Prelim

1) Walter Johnson - also my #1 pitcher all-time
2) Pete Alexander
3) Cy Young
4) Joe Williams - top 4 are in my top 10 all-time
5) Christy Mathewson
6) Kid Nichols - pretty clear top 6
7) Dick Redding
8) Amos Rusie
9) Jose Mendez
10) Eddie Plank
11) Stan Coveleski
12) Ed Walsh
-- Urban Shocker
13) Rube Foster (needs re-evaluation)
14) Eppa Rixey
15) Rube Waddell
16) Clark Griffith
17) Red Faber
-- Wilbur Cooper
-- Babe Adams
18) Mordecai Brown - marginal pick
-- Dolf Luque, Vic Willis
19) Joe McGinnity - not PHoM

Pitcher Name - MMP Points

Cy Young 134.25
Walter Johnson 124.19
Christy Mathewson 78.13
Pete Alexander 72.38
Kid Nichols 70.92
Ed Walsh 51.13
Joe Williams 45.91
Joe McGinnity 40.54
Amos Rusie 39.83
Rube Waddell 37.78
Vic Willis 31.19
Ted Breitenstein 29.10
Mordecai Brown 25.54
Eddie Cicotte 24.29
Red Faber 22.29
Stan Coveleski 21.34
Noodles Hahn 22.06
Rube Foster 21.28
Dolf Luque 19.88
Clark Griffith 18.07
George Cuppy 17.04
Pink Hawley 15.13
Dick Redding 14.45
Jose Mendez 14.25
Jack Chesbro 13.67
Jouett Meekin 13.50
Cy Seymour 13.30
Babe Adams 8.70
Urban Shocker 7.71
Eddie Plank 7.53
Eppa Rixey 1.88
Wilbur Cooper 1.43

DL, what's the plan for pre-1893 pitchers?

Preliminary Ranking

1. Walter Johnson
2. Cy Young
3. Pete Alexander
4. Joe Williams (needs review)
5. Kid Nichols
6. Christy Mathewson
7. Eddie Plank
8. Ed Walsh
9. Amos Rusie
10. Stan Coveleski
11. Clark Griffith
12. Jose Mendez (needs review)
(Urban Shocker)
13. Rube Waddell
14. Red Faber
(Vic Willis)
15. Joe McGinnity
16. Mordecai Brown
17. Dick Redding (needs review)
(Babe Adams, Wilbur Cooper)
-in-out line-
(George Uhle, Theodore Breitenstein, Jack Quinn)
18. Eppa Rixey
(Nap Rucker, Hippo Vaughn, Noodles Hahn, Charles Bender, Al Orth, Eddie Cicotte, Doc White, Jeff Pfeffer, Addie Joss, Jesse Tannehill, Jack Powell)
19. Rube Foster (needs review)

This is a group that the HoM electorate has struggled with more than most. After the top 10, there are a lot of players who sat in the backlog for many years before being elected, and some that went in faster are now probably not viewed as among the stronger picks from this group. Several unelected players still draw more than token support as well. Factors in the difficulty are the large gap between Win-Share and WAR assessment of these players, the divide between long-career, low-peak pitchers and short-career high peak pitchers (heightened by uncertainty over the value of innings-eating in this period), the sudden beginning and sudden end of the deadball era, and the very limited data available for pitchers from Black Baseball before the creation of the Negro major leagues.

I need to re-examine the Negro-League pitchers and consider how I am weighing the deadball pitchers against the pitchers of the high-offense stretches at the beginning and end of this period designation.

DL, what's the plan for pre-1893 pitchers?

We can do those last, I suppose. There aren't any active candidates and no changes to the list since the last time we ran it.

I think Jose Mendez was overlooked in MMP voting. I know there was one of his two peak seasons I forgot about him where I definitely would have voted for him.

Prelim:

1. Walter Johnson
2. Cy Young
3. Pete Alexander
4. Smokey Joe Williams
5. Kid Nichols
6. Christy Mathewson
7. Jose Mendez
8. Eddie Plank
9. Amos Rusie
10. Ed Walsh
11. Stan Coveleski
12. Rube Waddell
13. Dick Redding
14. Joe McGinnity
15. Rube Foster
16. Clark Griffith
17. Red Faber
18. Mordecai Brown
19. Eppa Rixey

Initial thoughts - this is a really tough era to separate the borderline-in from borderline-out.

1. Walter Johnson
2. Cy Young
3. Pete Alexander
4. Christy Mathewson
5. Joe Williams - I might push him above Matty, will need to review further
6. Kid Nichols
7. Dick Redding
8. Eddie Plank
9. Ed Walsh
10. Rube Waddell
11. Amos Rusie
-- Babe Adams
12. Jose Mendez
13. Eppa Rixey - Seems liable to be a controversial guy here - personally I think he is solidly over the line, but I can very easily see the case against him.
14. Mordecai Brown
15. Stan Coveleski - I always feel disappointed with his showing, but then I remember his bat.
16. Clark Griffith
17. Red Faber - I don't think he'd quite make my pHOM, but he's close to the cutoff.
-- Eddie Cicotte
-- Wilbur Cooper
-- Urban Shocker
18. Rube Foster - I'm inclined to give him the benefit of the doubt, but his MLEs really don't look great compared to the other guys in this era, and I'm quite high on most of them as is.
-- Vic Willis
-- Noodles Hahn
-- Chief Bender
-- Jack Quinn
19. Joe McGinnity - He rates out as one of the worst overall HoMers for me.


Trivial note - Redding is the only pitcher from this era that we've elected that isn't in Cooperstown.

A quick question to make suer I don't do something stupid: When ranking pitchers whose careers don't fall completely within the time frame, are we supposed to rank the whole career, or just the part that is within the frame? I personally would prefer to rank the whole careers.

Oh, Right. Amos Rusie. Are we supposed to make playing time adjustments for "was owned by a team that was owned by Andrew Freedman?" Rusie would look MUCH better if we could do that.

When ranking pitchers whose careers don't fall completely within the time frame, are we supposed to rank the whole career

Yes

Are we supposed to make playing time adjustments for "was owned by a team that was owned by Andrew Freedman?"

I don't know what you're talking about.

Are we supposed to make playing time adjustments for "was owned by a team that was owned by Andrew Freedman?" Rusie would look MUCH better if we could do that.

Brock, if you are making reference here to Rusie's missing a season due to contract disputes, then the answer would be that you can give him credit for time missed due to a labor dispute, but you are not obligated by Hall of Merit rules to do so. If you want to boost Rusie's on-the-field performance by arguing that he would have pitched better if he hadn't been in conflict with the owner, that would not be allowable. A player's on-the-field performance is what counts, except insofar as circumstances that were unrelated to the conduct of the game and not the player's fault kept a player otherwise able to play from performing on the field. Military service and work stoppages are the two most typical cases for which gaps get filled in. Even in these cases, it's still up to individual voters to decide to give credit or not. The majority have generally done so, but not everyone has.

Individual contract disputes are a grayer area, since that's an individual decision by the player and not a collective decision that individual players have to honor, but in cases like Rusie's or Edd Roush's, some voters have decided to give credit for missed seasons. On the other hand, I don't think anybody has ever given Arky Vaughan credit for the seasons he missed when he retired because he wouldn't play for Leo Durocher anymore. (At least, that's the story of Vaughan's early and temporary retirement that I've heard--there may be other accounts.)

DL, could you explain how you have Cy Young over Grover Alexander? It looks to me like their career rates of WAR/IP are about the same in both BWAR and FWAR, and Young has about 2000 more innings at that rate than Alexander does, which seems like a pretty big gap to close: by BWAR, it's pretty close to the value of Johan Santana's whole career!

I can see some adjustments that would be somewhat favorable to Alexander, but I can't quite see a set of adjustments that would be large enough to close that large of a gap in value between the players.

What are you seeing here?

I think you are asking DL why he has Pete Alexander over Cy Young.

Chris - You've got my question right, although I extend it past the 1896 season, when Rusie held out all year in a contract dispute with Freedman. Rusie retired from baseball after the 1898 season. He was 27 years old. He had shown no signs of injury. He retired because he could not stand working for Freedman any more. He stayed out of the game for two years, and then went to Cincy for a VERY short time, because he was awful. His arm had not survived two years of non-use. I'd like to give him some credit fo 1899 and 1900, although I'm not sure how much. I just wanted to know if that was allowed.

although I'm not sure how much. </i>

It's well documented that Rusie suffered what would be a career-ending arm injury late in the 1898 season. Here's the account, with quotes from Rusie, from his SABR biography:

In 1898 Rusie suffered an injury that ended his effectiveness as a pitcher. He was pitching against the Chicago Cubs one day in August. Chicago’s speedy outfielder Bill Lange was on first base. Lange had led the National League in stolen bases the previous year. Although his productivity was down in 1898, due to physical and attitude problems, he was still a threat on the basepaths.20 Rusie resolved to pick the speedster off base. Instead of taking the usual step when throwing to a base, Rusie made a quick throw to first base without moving his feet. Something snapped in his shoulder. He got his out, but never regained his fastball. “My arm felt dead,” Rusie said. “I finished the game throwing floating curves. The following day saw the start of a parade of doctors. Each examined my arm. Each had a different diagnosis The x-ray was unknown then, so their job wasn’t an easy one.”21 (https://sabr.org/bioproj/person/amos-rusie/)

Credit for 1896 is reasonable, but it was injury, not contract disputes, that ended Rusie's effective career.

I think you are asking DL why he has Pete Alexander over Cy Young.

Yes! Thanks for correcting my reversal of the names.

1. Cy Young
2. Walter Johnson - I have Young and Johnson as #7 and #8 all time players

3. Pete Alexnder
4. Christy Mathewson
5. Joe Williams
6. Kid Nichols

-- big ol' huge gap

7. Ed Walsh
8. Rube Waddell
9. Eddie Plank
10. Stan Coveleski
11. José Méndez

12. Amos Rusie
13. Red Faber
-- Eddie Cicotte
14. Mordecai Brown
-- Babe Adams
15. Eppa Rixey
16. Dick Redding

-- Jack Quinn
17. Clark Griffith
18. Joe McGinnity - I have back-to-back with Clark Griffith
-- José Muñoz
-- Urban Shocker
-- Noodles Hahn
-- Vic Willis
-- Hippo Vaughn
19. Rube Foster - have pHOM, but in a re-do I wouldn't

Chris - THANK YOU for the info. I don't believe that, in all my reading about Rusie, I have never run across mention of the injury - or at least, a mention of it that I remembered. And that, of course, solves the problem. There is no credit due to Rusie for 1899 or 1900. THANKS again!

I am less of a Christy Mathewson fan than everyone here, apparently, and more of aRube Foster fan. My top six would be:

Johnson
Young
Alexander
Williams
Nichols
Foster

Some of you may know a lot more about Rube Foster than I do. I am willing to be educated.

Foster is probably a good figure to discuss. The recent Major League Estimates have been quite unkind.


What we have of his career takes a major downturn starting in 1913 by just about any standard. He was transitioning into managing at this point and was likely hampered by his leg injury in 1909, so this downturn isn't surprising or anything, but it really keeps him from building up a solid career case - I don't think he provides any signficant value for those last 5 seasons.

So we are stuck with 11 years where he was a viable major league contributor: 1902-1912. And 1909 is definitely a truncated year since he broke his leg that season. The problem then is that this is weak competition (no organized league, or even anything coming close) and and we have quite limited information - for many of these seasons we have <5 games. His reputation is really strong, but I have to imagine at least some of that is clouded by his immense talents as a manager and organizer of the game.

There are some good examples of HoM pitchers in this period who got a lot out of shorter careers - Amos Rusie, Ed Walsh, Stan Coveleski, Joe McGinnity - that's who his immediately comparable cohort contains. It seems unlikely to me that his peak is higher than Walsh or Rusie, who look to me to be the strongest of the group - after that, I don't think any interpretation is unreasonable. I tend to think Wilbur Cooper is a very nice point of comparison - very similar career arcs. I like Cooper, I don't think he's far at all from being a viable HoM candidate. But he's definitely be towards the bottom of this group, which is why I rate out Foster so low.

I think Jaack’s analysis hits the key features of Foster’s career that are guiding the lower placements of him.

I’ll add a few more notes. The first is purely informational. The first is that the fullest statistical record for Foster and other pre-1920 Negro League players is available at

www.seamheads.com/NegroLgs/index.php

At Seamheads, one can obstain player records and team and “league” data for all seasons in Black Baseball from its beginnings through 1948. Baseball Reference and Fangraphs have the same foundational data for the Negro Leagues now recognized as major, but that data only covers the period of 1920-48, and the major independent teams during that period are missing from its data but are included in the Seamheads data.

It's the data at Seamheads that indicates the career shape that Jaack has described. From 1902 up until his injury in 1909, Foster has a reasonable case to be the #1 starter on his team for most seasons, but after 1909, his number of starts and innings declines into the 2-4 range through 1914, and his ERA+ numbers worsen, including a couple of really poor years in this later period.

Based on this data, I can only project him to a major-league equivalent career of 3300-3700 innings and a level of effectiveness that aligns, at best, with the lower-tier Hall-of-Merit pitchers form that period: Waddell, McGinnity, and Brown. I’d agree with Jaack that Walsh and Rusie are Foster's highest upside, but I don’t think that’s likely: the statistical record, such as it is, doesn’t suggest that kind of dominance. His reputation might, but even with his reputation accepted, his career is not of a scale to put him into the top tier of pitchers from this period. Those pitchers all have more than a decade of dominance. Foster has at best about eight years at that level.

Because his peak came in the 1902-09 period, which is much less well documented that his decline years, it truly is hard to say just how good he was. But his two seasons in the Cuban Winter League, which are the best documented stretches of his career in this period and probably the strongest league he played in, are less impressive than his documented play against Black Baseball teams in the U.S.

1. Walter Johnson
2. Cy Young

3. Joe Williams
4. Pete Alexander
5. Kid Nichols
6. Christy Mathewson

7. Eddie Plank
8. Dick Redding
9. Ed Walsh

10. Stan Coveleski
11. Mordecai Brown - large strength of schedule boost
12. Amos Rusie
(Urban Shocker)
13. Rube Waddell
14. Jose Mendez
15. Joe McGinnity
(Babe Adams)
(Vic Willis)
-in-out line-
(Eddie Cicotte)
16. Red Faber
17. Clark Griffith
(Charles "Chief" Bender, Theodore Breitenstein, Wilbur Cooper, Noodles Hahn, Carl Mays, Jose Munoz, Jack Quinn, Nap Rucker, George Uhle, Hippo Vaughn)
18. Eppa Rixey

19. Rube Foster

I took a look at Rube Foster on the new (to me) database that actually have stats for his early years. I realized that my impression of Foster was heavily weighted by reading Only the Ball was White, which was published in 1970, and had little access to stats. The impression I gained from that book was that Foster was more or less like Al Spalding. Spalding was probably the best pitcher in the National Association every year of its existence, and also the best in the first year of the National League. Given the importance of individual pitchers in that game, Spalding was almost certainly the MVP every year, as well as the Cy Young. Then he retired to become a groundbreaking team owner and a sporting goods magnate. He was 25 years old in his last good year. I rate Spalding highly because I think that his contributions to the game from 1976 on are like war credit - their importance is so great that they overwhelm any level of playing performance. Doing them was certainly "under his control", but the need for them to be done was too obvious to ignore, if you had Spalding's ability.

I had thought of Foster in the same way. Then i took a look at a modern site, and found out that Foster was 40 years old when he put that Negro League together. That's completely different. His career as a star pitcher was over. There's no credit to give. And that's why I've just revised my opinion of him way down. I'd just never known the age data before. Thanks to everyone who pointed out the weaknesses in his resume.

Definitely subject to tweak this upon further review.

Trying to rank Waddell, Brown, Griffith, Rusie, Faber and McGinnity in a way that was at all internally consistent had me twisted in knots, and I wound up leaning heavily on WAR/162.

I think it's worth noting, since no one else has yet, that Mathewson, Alexander, Coveleski, Brown, Faber and McGinnity all had huge World Series performances that helped their team win championships. More than a few of them, I imagine, would have been easy World Series MVPs had the award existed at the time. Of course, accounting for that is a bit complicated given that many of the other pitchers in this exercise played before the World Series existed, but it's certainly part of their stories.

Anyway:


1. Walter Johnson
2. Cy Young

3. Pete Alexander
4. Christy Mathewson
5. Joe Williams
6. Kid Nichols

7. Eddie Plank

8. Ed Walsh
9. Stan Coveleski
10. Dick Redding
11. Jose Mendez
12. Rube Waddell
13. Mordecai Brown
14. Clark Griffith
15. Amos Rusie
16. Red Faber
17. Joe McGinnity

18. Eppa Rixey
19. Rube Foster

I've been toying with a modified version of Bill James's starting pitcher rankings for a while now; I have them processed from 1901 (as far back as Retrosheet has box scores) through 2011. The method has imperfect applicability to this group, since it's MLB pitchers only and isn't really reliable until 1902. It also doesn't give any credit for hitting or relief work, which affects most of these pitchers to varying extents (particularly Walsh and Brown). But for the pitchers who I have complete careers for in this group, here's how they shake out:

All-timers
Walter Johnson (#2 all time) - Second all-time in days at #1 (1138) AND second all-time in days at #2 (815), which is rather remarkable. Can rank ahead of Clemens depending on how you break things down. Six year-end #1 finishes ('12, '13, '14, '16, '18, '19), tied for the most ever. Year-end scores in 1913 and 1912 were the second- and third-highest ever to this point.

Grover Cleveland Alexander (7) - Year end #1 in 1915 and '17. Over 500 days at #1 despite both competing directly with Johnson and missing a peak season due to the war. This is backed up with significant amounts of time at every position from 2 through 6. Depending on how you break down the numbers and adjust for league size, he can finish as high as third in my current sample.

Christy Mathewson (15) - Year end #1 in 1905 (and with the fourth-highest year-end score ever). Surprisingly had "only" 155 days at #1, quite low for a pitcher of his caliber - but over 1200 combined days in second and third.

(Cy Young just from 1902 on ranks at #22; with his whole career, I suspect he would push Johnson for the top spot in the group. Finished #1 each year from 1901-04 and again in '07.)

Solid HOMers
Eddie Plank (31) - I expected him to contend for "best pitcher never to reach #1," but he actually did grab the top spot for 36 days in mid-1906 (very tumultuous year for #1, as Young and Mathewson both had bad seasons). Predictably steady; finished in the top 10 thirteen times.

Ed Walsh (46) - Extremely impressive peak; has two top-10 all-time year-end scores (1908 and '10), and joins Young and Johnson as the only pitchers in this group to hold down #1 for an entire season (1911). 478 days at #1 is currently a top-20 all time total, pending the last decade of data. (Is right on the edge of high-borderline; I gave him the upward nudge for his extensive relief work in peak seasons.)

High borderline
Stan Coveleski (54) - Year end #1 in 1920, edging out Alexander in a great race (Coveleski actually made the final pass during the World Series). Five other top-10 finishes.

Low borderline
Mordecai Brown (62) - Year end #1 in 1906 and '09. Relatively short career by HOM standards. Had a lot of relief work, so he probably deserves higher than this.

Red Faber (64) - Year end #1 in 1921 and '22, but only one other top-10 finish. Spent a lot of time being good-not-great; he leads this group by a wide margin in time spent as a #2 starter (between ranks 17-32 during this period, or 25-48 in 1915).

Eppa Rixey (68) - Only HOM pitcher in this group not to reach #1, but did reach #2 in three separate seasons ('23-'25) and finished in the top 10 six times. Not sure I'd vote for him, but consistent starters were hard to come by in the '20s. (Which will play heavily into how the rankings see the next cohort of pitchers.)

Among the pitchers for whom my dataset captures partial careers, I have most of Rube Waddell and Joe McGinnity's work. I would guess that Waddell would end up in the low-borderline or high-borderline group (he's just shy of that as is, but it takes a couple years for pitchers to establish themselves in the rankings and 1901 is his first 30-start season). McGinnity is harder to get a read on because more of his bulk predates the rankings, but I would tentatively predict low borderline for him, or just below that. (A lot of McGinnity's value - and several other pitchers in this group, for that matter - comes from making huge numbers of starts per season. That helps in the rankings to some extent, but not as much as it would in a WAR-type system.)

Highest-rated pitcher in this period who's not in the HOM is Addie Joss, who finishes behind Rixey. Babe Adams isn't too far behind Joss and could be in with minor league/war credit for his mid-career gap.

Eric J, what is the system behind these rankings? What is it seeking to measure, exactly? The link goes to the current rankings, but without signing up for the site, it doesn't seem possible to get more information about what the numbers mean or where they come from. Any additional background would be appreciated!

The explanation for James's original system can be found here. My numbers are built similarly (it's a rolling weighted average of Game Score), but there are a number of adjustments between James's system and mine that I can expound on if there's interest. Among them are the park adjustment (which James does rather lazily), the adjustment for missed time, and the version of Game Score used (older box scores don't have reliable earned run totals, so I use the Fangraphs version). I have day-to-day rankings for every season from 1901-2011 (working on data entry for '12 at the moment), hoping to be at least caught up enough to have numbers available for all eligible pitchers before election time.

Once the rankings are calculated, I have a number of ways to break them down, some of which are mentioned above (days at #1, year-end #1 finishes, time spent in the top 10, etc.) My best attempt at a comprehensive metric assigns a value to every day a pitcher spends at any spot in the top 2N in the rankings (N being the number of teams in the league). In a 16-team league, the pitcher gets (1/sqrt(rank)) points per day at a given ranking, so a day at #1 gives 1 point, #4 gives half a point, #16 gives a quarter of a point. In larger leagues, this is scaled to keep #1 at 1 point and #N at a quarter of a point. The intent is to capture "how much of an ace" the pitcher was during his career. (I also have this split out for just the top N pitchers, and for the top 32 without scaling for league size. But I like the top 2N version best.)

It's not an intensely rigorous metric - but then, I'm not particularly blown away by any of the commonly available pitching WAR systems either, and I do think that there's something to be said for comparing pitchers to their exact contemporaries in an ordinal sense rather than comparing cross-era WAR totals to each other without accounting for things like expansion and integration.

I'm realizing now that I technically don't have all of Mathewson's career either. If you add in what he would have accrued in 1901, he would likely move up one spot in the overall rankings, though not nearly enough to catch Alexander. (Matty's '01 is very good, but it was effectively his rookie year and it takes a little while for a new pitcher to establish himself in the rankings.)

1. Walter Johnson PHOM 1933
2. Cy Young PHOM 1917
3. Pete Alexander PHOM 1936
4. Kid Nichols PHOM 1912
5. Joe Williams PHOM 1938
6. Christy Mathewson PHOM 1922
7. Eddie Plank PHOM 1924
8. Dick Redding PHOM 1942
9. Ed Walsh PHOM 1925
---Babe Adams PHOM 1933---
10. Jose Mendez PHOM 1932
11. Rube Waddell PHOM 1916
12. Stan Coveleski PHOM 1937
---Urban Shocker PHOM 1937---
13. Amos Rusie PHOM 1908
14. Clark Griffith PHOM 1921
---Eddie Cicotte PHOM 1927---
15. Mordecai Brown PHOM 1929
---Vic Willis PHOM 1919---
16. Red Faber
17. Joe McGinnity
18. Eppa Rixey
19. Rube Foster

v1.4 MLEs

Pitcher Name pWAA pWAR WAR
Joe Williams 56.6 107.6 116.2
Dick Redding 40.7 83.3 88.5
Jose Mendez_ 34.9 61.4 62.4
Rube Foster_ 8.1 34.8 40.6

As noted in post # 31 above, Dr C’s latest MLEs for Dick Redding show a career total of 88.5 WAR. I use my own back-of-the-envelope to check on some of the MLEs, and I have a problem with this one. For Redding’s prime seasons from age 21 to 32, Dr C shows an MLE of 55.8 WAR, which looks okay to me. But for Redding’s last 9 seasons (ages 33 to 41, or 1923 to 1931), Dr. C reports an MLE of 32.7 WAR (3.6 WAR per season), which seems much too high to me.

For 1911–22, Redding’s raw ERA (not MLE) against NgL teams was 2.60 and his ERA+ was 140. In contrast, while Redding pitched for the Brooklyn Royal Giants from 1923 to 1931, his raw ERA (not MLE) was 4.32 and his ERA+ was 106.

Looking at other indicators of quality, on NgL teams the perceived best pitcher almost always pitched the most innings. Every season during 1911–22 except 1913 (when he appears to have been injured) and 1918 (when he entered the army and spent most of the season in France) Redding either led his team in IP or came very close. For 1923-31, Redding led his team in IP in 1923, but in subsequent seasons he never led his team, usually ranking third or fourth.

From 1911–22, most of Reddings’ teams had winning records, and they included some of the best teams of the era. From 1923–31, the Royal Giants never had a winning record, and got steadily worse as time went on. By the last two seasons of the ECL, 1926 and 1927, they were in last place (of teams that finished the season) with records of 11–24 and 11–23. From 1928 to 1931 they were an independent team, and not a good one, with a 9–23 record against NgL teams over the 4-year period.

For my back-of-the-envelope MLEs, I like to group several seasons together, which counteracts the small-sample-size problem of NgL data and I think helps give me a better picture of trends in player performance over the stages of a player’s career. Let’s start with the first 5 years (1923–27) when Brooklyn was in the ECL.

My back-of-the-envelope MLEs are based on on two factors, an estimate of MLE ERA+ and an estimate of MLE IP. To account for differences in quality between NgL and the AL/NL, I generally multiply a pitcher’s ERA+ by 0.85 (I believe this factor was also used by Chris in his MLEs). For independent teams that were not part of the recognized major NgL’s, I use a factor of 0.80.

Estimating MLE IP is tricky. The raw IP numbers for the NgLs are dominated by differences in team schedules and in the number of box scores available, so I don’t find them to be useful on their own. Instead, I look at relative IP within a team. Because NgL teams had such small rosters, using a pitcher’s percentage of his team’s IP would result in implausibly high estimates of IP (such as 400+ IP). What I do instead is prorate a NgL team’s top pitcher in IP to the MLB median “ace” (median of the top 16 pitchers in the AL/NL each season), which I rounded to 300 IP for 1911–22 and 270 IP for 1923–31. For the other pitchers, I prorate the IP to the ratio of their IP to the team’s top pitcher. For example, in 1925 Redding pitched 43.3 innings, and Brooklyn’s top pitcher was Bill Holland with 76.7 IP. The ratio is 56.5%, which is multiplied by 270 to give an MLE IP of 153 innings.

Here are my MLE estimates.

1923–27
Unadj. IP: 247.7
MLE IP: 866
Unadj. ERA: 4.43
Unadj. ERA+: 104
MLE ERA+: 88
WAR: 4.2 (0.8 per season on average)

For 1928–31, Brooklyn was an independent team, and the number of available box scores goes way down. I believe that for the entire team only 31 games are recorded over the 4 seasons, and for Redding we only have 5 games (40 IP). But he did pitch better on average in those 5 games than he had in the
1923–27 period.

1928–31
Unadj. IP: 40
MLE IP: 475
Unadj. ERA: 3.60
Unadj. ERA+: 126
MLE ERA+: 101
WAR: 6.1 (1.5 per season on average).

So, my back-of-the-envelope gives Redding 10.3 WAR for his last 9 seasons, versus 32.7 by Dr. C. And my intuition tells me that 10.3 is probably too high—the sample size for the modest uptick in performance in his last 4 seasons is terribly small. And realistically, how many 38-year-old pitchers are given the chance to keep pitching after 5 seasons of near replacement-rate performance?
But my main point is that I think Dr. C’s MLE estimates for these seasons are pretty clearly inflated.

Thanks, Brent, for this analysis. I agree that arriving at a reasonable estimate of MLE innings pitched is hard, and that Dr. C's MLE IP estimates are too high.

That said, I think the evidence of your own projections, which use an IP estimation method that is actually pretty similar to what Dr. C uses, indicates that the discrepancy between his evaluation of Redding and yours is not due to IP estimates but to estimates of league quality of play and, perhaps, Redding's relative quality.

Here's a direct comparison between the two sets of IP estimates for Redding:

1923-27
Brent: 866
Dr. Chaleeko: 850

1928-31
Brent: 475
Dr. Chaleeko: 560

That difference is obviously not producing the significant WAR differences that you've documented: it's a quality of play/quality of player issue. Dr. Chaleeko's quality-of-play spreadsheet indicates that the conversion factor he is using for the 1923-31 period is pretty steadily around .9, which is significantly higher than your .85 or .80. I don't have a specific enough knowledge of Dr. Chaleeko's whole MLE system off the top of my head to be able to say whether that difference in conversion factors is sufficient to account for the whole difference between the WAR estimates or whether there are other modeling differences involved as well.

We can safely say, though, that IP estimates are not the source of the difference.

Chris, I agree that the innings pitched estimates are not the source of the difference. However, it's also clear that the difference in conversion factors between his .9 and my .85 or .8 can't explain most of the difference either. If I use a .9 conversion factor, I get an estimated MLE WAR from 1923-31 of 16.2, which falls well short of his 32.7. Indeed, even a conversion factor of 1.0 gets me only to 23.7, still short of his 32.7.

My guess is that Dr C's regression/smoothing method for producing annual estimates may lead to his estimates for this period "borrowing" from Redding's better seasons prior to 1923. And I would also guess that this is not symmetric--so that Redding's MLEs for 1919-22 are not being significantly dampened by his decline in performance post-1923. Hopefully, Eric can take a look and explain.

While I understand that some smoothing is necessary to produce reasonable season-by-season estimates, I do worry that there's a risk that it could distort actual changes in player performance, especially toward the end of a player's career. So many of the NgL players have MLEs showing them playing at above-average levels into their late 30s and early 40s, and I wonder if that may sometimes be the result of smoothing methods applied to small samples.

Brent, I think that the smoothing may be having an effect, but it's still not the main cause of the difference. If I am correctly following the process Dr. C's blog post on his pitching MLE process, it indicates that the smoothing effect kicks in for pitchers only when their IP fall below 45 documented IP for a season.

This component of the process should affect Redding's numbers very little in the 1923-27 period, as his documented innings are above 45 for 1923 & 24 and are either 42.3 or 43.3 for each of the 1925-27 seasons, so the smoothing will be small for those years. It's only after 1928 that the smoothing would have a major impact.

The discrepancy between your estimates and Dr. C's are much larger for 1923-27 than they are for 1928-31.

1923-27
Brent: 4.2 WAR
Dr. C: 19.6 WAR

1928-31
Brent: 6.1 WAR
Dr. C 11.1 WAR

Moreover, about 30% of the difference between the MLEs for 1928-31 comes from Dr. C's larger IP estimate, while the vast 15.4 WAR difference for 1923-27 is virtually unaffected by IP differences.

So, neither IP estimates nor smoothing procedures seem in any way to account for a difference in estimates of fully 3 WAR/season from 1923-27, and it looks like changing Brent's quality-of-play adjustment for those seasons only accounts for 2.4 of the 15.4 WAR difference. 2.6 WAR/season remains unexplained.

One place to look for further explanations would be the possible gap between an ERA+ based estimate and an RA/9 with fielding adjustment estimate. I don't have time at the moment to run numbers in depth, but just for 1924, Redding had 32 RA, of which 29 were earned, so 90% of his runs allowed were earned. For the league as a whole only 77% of runs allowed were earned. There's no fielding data for this NeL season, so there's no fielding adjustment to Redding's RA/9 at all. His ERA+ is 90 (with park factors that I can't see baked in), but his RA+ is 109 (after applying a park factor estimate). Using these ERA+ and RA+ numbers, I find that in Redding's 1924 league environment, using Dr. C MLE innings estimate of 200 IP for the season, using RA+ rather than ERA+ to project Redding's runs allowed leads to a swing of 17 runs for Redding relative to league average, from 10.5 runs allowed above average to 6.9 runs allowed below average. Those are rough calculations, but it looks like an RA-based projection of Redding for 1924 would see him as being at least 1.5 wins better than an ERA-based assessment.

In reaching an understanding of the differences between Brent's numbers and Dr. C's, I would now propose that a possible significant contributing factor is the choice of ERA vs. RA in the calculations.

Chris,

I ran the RA numbers for 1923–27, and it closes a little bit of the gap between my estimates and Dr C's, but still leaves a large difference.

The following are Seamheads data (not MLEs) or my calculations from those numbers

. . . . . League . . . . . . . . Redding
Year . RA . . ERA . Ratio . RA . . ERA . RA+ . ERA+
1923. 5.67 . 4.19 . 74% . 5.50 . 4.33 . 103 . 97
1924. 5.47 . 4.24 . 77% . 5.65 . 5.12 . 105 . 90
1925. 5.63 . 5.03 . 89% . 5.19 . 4.57 . 114 . 116
1926. 5.51 . 4.35 . 79% . 3.83 . 2.98 . 143 . 145
1927. 5.21 . 3.97 . 76% . 7.02 . 5.10 . 91 . 95
Total. 5.49 . 4.36 . 79% . 5.45 . 4.43 . 108 . 104

So, over the 1923–27 period, Redding's RA+ is 4 points higher than his ERA+.

How far are we toward reconciling my estimates with Dr C's for this period?

First, as I was checking against Dr C's spreadsheet, I see that there is actually a bigger discrepancy in IP than you reported. (I think you must have missed one season.) For 1923-27, Dr C assumed 1,050 IP versus the 866 in my first estimate in #32. We also have the effects of using different conversion factors (0.85 for me, vs. 0.90 for Dr C) and using RA+ rather than ERA+.

My original estimate (IP=866, CF=0.85, ERA+) WAR = 4.2
Switching to Dr C's IP = 1050, WAR = 5.1
Adding a CF=0.90, WAR = 8.7
Adding RA+ rather than ERA+, WAR = 11.2

These cumulative adjustments still leave us well short of Dr C's pitching WAR estimate of 19.6 (or his total WAR estimate of 20.9), so there is still a mystery. But identifying these differences in assumptions and methods does move us partway toward explaining the difference.

Brent, thank you for catching my IP error (if I was off by 200, that would almost certainly be from leaving out a season) and for quantifying the impact of each of the differences between your MLEs and Dr. C's.

I've been reviewing Dr. Chaleeko's full methodology for pitching MLEs, and I suspect that the remaining difference is probably arising from the complex steps the process includes for moving between league contexts: using z-scores to adjust for standard deviation and calculating artificial mean league RA9s in connection with calculating the z-scores. I confess that I am not fully able to follow the rationale for all of the steps. The use of derived mean league RA9 values after discarding extreme players before calculating standard deviation is particularly puzzling to me, and I think it could have unpredictable and possibly unjustified inflationary or deflationary effects on players' values. These are steps 3-7 in the full pitching MLE process as described at

https://homemlb.wordpress.com/2021/07/22/mle-pitching-method-full-detail/

Anyway, my guess is that the remaining divergence between the two MLEs happens when Dick Redding's 1923-27 RA9 values are put through this part of Dr. C's process. I don't have the data or the statistical know-how to trace that out, but someone with the requisite data and know-how could take a look. If it turns out that these steps are indeed the source of the remaining divergence, then we would have the clarity needed to assess the appropriateness of each step in the process.

At this point, my own view is that the IP estimate in Dr. C's MLEs for these seasons is probably too high and that this problem can be rigorously demonstrated. His conversion factor is probably higher than I think is warranted, but that's a much harder point to establish with real rigor. In general, I think Dr. C's use RA+ with a fielding adjustment is preferable to using ERA+ as a measure of a pitcher's quality, but when there's no fielding adjustment to be had, I'm less sure. And the mysterious elements are still mysterious, of course.

So I think there's a bit further to go to have a full assessment. What we've learned so far provides evidence that convinces me of an inflation problem, but the magnitude of that problem is still not fully in view for me.

Ballot is up. Will try to end the election Wednesday 10/4 @ 4:00 eastern