is accurate (on average). You would, of course, in a value metric, adjust the raw value, because you know that 35% of the data is missing. But we already pad those missing games by virtue of the playing time estimates. I.e., if we know he had steals in 40 games out of 65, we would pro-rate that to 60 steals in 100 games (the 60% rate from his missing AND non-missing data), but not pad the rate to 80% which would give him credit for steals in 80 games out of 100.

Another reason you can know this is wrong, is to ask, are we adjusting any players down in rate? If not, then padding steal% for select players simply has the effect of making the entire league look like it stole bases at a higher rate than it did. If players above some threshold are adjusted up, then players below the threshold would need to be proportionally adjusted down to keep the league steal rate appropriate. But this isn't even possible once you think about it. There are plenty of players with rates at or near zero - you can't adjust them downward. And besides that, what should the threshold even be?

You may want to reconsider this adjustment to baserunning.
Page 2 of 2 pages  < 1 2

You must be Registered and Logged In to post comments.



<< Back to main

BBTF Partner

Dynasty League Baseball

Support BBTF


Thanks to
for his generous support.


You must be logged in to view your Bookmarks.


Page rendered in 1.1152 seconds
41 querie(s) executed