hot topics

1921 Oscar Charleston
NNL St. Louis Giants

*-led league
G-64 (team 64)
AB-225
H-98
D-12
T-10
HR-12
R-88*
W-34 (2nd in league)
HP-5
SH-8
SB-32* (tied for lead with teammate Joe Hewitt)
AVE-.436* (NeL .263)
OBA-.519* (NeL .324)
SLG-.738* (NeL .361)

Remember, this season was played in Giants' Park, which was *not* the same park as the later hitters' paradise, Stars' Park. This one had raw park factors of 68 and 102 in 1920 and 1921, respectively.

From Patrick Rock:
1923 Oscar Charleston
Indianapolis ABCs

*-led league
G-84 (team 84)
AB-308
H-112 (4th)
D-25 (2nd)
T-6
HR-11
R-68
RBI-94 (2nd)
W-49*
SB-25*
TB-182 (4th)
AVE-.364 (4th)
OBA-.443 (3rd)
SLG-.591 (3rd)

1928 Oscar Charleston
Hilldale Club

Batting
G-62 (team 62)
AB-219
H-74
D-8
T-5
HR-10
R-58
W-34 (2nd)
HP-0
SF-2
SH-2
SB-12
AVE-.338 (eNeL .282)
OBA-.424 (3rd; eNeL .333)
SLG-.557 (4th; eNeL .383)

Fielding-cf
G-56
DI-492
PO-115
A-4
E-5
DP-1
RF-2.18 (eNeL cf 2.52)
FPCT-.960 (eNeL cf .965)

Hilldale cfs were responsible for 38.2% of the team's outfield putouts; the eastern average was 41.3%.

Charleston also played 7 games at first base, all at the end of the season, fielding .979 (eastern average .979).

Gary,

Given Charleston's 1928 defensive statistics, which are average at best for the Eastern League, is his great defensive reputation based in facts? I admit ignorance.

Well, in reality this is a small sample (one 62-game season). I get all excited about being able to do this stuff, but we really need 2-3-4 consecutive seasons to have more confidence in the results.

This is also very near the point in Charleston's career when he moves permanently to first base. It's probably a good bet he was deteriorating defensively--if you've seen pictures of him over his career, it's not hard to guess why (he begins to look a little chubby during his Harrisburg days).

That said, I do suspect (but only suspect) that he may be somewhat overrated as a defensive centerfielder. It's true that Torriente was regarded as better in the late 1910s and early 1920s--when they played side by side on the American Giants, I believe Charleston moved to left (though now that I think of it, I'll check some box scores to make sure that's true). At least one writer's all-star team I remember from the early 1920s also put Torriente in center, Charleston in left.

But that doesn't mean Charleston wasn't very good, at the least. Now I think I'm going to have to compile those 1921 fielding stats, just to have another set of data...

I haven't updated this with Gary's data, but here were my career MLE's for Charleston:

AB-11726
H-3697
2B-710
3B-226
HR-761 (!)
BB-1385
SO-2271
SB-742 (!)
CS-304
R-2706
RBI-2533
AVE-.315
OBP-.388
SLG-.609
OPS-.997
OPS+ - 179

raw park factors of 68 and 102 in 1920 and 1921

! were some literal bandboxes abandoned as league ballparks?

By the way, Pete Palmer uses 3-year park factors for mlb in order to increase the sample size, because he considers a 66 to 81-game home season (since the mid 1880s) too short. KJOK suggests 3-year park factors for the Negro Leagues, because of their much shorter home seasons, but not for the white majors.

Yeah, clearly we need at least 3 seasons for real NeL park factors. In fact, I'd say five wouldn't be out of the question (allowing for park changes, alterations, etc., when known).

The change in St. Louis's park factors from 1920 to 1921 probably has as much to do with sample size as anything. 1920's only based on 16 home games, after all (1921 is 45)--it's one of the more poorly-represented NNL parks in my stats for 1920. Much better are places like Indianapolis's Washington Park (96 games between 1920 and 21), KC's Association Park (98 games), and Detroit's Mack Park (94).

By the way, the only NNL park that departed the scene after 1920 was Dayton's Westwood Field (about which I know nothing)--the other five were still in the league in 1921, while Redland Field and Columbus's Neil Park were added. We know Redland wasn't a bandbox, anyway.

Of course he was the first and only Oscar Charleston; he wasn't anyone else. But sometimes we like to talk in terms of players we're more familiar with just to get an image. What am I hearing here?

He was a tremendous, and tremendously feared, power hitter.

He could work the count; he drew plenty of walks.

His batting average was good enough to make him a triple crown threat (including the observation that a HR is a hit, too), but we're more impressed with the power.

He was a centerfielder, and in his prime a good defensive centerfielder. But he wasn't really in that Speaker/Mays category on defense. You might consider moving him to LF or RF if you had the right someone for CF. Late in his career, he moved in to play some 1B.

So ... searching for familiar image ... Mickey Mantle, only with a longer career?

I think that Mantle is apt. I would put Charleston behind Mantle though, but better than Speaker and Dimaggio.

Of course he was the first and only Oscar Charleston; he wasn't anyone else. But sometimes we like to talk in terms of players we're more familiar with just to get an image. What am I hearing here?

I think his skill set is similar to Barry Bonds.

from baseballlibrary.com

multi-talented Oscar Charleston was often compared with three great white contemporaries: his hitting and speedy, aggressive baserunning (and hard-sliding style) brought favorable comparison to Ty Cobb;
his physique (he was barrel-chested, with spindly legs), power, and popularity, particularly with youngsters, were reminiscent of Babe Ruth;
and his defensive style and skills, playing a shallow, far-ranging centerfield with a strong, accurate arm and excellent fly ball judgment, brought visions of Tris Speaker.

Charleston had a famous temper, and enjoyed brawling, resulting in legendary encounters with umpires, opponents, agents raiding his teams, a Ku Klux Klansman, and, on one occasion, several Cuban soldiers. As his legs gave out, he moved from centerfield to first base, yet as long as he played, he never lost his home run power, nor his meanness on the basepaths.

One story you see quite a bit is that Charleston played a shallow center field because he had a weak throwing arm--his speed enabled him to make up for this weakness. On the other hand, he did pitch a little in the early part of his career (not a lot; he was never a regular starter or anything), which I'd think is inconsistent with having a *really* weak arm. I don't know of contemporary newspaper sources for this; I think it comes mostly from later interviews and oral histories. Would be interesting to know exactly how many sources there really are.

Going by Holway, he appears to have moved to first base permanently in 1930, at the age of 33. He played there occasionally before that; and in 1922, he played a chunk of the season at first when his manager Ben Taylor was out with an injury. Iirc, the newspapers made it sound like Charleston needed to step in to fill this gaping hole at first, as if first base were more important than center field.

Actually, the Negro League defensive spectrum is pretty interesting...first base, for example, was I think attributed more defensive importance than would be the case today. Defensive liabilities might often be parked in right or left field, rather than first; and I've read several accounts of fans being upset because their first baseman wasn't a fielding whiz--there was fan discontent in Atlantic City in 1928 when George Carr played first, instead of the slick-fielding Chance Cummings.

Third base was a combination of big sluggers (Wilson, Beckwith) and more agile, defensive-minded players (Marcelle, Johnson, Dandridge). Second base, on the other hand, was pretty much dominated by defense--there really weren't any Lajoie/Hornsby types.

Here's Oscar Charleston's data in Holway.

In Negro League Play

1914 No data
1915 .203 for Ind ABCS; 11 2b (2nd), 5 3b (1st), 19 sb (3rd); cf; 1-1 as pitcher median avg. for West teams is 226.
1915 2-5 vs. major-league competition
1915 11-69 for ABCs in Cuba
1916 .152 for Ind ABCs; cf
7-18, .389 in playoff vs. Chi Am Giants
2-3 vs. major-league pitching
1917 .283 for Ind ABCs; ba 2nd, 6 2b (2nd), cf; all-star
1-4 vs. major-league competition
1918 .429 for Ind ABCs (18-42); ba 1st; cf; all-star, MVP, drafted
1919 .356 for Det Stars/Chi Am Giants; ba 2nd, 4 2b (1st), 3 3b (1st); cf; all-star, MVP
2-0 as pitcher
1920 .338 for Ind ABCs; 7 hr (3rd), 18 hr/550 ab (4th); cf; all-star
2-7 vs. major-league competition
16-39 in Cuban Play
1921 .437 for Stl. Giants; ba 2nd, 17 hr (1st), 39 hr/550 (1st), 19 2b (3rd), 13 3b (3rd), 37 sb (1st); cf; all-star, MVP
7-23 vs. major-league competition
7-12 vs. major-league pitchers in California
1922 .391 for Ind ABCs; ba 5th, 20 hr (1st), 50 hr/550 (1st), 29 2b (2nd), 13 3b (2nd), 28 sb (1st); cf; all-star
6-13 vs. major-league competition
41-92 in Cuban Play
1923 .314 for Ind ABCs; 13 sb (1st); lf
.375 and .365 in 2 Cuban short seasons
1924 .342 for Harrisburg; ba 5th, 9 hr (1st), 16 hr/550 (1st), 23 2b (1st); cf; all-star
1925 .416 for Harrisburg; ba 2nd, 19 hr (2nd), 39 hr/550 ab (2nd); cf; all-star
1926 .281 for Harrisburg; 8 hr (5th), 19 hr/550 (4th); cf
1927 .342 for Harrisburg; 12 hr (1st), 24 hr/550 (3rd), 18 2b (2nd); cf; all-star
1928 .380 for Hilldale; ba 3rd, 10 hr (3rd), 26 hr/550 (5th), 9 sb (2nd); cf; all-star (dh)
1929 .370 for Hilldale; 25 2b (2nd), x 3b (4th); cf; all-star
1930 .337 for Homestead; 6 hr (4th); 42 hr/550 (3rd); 1b
1931 .341 for Homestead; 6 hr (3rd), 19 2b (1st), 5 3b (3rd); all-star (dh)
1932 .314 for Pgh Crawfords; 10 hr (1st), 10 hr/550 (3rd), 11 2b (5th), 10 3b (1st), 9 sb (2nd); 1b
1933 .388 for Pgh Crawfords; ba 4th, 22 hr (2nd), 35 hr/550 (4th), 24 2b (1st), 12 3b (1st), 14 sb (1st); 1b; all-star, MVP
1934 .285 for Pgh Crawfords; 12 hr (4th), 21 hr/550 (5th); 1b
1935 .294 for Pgh Crawfords; 9 hr (3rd); 1b
1936 .322 for Pgh Crawfords; 1b
1937 .213 for Pgh Crawfords; 1b
1938 no data; listed as manager for Pgh Crawfords
1939 .250 for Toledo Crawfords; 1b, manager

In Championship Play

1929 6-13 vs. Chi Am Giants in World Series
1930 5-29 in playoff vs. New York
1931 22-43 in World Series vs. Monarchs
1935 10-27 in Playoff vs. Cuban Stars


In Cuba

1924 40-153
1926 61-151
1927 42-120
1928 38-114


Versus Major-League competition

1923 8-18 vs. major-league competition
1928 2-12 vs. major-league competition
1932 9-28 vs. major-league competition
1936 4-10 vs. major-league competition


Career
1689-4972, .340
197 hr, 22 Hr/550 ab
66-200 vs. major-league competition
Black Ink 54, Gray Ink 174

From all I've read, Mantle seems like the best comp. Power, great speed when young, walks, avg, less valuable late in career.

NeL expert Ted Knorr said "Bill James got it right", when putting Oscar as the best NeL player above the more-typical claims of Josh Gibson and Satchel. I'd peg him as probably one of the 20 best players ever, and in a rare display of prudence not speculate any more precisely.

By the way, the only NNL park that departed the scene after 1920 was Dayton's Westwood Field (about which I know nothing)--the other five were still in the league in 1921, while Redland Field and Columbus's Neil Park were added. We know Redland wasn't a bandbox, anyway.

That fits. In order to cause a jump in Park Factors of incumbent ballparks, batters parks (bandboxes must be abandoned and pitchers parks added. Based on 16 home games, no doubt the jump is spurious.

Here’s more information on Charleston’s play in the Cuba, mostly coming from Cuban Baseball: A Statistical History, 1878-1961, by Jorge S. Figueredo.

1915

Charleston was a rookie with the Indianapolis ABCs when they visited Cuba in the fall of 1915 and played 20 games with the Cuban League teams. Their record was 8-12. His teammates included Dick Redding, Ben Taylor, Bingo DeMoss, and George Shively. I do not have Charleston’s statistics for this series.

1920-21

The Bacharach Giants joined the Cuban League. Despite having star players including Charleston, Lundy, Santop, and Redding, by January 13 they were only 4-11 in the 3-team league, when they returned to the U.S. and forfeited their remaining games.

39 AB, 7 R, 16 H, 2 2B, 0 3B, 0 HR, 2 SB, .410

Note: The batting champion was Pelayo Chacón, with .344 in 93 AB.

1922-23

In an attempt to expand from their base in Havana, the league added a club from Santa Clara in the center of the island, and Charleston, Alejandro Oms, Dave Brown, and Bill Holland played for them. Santa Clara was 14-13 when they suddenly pulled out of the league in reaction to a league decision that took away a win, robbing Charleston of an opportunity to qualify for the batting championship. (Bernardo Baró won the championship with .401 in 152 AB.

92 AB, 24 R, 41 H, 7 2B, 3 3B, 0 HR, 7 SB, .436

1923-24

Considered the greatest team in the history of Cuban baseball, the Santa Clara team loaded up on stars, including Charleston, Frank Duncan, Frank Warfield, Oliver Marcelle, Dobie Moore, Alejandro Oms, Pablo Mesa, and pitchers Bill Holland, Rube Currie, Dave Brown, Merven Ryan, and José Méndez. Their team batting average was .331. The season was terminated early when Santa Clara reached a record of 36-11, 11½ games ahead of second-place Habana in the 4-team league.

176 AB, 59 R, 66 H, 9 2B, 5 3B, 3 HR, .375

Charleston led the league in runs scored and stolen bases (31? looks suspicious). Marcelle led the league in batting with .393 and Bienvenido Jiménez led in HR with 4.

1924

After the early termination of the regular season in mid-January, the weakest team (Marianao) was folded and its best players were dispersed to Habana and Almendares for a special season of 25 games. Santa Clara fielded the same team that had blown away the opposition. This time there was near parity Santa Clara won by the slimmest of margins with a 13-12 record, ½ game ahead of Habana and 1 game ahead of Almendares.

63 AB, 11 R, 23 H, 4 SB, .365

(Note: The batting champion was Torriente with .377)

1924-25

Charleston played for Almendares, which ran away with the title with a 33-16 record, 8½ games ahead of Habana in a 4-team league. Charleston’s teammates included Bullet Rogan, Adolfo Luque, José Acosta, Biz Mackey, John H. Lloyd, Newt Allen, and Dick Lundy.

153 AB, 25 R, 40 H, 10 2B, 4 3B, 4 HR, 0 SB, .261

At the end of the season, the players were organized by nationality into “All-Yankees” and “All-Cubans” teams, which played a series of 8 games. Charleston played for the All-Yankees, who won 5-3.

1925-26

Charleston played for Habana, which ended in second place in the 3-team league, 2 games out with a 32-15 record. However, 13 of their victories represented forfeits by San José, which withdrew with a 3-16 record and 25 games left to play. Charleston recorded only 40 AB, so I wonder if he may have left the team early or arrived late.

40 AB, 14 R, 14 H, 3 2B, 2 3B, 2 HR, 0 SB, .350

1926-27

Adolfo Luque and Miguel Angel (Mike) González led the better players to an alternative league, the Triangular, which played their games at the University of Havana Stadium. It appears to have been more batter-friendly than spacious Almendares Park (where most games were played through 1930), since 4 players had averages above .400. (Pablo Mesa won the championship with .433.) Charleston played for the Leones, who finished in second place in the 3-team league with a 16-19 record, 5 games behind. Notable teammates included Marcelle, Emilio Palmero, and José Acosta. Charleston led the league in hits and in AB.

151 AB, 25 R, 61 H, .404

1927-28

Charleston played for “Cuba,” which placed second in the 3-team league with a 16-21 record, though 4 of their victories represented forfeits by Almendares, which withdrew before the season ended. Teammates included Bill Foster, Pelayo Chacón, and Judy Johnson.

Charleston led the league in home runs (5) and stolen bases (11). According to Roberto González Echevarría in The Pride of Havana, Almendares Park had huge dimensions. “The left-field fence was well over five hundred feet away, and according to Robreño no batter ever clouted one over it...The low fence in right field was four hundred feet away, and only the great left-handed sluggers, like Torriente, Alejandro Oms, Oscar Charleston, Jud (Jorocón) Wilson, and Esteban (Mayarí) Montalvo were able to hit one past it.”

120 AB, 22 R, 42 H, 6 2B, 2 3B, 5 HR, 11 SB, .350

1928-29

Charleston again played for Cuba, but when the team dropped to a 17-22 record they withdrew from the league and forfeited their last 13 games. Teammates included Adolfo Luque (9-2!), Judy Johnson, and Oliver Marcelle.

114 AB, 20 R, 38 H, 6 2B, 4 3B, 4 HR, 6 SB, .333

1930, Special Season

Roberto González remarks that the seasons after 1929-30 found the Cuban League in disarray. In the fall of 1930 the regular season ended after only 5 games, apparently as a result of contract problems with the new stadium, La Tropical. A special season was organized in November in the old ballpark, Almendares Park. Charleston played for Marianao, which finished second with a 9-5 record, ½ game behind in a 4-team league. Teammates included a number of future or former major leaguers: Billy Herman (.229), Howard Freigau (.239), Elias Funk (.346), Al Moore (.292),Tom Angley (.267), Claude Jonnard (5-1), Roy Wilkinson (2-1), and Bill Burwell (1-1).

Charleston led the league in average, runs scored, and triples, and tied with Lundy for the lead in hits.

51 AB, 12 R, 19 H, 0 2B, 5 3B, 0 HR, 4 SB, .373

I checked 1919 box scores, when Charleston and Torriente both played for the American Giants. In a preseason look at the Giants' roster in the 4-12-19 Chicago Defender, Torriente's listed as the center fielder, Charleston as the left fielder; BUT, in all 19 box scores I found in the Defender, Charleston's in center, and Torriente's in left. It pays to check these things.

Btw, the standard microfilm edition of the Defender is missing several issues from that summer.

For the 1923-24 Cuban League season, I have 14 Santa Clara box scores; Charleston plays center field in 8 of those games, right field in 4, and left field in 1. Oms plays center in 4, and Mesa in 2. In one game, Charleston is listed as a second baseman, but I strongly suspect that's a typo, though it's very unclear what position he did play (there's a rf, lf, and cf listed aside from Charleston).

Chris Cobb,

Do you think you'll be able to do WS projections for Charleston? I'm not trying to pressure you -- if you don't have time, not a problem... he figures to be number 1 or 2 regardless.

I'm curious because on the Beckwith thread you emphasized how in their primes Beckwith was a similar hitter to Charleston, and you have Beckwith's best seasons rounding to 34, 30, 29, and 28 (3 times) WS. Those are great seasons, but if Charleston's best seasons aren't higher than that, it really wouldn't really support the peak comparison of Charleston to Mantle, Mays, Cobb, and Speaker, all of whom had multiple seasons above 40 WS. Of course the length and quality of Charleston's career makes him an automatic elect-me regardless, but if his peak really isn't as high as the other great CFers, then I might conclude that Bill James placed him too high as the # 9 greatest player of all time.

Thanks again to you, Gary A, KJOK and the others who have helped us so much in deciphering the data from the Negro leagues.

Thanks again to you, Gary A, KJOK and the others who have helped us so much in deciphering the data from the Negro leagues.

I second that. Without their collective input, I would be lost trying to decipher Negro League stats.

I intend to do WS for Charleston, but I don't know when I will get to it. Just from the raw statistics, I believe that Charleston's peak will be higher than Beckwith's, for three reasons.

1) Charleston was a _complete_ ballplayer, so during his peak in the early 20s, his win-share totals will be boosted by high fielding win shares, and by stolen bases.

2) Beckwith is projected as missing a lot of games. Charleston is more durable, so that will add a couple of win shares over Beckwith in some seasons.

3) I think that Charleston's peak as a hitter is a little bit higher than Beckwith's. Beckwith put up his best numbers in a high-hitting environment in the mid-20s. Charleston's 1921 numbers, in a hitting environment 10% below the majors, are just scary.

For these reasons, I'm guessing that we will see Charleston peaking at 40-42 win shares. I think James might have him a little high at #9 all time, but I'm pretty sure that he's top 20.

Thanks, Chris. That's good enough for me to place him number one, ahead of Cochrane.

Charleston's 1921 season is indeed scary. But look at his teammate, right fielder Charles Blackwell, who batted fourth, behind Charleston:

G: 64
AB-235
H-101
D-17
T-9
HR-10
R-74 (2nd)
W-26
HP-8
SH-6
SB-25
AVE-.430 (2nd)
OBA-.502 (2nd)
SLG-.706 (2nd)

Easily the best season he ever had...a couple of other guys on that St. Louis team (Dudley, Kennard) hit decently well, but check out the team's regular infield:

2B Joe Hewitt: 237/368/280
3B Sam Mongin: 223/368/277
1B Tully McAdoo: 217/329/333
SS Eddie Holtz: 173/249/197

A very unbalanced offense, to say the least. Another weird thing about this team was that its hitters were hit by way more pitches than anybody else--69 in 64 games (the next highest total being Indianapolis, with 34 in 72 games). This isn't a reporting bias, either--St Louis pitchers only hit opponents 25 times, 7th in the league, and easily the best ratio of HBP to HP.

Sticking with the usual standards for HOMers (fulltime - at least half his team's games; parttime - more than equivalent of 10 G per 154, up to half the games; or token - equivalent of fewer than 10 G per 154).

Please make the fixes to this rough draft (isn't there a NY Lincoln Stars in there somewhere?):


1915 - Ind ABCs (partttime)
1916-18 - Ind ABCs
1919 - Det Stars/Chi Am Giants
1920 - Ind ABCs
1921 - STL Giants
1922-27 - Harrisburg
1928-29 - Hilldale
1930-31 - Homestead Greys
1932-37 - Pitt Crawfords
1939 - Toledo (token)

Howie,

Your list matches mine except for 1922-23. Charleston was back with the ABCs for these seasons.

I'd suggest an ABCs hat for Charleston's plaque, in fact, though there's a case to be made for the Crawfords also.

I'd suggest an ABCs hat for Charleston's plaque, in fact, though there's a case to be made for the Crawfords also.

The ABC's is the one that belongs, IMO.

ABC's? The prime of his career was actually spent with Harrisburg?!

Charleston was with the Lincoln Stars for much of 1916.

I was browsing baseball-reference and stumbled on the fact that they have pages for the Negro League players who have been elected to the Hall of Fame. No baseball data - just names and birth and death data.

Of all of the ones they had listed, Charleston was born the furthest north, in Indianapolis. After that: Maryland, Oklahoma, Cuba - but the rest of them were sons of the South, playing for teams based in northern cities.

ABC's? The prime of his career was actually spent with Harrisburg?!

He was with the ABC's from '22 to '23. He was only with Harrisburg from '24 to '27.

Of all of the ones they had listed, Charleston was born the furthest north, in Indianapolis. After that: Maryland, Oklahoma, Cuba - but the rest of them were sons of the South, playing for teams based in northern cities.

jimd counted 2 of 21 NeL HOFers born in "free states" of 1860, including Campanella in PA and Doby,Irvin,Robinson in the Deep South.


jimd's racial and regional demographic Adventure

NeL HOMers to date, on the other hand, show a slightly different birth state profile:

Rube Foster-Texas
Frank Grant-Massachusetts
Pete Hill-Pennsylvania
Grant Johnson-Ohio
J.H. Lloyd-Florida
Bullet Rogan-Oklahoma
Louis Santop-Texas
Cristobal Torriente-Cuba
Cyclone Joe Williams-Texas

I once did a study, some years ago, of Negro Leaguers' birthplaces, based on Riley's Biographical Encyclopedia. Unfortunately the data haven't survived so I can't provide details, but I was surprised to find that only 55 percent of all Negro League players with known birthplaces at that time were born in Southern states, which I think were defined as Confederate states plus Kentucky.

That was 55 percent of U.S.-born Negro Leaguers--Latin American players were excluded.

I've got Charleston having played with the following already-inducted HOMers:
with Louis Santop in 1916, NY Lincoln Stars
with Pete Hill in 1919, Detroit Stars
with Smokey Joe Williams in 1930-31, Homestead Greys

Wow, Charleston with Williams on the same team.
Most dazzling HOMer combos so far:

Cy Young and Tris Speaker, Boston 1908
Young and Nap Lajoie, Cleveland 1910
Pop Lloyd and Smokey Joe Williams, NY Linc Gia 1911-13
Babe Ruth and Speaker, Boston 1915
Grover C. Alexander and Rogers Hornsby, St. Louis 1926
Walter Johnson and Speaker, Washington 1927
Ty Cobb and Speaker, Philadelphia 1928
Eddie Collins and Speaker, Philadelphia 1928
Cobb and Collins, Philadelphia 1927-28
Charleston and Williams, Home Gre 1930-31

P.S. The only "normal" HOMer that Speaker ever played with was Stan Covaleski.

Frank Grant-Massachusetts
Pete Hill-Pennsylvania
Grant Johnson-Ohio

So the earliest ones, the pre-league players, were much more northern. That fits, of course. Fleet Walker was a northerner, right?

Campanella, Doby, Irvin, and Robinson have playing records on bb-ref, so their pages are filed in different places.

Bullet Rogan-Oklahoma

If you try to define North and South by relating it to the Civil War and/or slavery, note the following:

In Indian Territory, a small number of Indians did own African-American slaves. The war was extremely divisive, with most tribal govenrnments taking no official position, but individual Indians fighting on both sides.

55 percent of all Negro League players with known birthplaces at that time were born in Southern states...

It's just a small sample, but it looks like the most famous stars of the league period, the ones who were elected to the HoF, didn't follow the same percentages. Was it easier for teams to find marginal players near where they played, but they got the stars from everywhere? Or is it just random?

Charleston and Williams, Home Gre 1930-31

I think these two were joined by a rookie named Josh Gibson in 1931.

The 1913 Lincoln Giants had Lloyd, Williams, Santop, and Grant Johnson, plus Spotswood Poles and Dick Redding.

The 1933-36 Pittsburgh Crawfords featured Satchel Paige, Josh Gibson, Oscar Charleston, Cool Papa Bell, and Judy Johnson, though Paige missed nearly the whole season in '35 playing for a semipro team in North Dakota.

The 1931 Homestead Grays had Gibson, Charleston, and Williams, but also Willie Foster and Jud Wilson.

The 1910 Leland Giants (the American Giants' predecessor) had four HOMers: Rube Foster, Pete Hill, J.H. Lloyd, and Grant Johnson.

My four-HOMer Negro League teams so far are:


1910 Chi Leland Giants HR Johnson Hill Foster Lloyd
1911-13 NY Lincoln Giants HR Johnson Santop Lloyd Williams
1915 Chi American Giants Hill Santop Foster Lloyd

I spotted one Cuban League team with four HOMers:

1912 Habana Reds Williams Johnson Lloyd Hill

Here are the birth states/countries of all the players with their own threads. This might give a slightly better idea of where star players were from, though it's still not a great sample (should be more Cubans, for example).

John Beckwith-Kentucky
Dave Brown-Texas
Pelayo Chacon-Cuba
Oscar Charleston-Indiana
Andy Cooper-Texas
Bingo DeMoss-Kansas
John Donaldson-Missouri
Bill Foster-Texas
Rube Foster-Texas
Frank Grant-Massachusetts
Pete Hill-Pennsylvania
Grant Johnson-Ohio
Heavy Johnson-Kansas
Judy Johnson-Maryland
J.H. Lloyd-Florida
Dick Lundy-Florida
Jimmie Lyons-Illinois
Dave Malarcher-Louisiana
Oliver Marcelle-Louisiana
Jose Mendez-Cuba
Bill Monroe-unknown
Dobie Moore-Georgia
Bruce Petway-Tennessee
Spotswood Poles-Virginia
Dick Redding-Georgia
Bullet Rogan-Oklahoma
George Shively-Kentucky
Chino Smith-South Carolina
Ben Taylor-South Carolina
Clint Thomas-Kentucky
Jules Thomas-unknown
Cristobal Torriente-Cuba
Chaney White-Texas
Sol White-Ohio
Nip Winters-D.C.

Texas (5)
Cuba (3)
Kentucky (3)
Ohio (2)
Kansas (2)
South Carolina (2)
Georgia (2)
Louisiana (2)
Florida (2)
Tennessee (1)
Indiana (1)
D.C. (1)
Virginia (1)
Illinois (1)
Oklahoma (1)
Pennsylvania (1)
Massachusetts (1)
Maryland (1)
Missouri (1)

Of 35 players, the birth states of 2 are still unknown; 3 are Cuban; 8 hail from free states in 1860; 7 from border states plus Oklahoma and D.C.; 15 from Confederate states (including Tennessee). It matches up roughly with my old survey: 60 percent (18 of 30) U.S.-born players here were from former Confederate states plus Kentucky.

Records of teams with 4 NeL HOMers:
1910 Leland Giants 11-0 (123-6 overall; later called by Rube Foster his greatest team)
1912 Habana Reds 22-12, Cuban League pennant
1913 Lincoln Giants 16-7, including a decisive victory over the American Giants for the black championship

In 1911, I have Williams with the Leland Giants.
In 1912, I have Johnson with Brooklyn.
In 1915, I have Santop with the Lincoln Stars. A quick check of box scores in the Indianapolis Freeman showed 30 American Giants games with no appearance by Santop; Santop did appear in 10 of 13 Lincoln Stars box scores (including a few games against the American Giants).

Lloyd and Johnson, btw, were the double play combination for all three clubs.

In other words, with regard to the birth state stuff in post #44, 2 out of 20, or 10 percent, of U.S.-born NeL HOFers were born in free states, but 8 of 30, or 27 percent, of this different (and slightly larger) sample were.

Ok, Gary A, I'll put the "inquiry" sign up on those for now. I have Santop playing for both of those teams in 1915, will await additional comments from the crowd...

What is the range of birthdates for the 33 black stars, less Grant and White if they are outliers.

Gary A-
Bill Monroe was born in Tennessee. Jules Thomas was born in Virginia. One correction: Jimmie Lyons was born in Indianapolis, Indiana; but raised in Chicago, IL.

Well, based on the initial results of my quasi-new MLE routine, as noted on the Judy Johnson's tread, I decided to hunt bigger game and see what Charleston's MLEs and WS might look like. My initial attempt follows.

Two things to note.
One: I have not made any attempt at this 1917-1919 seasons. Gary A's website gives me the 1916 data, and SoG gives me 1920-1941, but there isn't yet a 1917-1919 source other than Holway. At this point, given what all we have to work with, I just chose to leave it be. When you see the MLE, you'll see why.

Two: The playing time shaping is minimal. I went 250 outs in 1916 (about 60-65% of a season, which is reflected in Gary A's team totals for Charleston's squads), then full-time play of 400 outs all the way through 1934. In 1935, I gradually started him into decline, with 350 outs, then 250, then 200 in 1937. There's no data for 1938, so I skipped it, and in 1939-1941 I accelerated the decline greatly: 75, 25, 1. That one is off a one-game line, so I feel OK with just leaving it be.

Three: Defensively I made him a 4.0 WS/1000 innings CFer through age 32. After that, he goes to 3.0. Looking at his neutral stats, his game just dropped down a level at that age, and I'm figuring that's probalby a good time to put him into a defenisive decline. Then at age 39, 2.25 WS/1000.

OK, here goes.



Given the four seasons that I haven't accounted for here, you can estimate there's 50-100 more WS in the ether that Charleston might end up with. I don't really know for sure. You can see his tremendous peak and prime from 1921 to 1928, then a leveling off at a slightly lower strata before a more pronounced decline in his late 30s (though some of that is due to the playing time shaping I've done. Conversely some of those pretty nice looking late-career OPS+s are the result of "regression" to his career totals in combination nwith season of less than 18 games. He appears, in the MLE rendition, to be a lot like Hank Aaron or Frank Robinson. A little faster, maybe; playing CF rather than RF (and well). Not as good a hitter as those guys (~155 OPS+s) but close in total value by dint of being a CF. Probably not as valuable as Mays overall since he didn't hit as much and Mays was an all-time-great defensive CF for a long time. But anyway, that's what the MLE says, and the truth could be very different.

Eric, it's interesting that you have Charleston's 1921 as only his third-best offensive season, since the prior comments on this thread seem to refer to it as of Ruthian proportions...what accounts for the discrepancy?

Sh*t I just lost a multi, multi hundred word post in answer to your question, Dan.

Well, I'll encapsulate.

1) I looked to see if my lgRC/G estimates were causing the issue. There are discrepancies between my estimations and, for instance, Gary A's actual R/G. Mine are always substantially lower. But I'm nearly certain those discrepancies are the result of there being many, many more errors in the NgL game than the MLB game. In fact, I was just putting together some tables about how often NgL 3Bs erred versus their MLB counterparts of the same era, and it was astonishing. Every NgL guy was at the very bottom of the list, even below the worst MLB hands. They mostly erred about once every 4.5-6.0 games, while MLB guys erred every 6-9 games. However, I rasoned that this should not produce a dampening effect of the sort you're inquiring about because I'm comparing the RC estimate of the player to a RC estimate of a an average player using the same formula. Since I'm not comparing a player's RC to the league's R/G, I think this is not causing the distortion you're talking about.

2) I think it more likely that the issue comes in when I'm filling in gaps. Specically my current protocol looks like this for Oscar's 1921 season.

Actual, translated to neutral MLB/4.5 environment = 106 outs at 408/465/644 49 RC, 12.5 RC/G

I'm allotting him 400 outs, which is a full-time season. So 73.5% of his outs are not covered by his neutral MLE translation. I fill the gap with his career average.

Career-average-based projection to fill in gaps = 294 outs at 305/379/502 84 RC, 7.44 RC/G

That makes his final, neutral, projected+translated line....

400 outs 335/404/543 133 RC, 9.0 RC/G

The reason it comes out third best is that he played fewer games and made fewer outs, so there's more gaps to fill between his actual playing time and the playing time of a full-time MLB player.


I'm very much open to improving on this. I've been using career average because it's safe, but I can think of four reasonable options, including it, for filling in the gaps between actual and projected playing time:

1) fill in with career average performance as I'm doing

2) fill in with average performance of prior 3 or 5 years (which is Chris's method I believe)

3) fill in with both career average AND prior 3 or 5 years performance, weighted 50/50 or whatever weighting is appropriate/sound

4) fill in with current season performance, career average, AND prior 3 or 5 years at 33/33/33 or whatever weighting is appropriate/sound.

Again, I hope that if anyone has a strong rationale for one of these options versus another that they'll offer them so that I can make a better system.

Dan, I hope that answers your question.

What weighting would a Marcel projection use?

Wouldn't a reasonably projection not have a player get 600+PAs in 15 consecutive seasons? That's unreasonably high durability.

Phil,

I don't disagree with you. The issue is that I don't have team games to work with for the SoG numbers. I looked at the previous Clark/Lester book, which included standings, and the team game totals are below Charleston's reported game totals in many cases. So I just ducked the issue by giving him full-time all the way through his prime. I could do something like tried on Easter, allowing for a certain amount of variance in his outs and using a random number generator to apply it + or - some set number (like 375 outs +/- 50 or something).

user,

i don't know what marcel would do. i also don't know if it's compatible since it's a projection system that is looking at unknown future performance based on known past performance. would it be applicable?

Given what little I know of it I don't see why not - a required adjustment would be to make it take data from seasons either side of that we're interested in - but I presume symmetrical weighting would be fine.

The reason it comes out third best is that he played fewer games <u>and made fewer outs</u>, so there's more gaps to fill between his actual playing time and the playing time of a full-time MLB player.

By tying it to number of outs aren't you sort of punishing Charleston for his .465 OBP? Isn't that a lot of why he made so few outs - because he was so good at getting on base?

For example, according to baseball-reference, Babe Ruth never made more than 376 outs in a season, despite having 7 seasons where he was in the top 10 in the AL in plate appearances.

What weighting would a Marcel projection use?

Marcel projections use a weighting of 5 for the most recent season, 4 for the season before that, 3 for the season before that, and 2 for the league average.

While I'm not an expert in this, might it be that regressing the stats of a Negro Leaguer to the length of a major league season unreasonably pulls his averages closer to the mean? Very few major leaguers of consequence have important seasons where you have to adjust from 106 to 400 outs. If you do this with every season, it seems like you would end up with suppressed variation.

Maybe a better way would be to use 3-5 year averages to fill in the regression outs, rather than career averages.

By tying it to number of outs aren't you sort of punishing Charleston for his .465 OBP? Isn't that a lot of why he made so few outs - because he was so good at getting on base?

Wow, you're absolutely right, thanks for the catch. I totally missed it. I'll have to do it by PAs instead to avoid penalizing players who get on base a lot.

I did this for his 1921 season and it did improve his rates to

.339/.408/.549

from

.335/.404/.543

Thanks!

Eric,

Great stuff, thanks. A few comments:

1) Charleston moved from center field to first base full-time in 1930, at age 33 (I think he also played first base for part of 1929). You have his FWS declining at about that time anyway; maybe they should decline a bit more sharply?

2) I'd concur with the concern about 15 consecutive seasons with 600+ PA. Neither of the contemporaries he was usually compared with, Cobb and Speaker, ever came close. Frank Robinson didn't do it, either. Aaron did, with exactly 15 (and a 16th with 598 PA), but had the advantage of 162-game seasons for much of his run.

On the other hand: Charleston hardly ever missed a game. Counting his NNL games 1920-22, Cuban League 1927/28, and 1928 Hilldale games against top black teams, Charleston played in 285 of his teams' 287 games (missing only two in 1921). I also can't say that I've ever heard of him missing substantial time (or having a major injury) in the 1920s or 30s. It's true, of course, that NeL conditions (small rosters, lower quality of players around them) tended to result in stars playing a higher percentage of games than they might have in the white majors.

It could be that the high number of PAs partly stems from forcing each season to have 400 outs--it could be that switching the basis to PAs would lower the PAs in the translation.

3) In #51 above, his 1921 rates are .339/.408/.549--which are the same rates given in #61 as the *corrected* rates (after switching from outs to PA as the basis).

On the other hand: Charleston hardly ever missed a game. Counting his NNL games 1920-22, Cuban League 1927/28, and 1928 Hilldale games against top black teams, Charleston played in 285 of his teams' 287 games (missing only two in 1921). I also can't say that I've ever heard of him missing substantial time (or having a major injury) in the 1920s or 30s. It's true, of course, that NeL conditions (small rosters, lower quality of players around them) tended to result in stars playing a higher percentage of games than they might have in the white majors.

I'm back on PAs instead of outs now, despite what's listed above, but I'm not going to post another MLE until I get a better sense of how I should approach the regression question. The playing time's really knotty for the reasons I listed above. I ran a little query with the SBE. I checked out which HOM CFs from 1910-1950 had played at least five years in CF before age 33. Then I figured out which HOMers of the same era had played 1B after age 33. I searched on these criteria for any seasons of 600+ PA, and what I found was that HOM CFs averaged 8.6 600 PA seasons; the 1Bs averaged 1.8. It should be noted that Greenberg missed opportunities in 1945 to rack up PAs due to war. The net rsult is 10-11 600 PA seasons for HOM-level players.

Charleston is shown with 15 such seasons, so it seems likely that reducing that by five or six years will make the MLE more reasonable. Based on this little mini-study, I think I would do it on the back end of his career, probably giving him a 600 PA year upon his move to first, then reducing downward right away. Given the more difficult offensive conditions for the seasons 1917-1919, which I haven't translated/projected, it might be that he'd come in under 600 PAs for those years anyway unless he were a leadoff hitter. Here's what I'd propose for PAs, please let me know what you think---I'm dealing in round numbers, of course:

1916 350
1917 500
1918 550
1919 575
1920 600
1921 600
1922 600
1923 600
1924 600
1925 600
1926 600
1927 600
1928 600
1929 600
1930 550
1931 550
1932 550
1933 550
1934 500
1935 500
1936 400
1937 300
1938 ??? (where was he that year?)
1939 100
1940 50
1941 5

That takes it down to ten 600 PA seasons, with gentle slopes on either slide indicating the rise to prominence and the decline. Within the group of 500+ PA seasons, I could always use some kind of randomizer to generate some variation around this pattern.

I'm not actually sure of his precise career shape yet because I haven't tried the missing 1917-1919 seasons. That might cause me to simply lop off 1939-1941. I don't know if 3000 hits was a big enough deal by 1936 that a player would hang on for it (as Paul Waner did in a few years), but I'd probably end the MLE with either his 3K year, or the first ineffective year thereafter.

Does that seem more reasonable to everyone?

While I'm not an expert in this, might it be that regressing the stats of a Negro Leaguer to the length of a major league season unreasonably pulls his averages closer to the mean? Very few major leaguers of consequence have important seasons where you have to adjust from 106 to 400 outs. If you do this with every season, it seems like you would end up with suppressed variation.

Maybe a better way would be to use 3-5 year averages to fill in the regression outs, rather than career averages.

This is my thought also. If you're going to regress a season, you need to regress the player TO HIS OWN PERFORMANCE LEVEL somehow. In other words, regressing should not BOTH:

1. reduce peak seasons where team games are less than 154
AND
2. significantly reduce 'overall career performance' (beyond the difference in league quality).

regression should only do #1, as those short-season peaks SHOULD be regressed to be comparable to MLB.

or maybe you are regressing the player to himself - guess I need to read the Judy Johnson thread....

Just testing, can't seem to post my ballot to the other thread ...

Here's what I'd propose for PAs, please let me know what you think---I'm dealing in round numbers, of course:

Here's a suggestion. If we can assume that Charleston's closest comps were Cobb and Speaker, why not assume that his PAs at each age are equal to the average of their PAs at the same age. (And if you think someone else makes a better comp, substitute them.)

I can't quite put my finger on what's not right. You seem to be regressing the way I would expect, but Charleston's looking more like Ken Griffey, JR or Duke Snider instead of like Mickey Mantle or Ty Cobb....

KJOK,

Here's a step-by-step of Charleston's 1921 season MLE. It includes some changes I've made since the first MLE in post 51. There may be small rounding errors in what I'm writing below.

1) Figure out player's RC for the season in question.
Charleston created 50 runs on about 106 outs (I estimated CS via an estimated 70% SB%), which is 12.74 R/27 outs for those interested.

2) Figure how many runs an average player in his league would create per game.
Charleston's league batted .256 and slugged .351. Based on my own estimates, I peg the league's bb/ab at .915 of MLB and SH/AB at 1.35 of MLB. I estimated .02 SB/AB with a 70% success rate. This leads me to a RC/G estimate of 3.95. (This is lower than the actual R/G for sure, resultant from the higher error totals in the NgLs.)

3) Figure Charleston's RC/G context by multiplying his league's RC/G times his park factor (if there is one).
3.95 * .98 = 3.87 Charleston was playing in a 3.87 RC/G environment.

4) Figure out what his RC would be in a neutral, 4.5 R/G league.
4.5 / 3.87 * 50 RC = 58

5) Multiply by .85 to find the player's RC/G in a neutral, 4.5 R/G, MLB context.
.85 * 58 RC = 49

6) Use the Willie Davis comment methodology to project the player into the new environment.
Charleston goes from .436/.687 with 50 RC in the NgL to .408/.644 with 49 RC in a 4.5 MLB league.

The summ of all neutralized, MLB-level lines is the basis for the career regression that will follow.

8) To project him to a full season's playing time decide on playing time in the form of PAs and figure how many "missing" PAs are out there.
I'm using 600. His NgL line contains 191, there are 409 remaining PAs.

9) Next figure the player's career rate of neutralized MLB RC/PA and multiply by the "missing" PAs.
Charleston's neutralized career RC/PA is .18. Multiply by 409 = 74 "missing" RC.

10) Use the Willie Davis comment method to fill in the PAs.
Charleston's 409 missing PAs turn out to be at 307/381/505, which is quite close to his neutralized 305/379/502 line (difference is using whole numbers, no fractions in the Hits column).

11) Combine the translated and projected statlines
TRANSLATED 408/465/644/49RC
PROJECTED 307/381/505/74RC
=============================
TOTAL 339/408/549/123RC

This is the total, nuetralized, projected 1921 season of Oscar Charleston. But we're not done yet!

12) Obtain the new R/G of his final MLB destination.
The 1921 NL was a 4.59 R/G league.

13) Divide the new league's RC/G by 4.5 (which is the neutral lg's R/G) and multiply by his RC.
4.59/4.5 * 123 = 126 RC

14) Use the Willie Davis comment methodology to adjust his line.
He goes from 339/408/549/123RC to 343/412/555/126RC

That is the final 1921 MLE line.

I see the following three steps as the most likely sources of potential error:

-Step 2: Without actual R/G for most seasons, this is a bit dodgy; my RC/G are all estimates.

-Step 5: Am I using too low a conversion factor? James uses .82 on runs for modern AAA, and I took it upward for various reasons. On the other hand, Davenport uses a .86 converter for some AAA leagues and higher yet for Japan; I think that's on runs. Should I convert at a higher rate?

-Step 10: Am I using the correct technique for "regressing" his projected time?

PLEASE let me know if there's some key step that I'm missing. I want to get this right, so I'm all ears.

Eric & I had this discussion off-line, but as I see it choosing a baseline to regress to (assuming you even SHOULD regress) involves weighing:

1. A career baseline gives you more PA's, and would hopefully provide a 'better' sample of the player's overall talent level in his career. The drawback is that a player's talent level changes over time, and using this method may 'smooth the peaks' more than is desired.

2. A 'surrounding season' baseline may give you a better sample of the player's talent level at that point in his career. The drawback is that those surrounding seasons, based on few PA's, could contain lots of 'noise' or random variation, be impacted by an injury or slump, etc. and not really be a good sample of the player's overall talent level, and by using it to regress you actually amplify those issues.

Would the alterantive to regression be a straight line extraoplation? If so wouldn't that be worse in terms of amplifying the small-sample? Or is there another, better option?

Eric,

1. What raw data are you using? The numbers in post #2 are totally different.

2. You can't just multiply the total RC by the translation factor, because there is an OBP effect--as his RC goes down, his outs consumed go up, leaving fewer outs for the rest of the team. You have to account for that.

3. The formula I use to "smooth out" fluctuations in BABIP and runner stranding for pitchers is 38% Year in question, 13% year before, 13% year after, 5% 2 years before, 5% 2 years after, 25% career.

4. I would do this the following way (using data from post #2).

a. Calculate RC (I use XR) and outs (70 and 141), and XR/Out (.499).
b. Multiply XR/Out by the .85 conversion factor (.424).
c. Calculate park-adjusted league XR/Out (.153).
d. Calculate the player's league difficulty-adjusted RC+ (.499/.153 = 277). This is between 1994 Jeff Bagwell (273) and 1924 Rogers Hornsby (287). Not bad.
e. Use the formula in #3 to calculate what RC+ to use for the missing PA's. I don't have this data. But I'll pick a number off the top of my head and say 200 is the "true" performance level of an inner-circle Hall member at his absolute peak (200 seasons include 1964 Mays, 1932 Ott, 1990 Bonds, 2004 Pujols, 1949 Musial, etc.). 336 PA at 200 plus 264 PA at 277 is 234 total.
f. The 1921 NL scored .187 runs per out. .187 * 2.34 is .438 XR/Out.
g. Given the shape of his production, that translates to a line of .406/.485/.691, a 217 OPS+.
h. A 234 RC+ in 600 PA in the 1921 NL is worth 9.4 wins above average. Assuming he was an average fielding center fielder, I'd add on 1.1 wins for replacement level, for 10.5 total, and then regress to the 2005 NL standard deviation, reducing him to 9.9 WARP2. That's similar value to Bonds '92, Mays '65, Hornsby '29, Morgan '76, Wagner '09.

Note: the biggest reason for this massive discrepancy is that I think Eric is using different raw data for Charleston's 1921 than the stats provided in #2. The second biggest reason is that he is regressing to career averages, while I am regressing to a true peak HoM performance level.

Questions:
1. do we know that Negro League play was roughly AAA level?
2. do we know what the standard deviation of performance in his league was? The data in #26 suggest it may have been exceedingly high, which would be cause for a lot more regression.

Questions:
1. do we know that Negro League play was roughly AAA level?
2. do we know what the standard deviation of performance in his league was? The data in #26 suggest it may have been exceedingly high, which would be cause for a lot more regression.

Dan, I am indeed using different data, the stuff in the Shades of Glory book that came out last spring. That accounts for some of the observable difference. And our sense of the quality of play is that it was likely around or slightly above AAA. Chris Cobb did the original research, and his finding was NgL batting averages were about .90 of MLB averages and SLGs about .81. At the time of his research, we had no leaguewide AVG and sLG figures to work with. Now we have the Clark/Lester/HOF figures, and I actally suspect that the conversion factor may be slightly higher than his study shows if it were replicated using relative AVG and SLG. The NgLs (with pitchers) from 1920-1948 batted 269/365, MLB (with pitchers) batted 275/388. If Chris found a .9 correspondence on AVG, it's possible this could increase by the 2% difference between the two leagues. Chris found a .81 SLG concordance, but the difference is 6% here. So it's possible that a new conversion factor could show .918/.859. And it's really on the SLG where it would be noticed because in many of the MLEs I've done I've thought to myself: Gee, this really seems to dampen his SLG....

The second biggest reason is that he is regressing to career averages, while I am regressing to a true peak HoM performance level.

Since I started doing MLEs a year or two ago, I've been thinking a lot about them and trying to learn a lot too. The former is easier than the latter. ; ) The major stumbling blocks are the league totals for leagues without a substantial body of reference work, the question of league quality, and lastly the question of extrapolation from shorter seasons. I've addressed them each in whatever way I can. And I've discovered that it can make a small difference or a big difference, but it's all relative.

That's why I've decided to take the step of putting all of a player's seasons into a common, nuetralized 4.5 R/G environment. That way, I have a snapshot of the man's talent as expressed in common units. This then allows me to make any adjustments by using those same common units. See you have to remember that in some cases, you're dealing with guys who played in 7-10 or more leagues. Not teams, leagues (check out the Marv Williams thread). And partly that's what brought me to this place. Each league can be wildly different and it causes some distortions. So the neutralizing really offers some interesting information. And is a sound basis for extrapolation either by regression, simple linear means, or some other method. And while I used to extrapolate only with linear proration, I'm now more of the belief that using career norms or some combo of career and recent performance weighted heavily to career is going to yield defensible results due to the presence of a more robust sample.

OK, round two, with feeling.

Changes from version one:
-Changed from full career regression to 55/45 career to previous three seasons, except where threre are no previous three seasons, in which cse, just career (so 1916, 1917-1919, 1920-1922). This doesn't actually change the final career totals too much, but it does create a little sharper peak delineation and does make the come down a little more sever.

-Changed from .85 conversion rate to .90. KJOK suggested this, citing AAA rates as slightly higher than my original conversion rate (~.86-.88), also citing that japan is generally around .92, so spliting the diff is .90.

-Changed fielding in light of Gary A's pointing out the position switch. This means that in the transition year, he went from 4.0 to 3.0 WS/1000 innings, then after the full-time switch, down to 2.0 WS/1000.

-Dropped the 1939-1941 token seasons from regression and MLE.

-Added the 1917-1919 seasons "from scratch." I don't have anything like league totals or anything totals for them, so I simply used the career figures and plugged them into the PA totals I was going to use (see below). Luckily they fit in nicely.

-Changed to PA-based extrapolation from outs-based in deference to smart comment made above about penalizing guys who don't consume outs.

-Changed the flow of PAs in Charleston's career to match post 63.

-Smacked self in head after realizing I'd used pitcher-inclusive figures for prior OPS+ calculation, made penance at the altar of David Foss, and recalculated with correct figures.



The OPS+ is lower than expectation, for sure. But the WS are in neighborhood you'd figure, as are the career totals. Charleston's MLE career totals come out at

PA: 11866, ~15th all time, between Speaker and Brooks Robinson. Others in vicinity are Frob, Wagner, Molitor, and E. Collins.

H: 3256, one more than Murray, 11th all time. Mays and Lajoie nearby.

TB: 5381 would be 11th all time between Palmeiro and Frob with Murray nearby.

W: 1364 28th all time between Reggie and Palmeiro; Dewey, Speaker, McCovey nearby.

RC: 2238 13th all time between Anson and Foxx; Wagner, Ott, and Rose nearby.

WS: 611 8th between Speaker and Musial. A little gap until E Collins and Mantle. This is a bit puffed up because he's got full-season credits for the shortened war years of 1918-1919. Surrounding players of the era don't have that luxury. Pull 8 WS off the total (5 for 1918, 3 for 1919), and he's 9th, between Musial and Collins.

OPS+: 144 t-51st with Sam Crawford, Mike Donlin, Bill Joyce and Hack Wilson. Nearby are Schmidt, Stargell, Thompson, Killebrew, Mathews, O'Doul, G Stone, Stovey.

So Charleston is coming out on the career side about where you'd expect him to, but for whatever reason, it's just not showing up in OPS+, and I don't have a very good explanation for why. So it ain't perfect, but I feel reasonably confident that I've made reasonable adjustments and gotten it to the best point I am able.

If the group concurs and thinks this is a viable method, then I'll reopen the files on a few of our backlog NgL hitters over the next few weeks to see what I can see (namely Wilson, Howard, and Clarkson, but maybe some other guys too).

So Charleston is coming out on the career side about where you'd expect him to, but for whatever reason, it's just not showing up in OPS+, and I don't have a very good explanation for why.

Looks like it's because instead of being Mickey Mantle, we now see the Charleston is a little more like a slightly better version of Ken Griffey JR, only more durable and with a longer career.

Like JR, he was performing at a level as one of the best players in his league, but also like JR, his level of play flattened out over the last part (9 years, in Charleston's case) of his career.

I guess as I do a few more of these, we can compare them to current MLEs and see if I'm just chronically coming in lower or if, indeed, Charleston's own career arc is the culprit.

The difference between the NeL and AAA is:

• In AAA everybody is more or less a AAAer

• In the NeLs you've got MLers and you've got AA and even A players, with an average of an AAAer

On a completely different topic (or is it?) I don't think there is any question whatsoever that MLEs:

• Over-estimate career length and in-season durability

• And under-estimate peaks

We know this to be so, we just don't know what to do about it, and we think it sort of roughly equals out, give or take, though on balance what we think and what we feel is probably two different things. We think it balances out, but we feel that it probably over-states most careers. But whose?

And of course it only equals out (if it does) for career voters. For peak and prime voters, no, not even close.

Still it's better than not doing the MLEs. You can pretty clearly rank order the NeLers among themselves and that is not nothing.

Another possibility is that NeL contraction in the 1930s caused competition levels to rise. It's quite possible, given Charleston's age and the dimuntion of his athletic skills marked by the shift to first base, that the lowered values for the 1930s are accurate, but it is suggestive that the decline happens at the time when an increase in quality of competition would be expected.

Especially considering other players--Cool Papa Bell, Biz Mackey, Willie Wells--also suffered similar declines, I believe.

It's also worth remembering that, as Gary noted, charleston switched from CF to 1B at age 32 or 33. That's a pretty big switch for a player who wasn't known to be injured. One logical suggestion about it is that his athletic slippage was not only obvious, but dramatic and led to his having to move to the least mobile place on the dimaond from the second-most mobile. He could surely have handled a corner outfield slot (esp with his bat), and he was a legend in his time, so it's a little odd that he made that move unless he was riddled by injury (a la Mickey Mantle). If there was some dimunitioin in his speed and athletic ability, it would surely also have translated into issues at the plate, if not immediately, then in due course. This may be reflected. On a technical note, he also has somewhat fewer recorded PAs in that era, so the regression will hit him harder then. He's probalby got around 30% fewer recorded PAs in the 1930s than the 1920s (avg of 236 in the 20s; 165 in the 30s).

I think these are overlapping effects though. In one decade he's his full self, in his prime, racking up PT; in the latter decade he's a diminished version of self who has increasingly less PT. Hard to seperate one from the other.

That's a very plausible career path--a ramp-up year, 11 out of 12 years with an OPS+ of 145 or more (with a two year peak right in the middle of that run at 177 and 182), then as a result of injuries or deterioration a decline to a lower plateau for seven out of eight years between 127 and 134, then a replacement level season and he's done.

A few observations:
(1) Obviously there is some smoothing out within the segments but their overall levels and lengths seems plausible.
(2) There is some possibility that the decline to the lower plateau is at least partially tied to changes in competition level--the problem is that it is hard to adjust for that--do we adjust the first plateau down or the second plateau up or some combination of the two?
(3) His decline is a little sharper than for many other top tier hall of famers but his time as a still good though declined player is a little longer than normal. Therefore, the overall value of his decline phase is about standard.

Basically, if you take into account points (1)-(3), Charleston looks like Frank Robinson with 2/3 of his career as an excellent CF. Since I have had Robinson in the mid-late 20's in my All-Time rankings and Charleston in the early 20's, these allegedly disappointing MLE's actually suggest that I might have been slightly underrating Oscar.

There could be something else other than injury in making Charelston's move from CF to 1B a permanent one. Charleston moved to 1B in 1930. This certainly could have been injury related. But he became a player-manager in 1932, and one can imagine that even if he were healthy enough to move back to OF he might have chosen to remain at 1B as an easier position from which to manage. Most sucessful player-managers were infielders: Charleston may have chosen to fit the mold.

There's some support for this idea that playing 1B may have had something to do with Charleston becoming a manager. At the end of 1928 (September and October), Charleston took over the player-manager role at Hilldale in place of Otto Briggs (the club's right fielder) for what was termed the "barnstorming season" (even though Hilldale wasn't actually in a league that year). When he did so, he moved to first base.

First base, btw, was always Charleston's secondary position, from at least 1922. That year Ben Taylor (the ABCs player manager) was out with an injury for a while, and Charleston was moved to first base to take his place.

Charleston also managed Harrisburg in 1924-25; I don't know if he played any at first then.

Of course, lessened mobility might have something to do with it, too, though I haven't found any contemporary mention of it. He had certainly put on some weight by the end of the 20s, though.

This is a bit puffed up because he's got full-season credits for the shortened war years of 1918-1919.

Charleston played a full season in 1919--haven't checked 1918.

Dr. Chaleeko wrote:

Chris Cobb did the original research, and his finding was NgL batting averages were about .90 of MLB averages and SLGs about .81. At the time of his research, we had no leaguewide AVG and sLG figures to work with. Now we have the Clark/Lester/HOF figures...

Do we have league-wide AVG and SLG for the Negro Leagues from 1920-1948? If so, where can one find it?

The NgLs (with pitchers) from 1920-1948 batted 269/365, MLB (with pitchers) batted 275/388.

Once again, where do these NgLs numbers come from? I'm new to this field of study and wonder if I'm missing some newly unearthed cache of "league total" type data from 1920-1948...?

Thank you for your help,

Brian

The League-wide AVG and SLG can be extracted from the HOF stats provided at the "Batter Stats" link here:

Special Negro League HOF Ballot Info

Those Charleston MLE's do not give him an extended inner-circle HoM peak, which is what I would have imagined him to have. Moreover, he seems to be coming in at slllightly below average playing time--perhaps because of missing HBP's? If so, that would explain why his listed FWS are translating to a bit better FWAA here than what I've read of his rep, since my spreadsheet thinks they are being accumulated in fewer innings. This is probably worth about the same on defense as the missing HBP would be worth on offense, so it should even out. Also, Dr. C has projected him for a far superior SB success rate to the league average for the 1920-25 period where CS data is available for the NL, resulting in terrific baserunning wins for that stretch. At any rate, here they are in my WARP (assuming they are all translated to the NL, as I believe they are):


Glossary

The following numbers are all standard deviation-adjusted. SFrac is the percentage of the season played (compared to a player with league average PA/G playing every game). BWAA is batting wins above average, BRWA is baserunning wins above average, and FWAA is fielding wins above average. Replc is wins above average a replacement player at the same position would have accumulated in the same playing time, and WARP is the first three minus the fourth (wins above replacement).



This comes out to a career salary of $332 million, comfortably in the inner circle, just above Rickey Henderson, Cal Ripken, and Frank Robinson, substantially below Gehrig, Morgan, and Schmidt.

Moreover, he seems to be coming in at slllightly below average playing time--perhaps because of missing HBP's? If so, that would explain why his listed FWS are translating to a bit better FWAA here than what I've read of his rep, since my spreadsheet thinks they are being accumulated in fewer innings.

I don't recall Charleston having an unusually high number of HBPs--really, just a handful a season--so I'm not sure how much effect that would have. (You can check out <url=http://agatetype.typepad.com/agate_type/2007/03/1921_negro_nati.html>1921</url>, <url=http://agatetype.typepad.com/agate_type/2006/10/1922_nnl_plus_v.html>1922</url>, and <url=http://agatetype.typepad.com/agate_type/2007/04/1923_negro_nati.html>1923</url> stats at my blog.) The Shades of Glory stats give his games played, and by cross-referencing with other sources you can figure out how much time he missed, which I think Eric must have already done to create his MLEs.

It is true that the HOF project didn't count HBP (though it did count HB for pitchers)--to my knowledge, Patrick Rock and I are the only ones to have counted them (likewise with full fielding stats).

Let me try those links again: 1921 NNL, 1922 NNL, and 1923 NNL.

Also, Eric, do your MLE's for 1918-19 take into account the shortened season? I was translating them to 128- and 140-game schedules, and they are the two highest SFrac's of his career.

The reason why Charleston's WS look so spectacular is that Eric has him with 12,000 career plate appearances. By my count, a league-average player with the same split of time in CF and 1B would have over 300 WS. (WS experts, please correct me). These numbers don't put Charleston anywhere near #9 all-time if you use any sort of plausible replacement level.

Right, but the point is that I'm comparing him to a league average that does count HBP, so even if it's just a handful, the fact that he's not being credited with any reduces both his playing time and his productivity.

Dan's numbers for Gibson and Charleston track my subjective reads on their MLE's. I have Gibson as number 6 All-Time (behind Ruth, Wagner, Bonds, Williams, and Johnson) and Charleston as number 25.

No Cobb, Andrew? The Georgia Peach could play some ball.

sure he could, but the top 10 is awful crowded. And if you're like me and worry about your team chemistry after 20 years of Tyrus Raymond.....

1 Babe
2 Barry
3 Honus
4 Say Hey
5 Walter
6 Splinter
7 Hammer
8 Josh
9 the Man
10 Ty
11 Mick
12 Rocket

Right, but the point is that I'm comparing him to a league average that does count HBP, so even if it's just a handful, the fact that he's not being credited with any reduces both his playing time and his productivity.

His actual HBP rates for those seasons could help figure out what's missing from the rest of his career, no?

And if it helps, Charleston was plunked once in 273 PAs in the 1920 NNL, and zero times in 146 PAs in the 1927/28 Cuban League.

Cobb is 7th on my list, dragged down by his similarity to Speaker (who only manages 15th). For anyone who cares, the top 50 on my list are:

(1) Ruth
(2) Wagner
(3) Bonds
(4) Williams
(5) Johnson
(6) Gibson
(7) Cobb
(8) Mays
(9) Musial
(10) Mantle
(11) Gehrig
(12) Aaron
(13) Clemens
(14) Collins
(15) Speaker
(16) Hornsby
(17) Young
(18) Morgan
(19) Schmidt
(20) Alexander
(21) Lloyd
(22) Grove
(23) A-Rod
(24) DiMaggio
(25) Charleston
(26) Maddux
(27) Foxx
(28) Ott
(29) F. Robinson
(30) Mathews
(31) Joe Williams
(32) Spahn
(33) Stearns
(34) Vaughn
(35) Brouthers
(36) Connor
(37) Lajoie
(38) Henderson
(39) Berra
(40) Mathewson
(41) Bench
(42) Feller
(43) Ripken
(44) Paige
(45) J. Robinson
(46) Delahnty
(47) Hamilton
(48) Mize
(49) Piazza
(50) Boggs

Honorable mention: Bagwell, Anson, Brett, Rose, Nichols, Seaver, Gibson, Ray Brown.

andrew, you and I could be clones. You are just frappin brilliant!

Whoa, TomH, Mays above Williams? Do you not give war credit? I mean, yes, Mays played a brilliant center field and was a great baserunner, and Williams was an indifferent left fielder. But we're talking *thirty five points of OPS+* here--that's the hitting difference between Mays and Chet Lemon. And the gap is really even bigger than that once you factor in war credit and OBP-heaviness. Could I trouble you to walk me through that one?

Similarly, Cobb below Aaron and Musial? Is there a timeline there? Not knocking either Aaron or Musial here, but their value is just well short of Cobb's, on any of peak, prime, or career. How are you calculating these rankings?

And Andrew Siegel: Why does the existence of Speaker make Cobb less Meritorious, or vice versa? Is it not possible that two of the top 10 greatest players of all time happened to be born around the same time and played the same position? I imagine the same thing is going on with having Collins at 14 and Lajoie at 37, when Lajoie's value was so close to Collins's. Pretty harsh on Rickey too, although I suppose if you're very peak-oriented it would be consistent.

But we're talking *thirty five points of OPS+* here--that's the hitting difference between Mays and Chet Lemon. And the gap is really even bigger than that once you factor in war credit and OBP-heaviness. Could I trouble you to walk me through that one?

Don't know how I'd rank it, and I won't go through any math, but in addition to the fielding and baserunning which you noted, consider that Mays was in the notably stronger league while Williams spent half his career in a notably weaker league. Consider that for the time we have data, Williams earned a ton of IBB, and it can be assumed he received a ton more (plus some unintentional IBB) in earlier years. These all boost his OBP and OPS but don't increase run scoring as much. Consider that Mays was very durable, while Williams topped 140 games only once after age 30. Do all these things overcome 35 points of OPS+? I dunno, but there they are...

But Jesus, it's just incredible that Williams received MVP votes every season he played more than 6 games -- including seasons of 37 and 98 games!