OK, league strength is a legitimate point. Again, I think you would have to argue that the gap between the AL and NL was as big as the majors-AAA gap for it to bring Mays and Williams together, but maybe it was.

The difference between a UIBB and an IBB is only .08 runs--it's not very substantial . Most IBB's are mistakes.

The difference in durability would just be reflected in a peak comparison, which obviously favors Williams, esp. with his WWII years.

Lajoie dominated some leaks that were relatively weak during a time fo great player movement and league flux plus reports of his defensive value differ wildly from system to system. I might have him a bit low nonetheless.

I have Cobb #7, so I obviously think very highly of him. When comparing him to the others in the top 12 though, I think it matters that (1) there were many teams in his league who would have traded him even up for someone who played the same position and (2) a bunch of guys (Speaker, Collins, Baker Jackson) in addition to Cobb were putting up WS/WARP totals that would have lead most major leagues throughout history by a wide margin.

As for Henderson, well, two things. First, peak matters to me at the very top of the list of All-Time greats. Second, at the margins, being a pain in the ass counts too.

That should say Lajoie dominated some leagues . . .

Certainly a discount for the 1901 AL is in order, but my understanding is that the AL was stronger than the NL for the succeeding years...and his final great years were in the era of total AL dominance (like 1910). I don't see why Lajoie would merit a league strength penalty like Williams would (a generalized timeline effect, sure, but that would take down Collins as well).

#2 is just a standard deviation issue. Part of it is that all the stars were clustered in the AL--if you were to put half of them in the NL, each league's WS/WARP leaderboard wouldn't look so wacky. Standard deviations on the whole in the deadball era (across both leagues) were not overwhelmingly high. I really think it's a star glut--you just happened to have some great players born at the same time and stuck in the same league--rather than systemic ease of dominance (which would be the case in, say, the 1936 AL or 1895 NL).

The difference between a UIBB and an IBB is only .08 runs--it's not very substantial . Most IBB's are mistakes.

I'm not sure you can make that assumption when looking at one specific player. The truly elite hitters are pitched around more because of the fear of what they could do if they put the ball in play. While there's no guarantee that those IBBs are issued optimally, I think it's safe to infer that those walks are not worth as much to Williams or Bonds as to another player.

But yes, in a major league lineup, you can't assume they are adding up to tons of runs' difference. I looked at Bonds's '04 in BaseRuns. It's a little tricky, because BsR looks at the player in the context of his team's offense, so it's not as simple as just changing all his IBB to UIBB, because that changes the team's results too. But it looks like the difference between Bonds's IBB and UIBB was about .14 runs. Of course, that assumes that all of his UIBB were purely random and unwanted by the pitcher -- clearly not the case.

Mmm I think that's a weakness in BaseRuns--Smyth simply counts IBB as half-walks for on-base purposes. I definitely think a linear estimator like eXtrapolated Runs has the edge there, since it has a precise regression weight for IBB.

Been out of town for a week...

-I ran 1918-1919 at 154 games, not at the abbreviated schedule.
-My PA estimates for Charlston are not as accurate as I might claim some other MLEs to be. I don't have team games for a lot of his NgL years, so I'm not entirely comfortable making estimates. In general, I simply put in 600 neutralized PA for full-time years (which then varies based on the league's offensive conditions), which in a 154 schedule at 4.1 PA/G means I'm showing him playing 95% of team games in those years.
-I assume a 70% success rate for SB for all leagues where no CS data is available, and where I don't have other evidence of a player's SB successes.

I've recently rerun Charlston, to include his Cuban League data and to update the method where applicable. Some abbreviated numbers:



The influence of the Cuban stats helps Oscar. He had an excellent record in Cuba, and adding it into his neutralized career record also helps support a slightly higher profile for the regression PAs. In addition, his SB% was previously too high, and I've corrected for that this time around.

Can I get the full updated seasonal lines for him? Do you have #'s for Lloyd?

Oscar Charleston MLEs



Notes
Source of NeL data on which MLEs are based is the HoF project, published in _Shades of Glory_, except for
1916 Gary A. http://agatetype.typepad.com/agate_type/
1917-19 John Holway, _Complete Book of Baseball’s Negro Leagues_ (some stats estimated from surrounding seasons)
1921 Gary A. http://agatetype.typepad.com/agate_type/
1922 Gary A. http://agatetype.typepad.com/agate_type/
1923 Patrick Rock via Gary A. http://agatetype.typepad.com/agate_type/

Conversion Factors
BA -- .9 competition, league context adjustment seasonally determined
ISO -- .81 competition, league context adjustment seasonally determined
BB -- .95 competition, league context adjustment seasonally determined when possible
SB -- .72 contextual adjustment, based on comparisons of top 5 SB totals in NL and AL with actual and estimated top 5 SB totals from NeL

Games Played modeled on Tris Speaker
PA set to 4.2 PA/game, except for rookie year and decline years

BA, SA, and BB rate are regressed to 5-year norms
SB are not regressed

From 1916-1930, even years are projected into NL seasons, odd years into AL seasons.
From 1931-1937, all years are projected into NL seasons

Comments
The career findings here are not all that different from Dr. Chaleeko’s MLEs, but the payoff for doing the MLE work is a) a better look at Charleston’s peak (wow!), regressed walk totals, and stolen bases, which are new to this projection. Charleston in his prime was pretty clearly one of the best, if not the best, base-stealers in the Negro Leagues. Since stolen bases are discretionary, and rates are affected by a league’s “culture” as much as players’ ability, it’s really hard to say what would have happened if baseball had integrated in, say, 1915, just as Charleston’s career was beginning. This projection basically scales down the rate of the top 5 NeL base-stealers to match the ML rate. Because these rates are quite volatile in all three leagues under consideration, I didn’t use season-by-season adjustments, since the NeL rate varied from 85% of the ML rate to slightly above twice the ML rate in individual seasons, but they scattered among those levels without any evident patterns over the full set of seasons I studied (1916, 1920-37), so I took the average rate difference for the whole period and applied it consistently. When Charleston broke the curve in the NeL, he shows up as breaking the curve in the majors as well. And I think that’s appropriate.

Since I’ve been working on systems for doing MLEs for walks and stolen bases, I haven’t had time to do equivalent win shares, either fielding or batting for Charleston, but I hope this data, plus repuational info on fielding, will enable Dan R to calculate estimated WAR for Charleston, which will give a better view of his value than WS would. I will try to do estimated win shares, but I want to get offensive MLEs for Stearnes and Torriente first. They should be pretty quick, now that I’ve updated my methodology to handle walks and stolen bases.

I probably should put this comment on the not-yet-posted CF discussion thread; any responses to it should probably go there.

As I'm sure you noticed, we had no Negro League players on our just concluded LF ballot and will have none on our RF ballot, but we have a whole flock on our CF ballot. This trend was present among infielders - a better list of Negro League SS than of either 2B or 3B - but not quite as strikingly as for the outfielders. We seem to have a strong case of the best players and best athletes gravitating towards the most valuable positions.

And, of course, that reflects the unevenness of the competition in the Negro Leagues, and the relative shortage of HOVG players to occupy the premium positions. It seems likely that had the major leages fully integrated in 1915 (not a chance, of course, considering the dismal nature of U.S. race relations at around that time), some of our CF candidates might have drifted out to a corner outfield position. But which ones?

Chris Cobb, I'll need Fielding WS to get WARP for Charleston.

Another way to put the same question: suppose Stan Musial played in a league whose talent density, top to bottom, approximated the Negro Leagues of the 20's and 30's. He'd have been primarily a CF, wouldn't he? Actually, he would also never have completely given up pitching.

I'll try to get FWS done in the next couple of days, then. They're a lot quicker to do that bws, because they really are estimates based on reputation, documented career changes, and normal career fielding patterns of peak and decline.

Over on the Pete Hill thread, burniswright addressed OCF's question about which NeLers would have been moved off centerfield in the majors:

I would have moved Brown to an outfield corner, definitely. Perhaps Bell to left too, because of his weak arm. But then there's the Ashburn precedent, etc. And Irvin wasn't a centerfielder at all. The rest are legit CFers though.

#16 John Holway via Chris Cobb
1918 .429 for Ind ABCs (18-42); ba 1st; cf; all-star, MVP, drafted

perhaps served after the Indianapolis ABCs shut down for the year?

20. Gary A Posted: January 15, 2005 at 01:33 AM (#1081209)
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.

Evidently, if Charleston served in the military or in civilian war work, he missed none of the 1919 baseball season. (Civilians missing none --that is to be expected, I think.)

I haven't projected Charleston as missing time for military service. Holway says he was drafted, but his stats show the ABCs with 11 games and Charleston with 42 ab, which doesn't look like he missed any time to me. He also seems to have a full season of at bats for 1919.

Riley's bio does not mention any service time during WW1, though both Holway and Riley note that Charleston did a stint in the army, serving in the Philippines, before he broke in with the ABCs.

How're those Fielding Win Shares going for Charleston?

I should have them finished tonight. The CF ones are done; now I'm working on first base.

Oscar Charleston Fielding Win Shares



My main model for the center field win shares was Ty Cobb, who had a quite similar career shape to Charleston, hitting an early-30s decline offensively and defensively at about the same time, though Cobb’s was not as steep. Cobb was a good but not great centerfielder. Charleston’s rep was great, but there is nothing in the fielding stats from his prime to suggest that he was far above average. I have given him (matching Cobb) two great years to set his reputation, with average performance elsewhere, until his decline. I also looked at Cy Williams.
For first base, I looked at the performance of other HoM outfielders who switched to first in their 30s: Musial, Stargell, Yastrzemski, and Rose. All were as good or better at first than they were in the outfield, and all had little seasonal variation in their quality, so I projected Charleston as a steadily B+ first baseman, once he learned the position in 1930.

I think that’s it.

I also looked at Cy Williams.

Is there some parallel, Oscar || Cy ?

Chris, could I get specific estimates of 2B/3B/HR? I can fill in the XBH at league-average rates, but the 3B are going to affect how many baserunning wins he gets since I use them in my estimator.

I was looking at the center fielders who were, roughly speaking, Charleston's contemporaries, to identify the players who had been around average center fielders during the 1916-29 period when Oscar was in center. Ty Cobb was one player that I identified, and I focused on him, because there were so many other significant similarities. Williams was the other around average CF with a long career who came up. His late start made him not a close match in career shape for Charleston, but his career helped to reinforce the range of values I was getting from Cobb, and to show what a decline period that would lead to a shift out of center field would look like.

Also, I imagine you're just subsuming Charleston's hit by pitch into his walks. Or should I credit him with a league-average plunk rate in extra plate appearances?

Chris, is there a typo in those FWS? There are a bunch of years where you have him at far fewer FWS than FWS/1000, despite his playing far more than 1000 innings.

Dan,

I don't have a real methodology for doing that, but if I use an ISO adjustment to scale down each category of extra base hit and then do a little fudging, I come up with the following career hit distribution for Charleston:

2369 singles
520 doubles
185 triples
344 home runs
----
1049 xbh total

that gets to 5340 total bases, and matches, more or less, the percentage of doubles, triples, and home runs in his extra base hits that occurs in the actual Negro-League data, so I think that's a pretty fair estimate of how many triples he would have hit, assuming there would not be a huge "style of play" adjustment to the majors. But his triples total seems to fit the era pretty well, too, so I'd say it passes the sniff test, at least.

OK, your numbers suggest Charleston was a *real* asset on the basepaths--my estimator is giving him over 2 non-SB baserunning runs a year at his peak. That's Carey/Cobb/Billy Hamilton territory. Is your sense that he was THAT kind of threat? In the 1926 NL, for example, you have him with 54 SB, which would have led the league by 19 over Kiki Cuyler.

Yes, I think he was.

In 1926, he and Jud Wilson both tore up the NeL in stolen bases. Their totals, projected out to a full season, are 73 stolen bases each. After them, the next closest player who shows up in (admittedly incomplete) data set that I have is at 40. As far as I can tell, Charleston and Wilson were curve-breakers on stolen bases, comparable to Ben Chapman in the 1931 AL, who stole 61 with the next highest at 33, and the number 3 guy at 19. Wilson was at this level for just a few years, but Charleston was either among the leaders or above the leaders for a full decade.

Now, Charleston would not have stood out quite as much as baserunner, I believe, if baseball had been fully integrated, but, when he shows up as a big outlier in the NeL context, I felt that it was proper to project as an outlier in the MLEs also, and the only way not to do so would have been to go outside the formulas entirely.

His reputation as a baserunner says very fast, very aggressive, great instincts. The numbers bear that reputation out, as far as I can tell.

We can't know for sure until we see full league data for more seasons, and until we have confirmation that the box scores behind the data are consistently recording stolen bases for all teams, that the picture the partial data give us is accurate. However, for the few seasons we do have full league data, Charleston shows up as a league-leader in stolen bases, so my view is that, absent dispositive evidence, the MLEs should reflect the data and the reputation.

For fielding, I don't see the data (or the career path) matching the reputation, so there I have not projected Charleston according to his Speakeresque reputation. But I spent a long time reviewing the SB data, and I can't see any other conclusion without going against both data and reputation.

Got it. There's some typo in your Fielding Win Shares estimates, so I'm just multiplying the FWS/1000 (which I imagine are right) by G*.009. (e.g., in 1923, you have Charleston with 3.2 FWS, while multiplying FWS/1000 by G*.009 gives me 6.1). Let me know if that's not a valid approach.

Out of curiosity, why the big dip in his fielding quality in 1922?

And you didn't answer me about HBP--are they already included in his walks, or should I add them at a league average rate?

Those 2.8 FWS would have comfortably led the NL in 1931--is that intentional?

OK, well assuming that I'm correctly fixing the FWS, that Charleston's HBP are already counted in his walks, and that Chris really did intend to have Charleston be a very bad CF for the two years before he was moved and then a well above average 1B throughout the '30s, here's how he comes out.



3-year peak: 30.4
7-year prime: 61.2
Career: 104.9
Salary: $329,398,528. Among CF, that's well below the top 4 (Cobb 560, Mays 521, Speaker 502, Mantle 429) and a clear notch ahead of DiMaggio with full credit for 1935 in the PCL and three war seasons (306). Overall, it's below Mel Ott (350), roughly tied with Rickey Henderson and Cal Ripken (331 and 327), and above Frank Robinson (318) and Jimmie Foxx (306). Definitely inner circle, but more like a top 20 position player than a top 10 one. He will be an easy #5 on my ballot.

On typos in the fielding win shares: the rates and the fws totals were reversed for 1923-24: a hazard of hand-entering data and deciding to re-order the columns. Ouch.

Here is the corrected complete set.

Oscar Charleston Fielding Win Shares



On HBP: I don't include HBP in my models: there's too little data to make it possible to include them. I would add them at league-average rates, along with extra PAs.

On seasonal fws totals that are high or low in 1922 and 1931: there's no historical justification for these. I try to establish what a player's career rates at positions would have been and then vary them in a realistic way while holding to the shape of a career path. Charleston's fielding peak 1920-21 fits plausibly with a fielding career curve and the establishment of his reputation; his decline in 1927-29 fits what's happening in his hitting/baserunning stats, the anecdotes about his decline, and his shift to first base. Other than those features, the variations are more or less random.

I don't have any statistical evidence for the peak in 1920-21 but my rationale for it is that 1) Charleston's reputation suggests that he was brilliant, so it was most likely early in his career but not so early that he was still developing at the position, and 2) Cobb had a two year fielding peak at this same point in his career, after which his fielding tailed off to around average for the rest of his prime, at least as WS and BP WARP see it.

That said, re 1931: my _Win Shares_ book has Sheely and Terry with 3.0 fws at first base each, so I'm not sure on what basis you see that as an NL-leading total.

Because I was looking at the wrong year. Yes, 2.8 would be third in the league. But my basic point is that you're positing Charleston was about a +5 1B well into his thirties, and I just wanted to confirm that that's what you think. Here's a redone Charleston, adding league-average plunk totals to his line. Looks like they're worth a little over two wins.



3-year peak: 30.8
7-year prime: 62.1
Career: 107.2
Salary: $338,742,339. No change in his ranking either within CF or in the all-time group.

No idea why that formatted goofily on the years with negative BRWAA. The fact that Dr. Chaleeko's MLE's were giving me $330M is encouraging; the two of you are definitely describing the same player! Neither set of MLE's provides any serious justification for placing Charleston higher than one of the Big Four at this position, although he was pretty clearly superior to Joltin' Joe.

But my basic point is that you're positing Charleston was about a +5 1B well into his thirties, and I just wanted to confirm that that's what you think.

It seems to me most likely.

Neither set of MLE's provides any serious justification for placing Charleston higher than one of the Big Four at this position, although he was pretty clearly superior to Joltin' Joe.

I agree. I would have to go back and look at your numbers for the Big Four, but my sense is that Charleston in his prime was probably as good as they were, but the second half of his career is a much bigger drop from his prime than theirs.

I think the shift to first base alone would set him below the Big Four, but I am concerned that his play in the 1930s is being underrated, because we just don't know with any precision how contraction affected the competition levels in the Negro Leagues from about 1931 to 1937. I doubt it could make so much difference as to catch him up to Mantle, though I could easily see it catching him up to Ott at $350 million.

When we have full seasonal data for those years, we can begin to work on the problem . . .

Now I'll start to work on full MLEs for Turkey Stearnes, which should be ready in a couple of days.

Here are Charleston’s actual HBP rates, along with league contexts:

1915: Charleston 0 HBP in 77 PA (.000); ABCs in Cuba series totals 8 in 1521 (.005)

1916: Charleston 0 in 110 (.000); Negro League totals 80 in 9258 (.009)

1920: Charleston 1 in 273 (.004); NNL 151 in 15015 (.010)

1921: Charleston 4 in 272 (.015); NNL+ 312 in 28648 (.011)

1922: Charleston 1 in 363 (.003); NNL+ 167 in 20481 (.008)

1927/28: Charleston 0 in 146 (.000); Cuban League 27 in 3885 (.007)

1928: Charleston 0 in 259 (.000); eastern Negro Leagues 88 in 12722 (.007)

If Charleston were plunked at league average rates, he’d have 12.8 HBP over the (exactly) 1500 plate appearances I have for him during these years; he actually got 6, so he was hit at a little less than half the contextual rate.

Ah, that's great to know. OK, well, let's split the difference and put him at $334M then.

Perhaps this issue is addressed somewhere in these 137 posts on Charlie (perhaps even by me), but some historical perspective is absolutely essential in understanding Charleston's move to 1B in the 1930s and what it was about.

Luckily, we don't need second-hand opinions on this subject, because Charleston talked about it himself, on more than one occasion. In his ABC, StL and Harrisburg years, Oscar had the body of those muscled sprinters we just saw in the Olympics. There probably wasn't an ounce of fat on him, and he was the consensus greatest defensive CF of his era. Yes, he didn't have a strong arm, but it was awfully hard to hit anything over his head.

If you look at those stats, through the late 20s, you see exactly what Charleston himself said about his career: he had absolutely nothing left to prove. He had mastered every phase of the game, was clearly both its greatest defensive and its greatest offensive player (Gibson had not yet emerged), and yet his chances of moving up to MLB were zero.

A profound change came over him at that point, physically, psychologically and emotionally. Just as Rube Foster had done, 20 years earlier, he "let himself go," because there wasn't the motivation to test his mettle against the Cobbs and Ruths of this world that had driven him as a young man.

As you can tell from the pictures, he didn't just put on a few pounds, he was genuinely fat. And he transferred a substantial amount of his energy from playing to managing. In his own words, he still got motivated to be an RBI man, when the game was on the line. But you don't see the kinds of monster seasons anymore that he had in the 1920s. As to his ranking as an overweight defensive firstbaseman in the 1930s, I can't imagine--even with his athletic skills--that he was much above average.

What might be the proper evaluation of a career that clearly took place in two parts that didn't bear much resemblance to each other, I leave to the sabermetricians among you. For peak types, it's not a problem. For career value, I think it's a big problem.

--David Lawrence

I guessed higher than that for defensive value but basically had the same line everywhere else. Knocks him down a peg from 22nd to 25th all-time.

I forgot to include Charleston’s 1923 HBP figures (from Patrick): 2 HBP in 358 PA (.006); NNL+ 177 HB in 24051 PA (.007).

Also interesting are Charleston’s positions played in 1923, which I hadn’t noticed before:

1b46 cf38 p5 rf1

Ben Taylor had left the A.B.C.’s after a dispute with his sister-in-law, C.I. Taylor’s widow (who owned the team), and they didn’t really have anybody to replace him at first. So Charleston filled in for more than half the season, with mostly Namon Washington (31 games) and George Shively (11 games) taking his place in center.

Oscar Charleston MLEs Version 2.0



Notes.

All the data sources for these MLEs are the same as for the earlier version.

All seasons in the new MLEs are projected into National League seasonal contexts.

All conversion errors, such as adjusting a season an AL context and then projecting it into an NL context have been corrected.

The regression equations have been improved. Instead of basing the multi-season average to which single season totals are regressed on raw hits, total bases, and walks, they are now all based on average BA+, SA+, and walk rate+. This probably has the largest effect of any change, as it smoothes out peaks and valleys that resulted from big changes in offensive context from one season to the next. Charleston’s 1921 peak has dropped considerably, for example, but his trough in 1929-30 has filled in considerably also. His mid-1920s peak remains pretty much unchanged.

I have refined the formulas for projecting walks to be more sensitive to seasonal variations in walk rates.

Overall, Charleston’s career totals and his counting stats have been little affected by these changes, but some of his seasonal OPS+ totals have shifted quite a bit.

I think that these MLEs now employ both “best available data” and “best available methodology” throughout.

Version 2.0 MLEs should be available for Turkey Stearnes pretty soon, then I’ll get back to work on adding CWL data to Torriente’s MLEs, which I’ll also upgrade.

I haven't projected Charleston as missing time for military service. Holway says he was drafted, but his stats show the ABCs with 11 games and Charleston with 42 ab, which doesn't look like he missed any time to me. He also seems to have a full season of at bats for 1919.

Riley's bio does not mention any service time during WW1, though both Holway and Riley note that Charleston did a stint in the army, serving in the Philippines, before he broke in with the ABCs.

Charleston was indeed drafted in 1918, at least according to the Chicago Defender. In fact, C. I. Taylor's team was so gutted by the draft that he disbanded it in late August. The last instance I can find of Charleston playing in 1918 was in a doubleheader on August 18 against the Pennsylvania Red Caps (of NY) in Indianapolis.

Taylor sat out 1919 as well, and only reassembled his team when the NNL started up in 1920.

Charleston had served in the 24th Infantry in the Philippines, where he was reputedly the best pitcher and only black player in the "Manila League" (about which I know nothing) around 1914/15. Charleston had enlisted in 1912 at the age of 15 (lying about his age, obviously); his height was reported as 5'5 1/2". He was honorably discharged on March 20, 1915. Bullet Rogan served in the 24th in the Philippines at the same time, and was supposed to have played baseball there; no idea if they were teammates. Geri Strecker is working on a biography of Charleston which may contain some answers.

In fact, C. I. Taylor's team was so gutted by the draft that he disbanded it in late August. The last instance I can find of Charleston playing in 1918 was in a doubleheader on August 18 against the Pennsylvania Red Caps (of NY) in Indianapolis.

Taylor sat out 1919 as well, and only reassembled his team when the NNL started up in 1920.

So that explains why Charleston played for Detroit and Chicago in 1919, but was back Indianapolis in 1920.

Given that the majors ended on September 1 in 1918, Charleston being drafted didn't cause him to lose much playing time relative to the major leagues (did major black teams other than the ABCs keep playing on into September?), so I think just projecting him into a 118-game season makes sense, rather than messing about with war credit.

Chris, and I imagine there's no change to the Fielding Win Shares above?

Chris, and I imagine there's no change to the Fielding Win Shares above?

Correct. The only part of the offensive MLEs system that affects fielding value is the playing time estimates, and those aren't affected by any of the changes in this version.

OK, well, assuming there are no changes to the Fielding Win Shares or the ratios of 2B, 3B, and HR among XBH, and adding in hit-by-pitches in extra plate appearances at half the league average rate as per Gary A's posts, this comes out to:



3-year peak: 30.8
7-year prime: 60.3
Career: 104.4
Salary: $322,280,744, a completely insignificant dropoff from the prior MLE's. Still is the equivalent of one big year ahead of DiMaggio, even after crediting Joltin' Joe for his 1935 PCL season and the war years.

This has probably been answered elsewhere several times, and if so my apologies for having asked again for the thousandth time, but I am curious about the .9 BA adjustment and the .81 ISO adjustment used for Charleston's MLEs.

Are these adjustments based upon some empirical analysis and comparison of data, or upon some agreed-upon norms without hard statistical evidence?

Are these adjustments based upon some empirical analysis and comparison of data, or upon some agreed-upon norms without hard statistical evidence?

The .9 BA adjustment is based upon empirical data. For all the position players who played at least two full seasons in the Negro Leagues and in the major leagues, their averages in the two leagues were compared, with one transition season to each new level of competition being dropped and adjustments made according to normal aging patterns. As far as was possible given the data available when the study was done, adjustments were also made for league context. .9 was the average amount that NeL batting averages needed to be reduced to match ML batting averages, after all other adjustments were made.

At the time the study was done, less data was available for slugging than for batting average, so .81 was chosen based on expected relationships between batting average and slugging average. .81 was not inconsistent with the available data, but there was too little to rely on. When the full data from the HoF study is released, these conversion factors could be re-checked, but we need full data on the non-HoFers who played in both leagues--guys like Sam Jethroe and Luke Easter and Bob Boyd--to have a decent-sized data set.

The application of this conversion factor consistently throughout the period of 1916-1948 is based on an assumption that quality of play across this period was fairly consistent, relative to the white major leagues. That is almost certainly an invalid assumption for the one-league years of the mid-1930s, when competition levels were probably higher in the NeL due to contraction, but we don't have enough data to estimate year-by-year competition-level fluctuations in the NeL. For other NeL periods, these conversion factors produce results that look quite plausible, and for the 1920s, Cuban Winter League data, which includes play by major leaguers at various points, suggests that the competition level is about right, but there are too few data points to consider the matter closed.

Charleston had served in the 24th Infantry in the Philippines, where he was reputedly the best pitcher and only black player in the "Manila League" (about which I know nothing) around 1914/15. Charleston had enlisted in 1912 at the age of 15 (lying about his age, obviously); his height was reported as 5'5 1/2". He was honorably discharged on March 20, 1915. Bullet Rogan served in the 24th in the Philippines at the same time, and was supposed to have played baseball there; no idea if they were teammates. Geri Strecker is working on a biography of Charleston which may contain some answers.

Geri did a presentation on Charleston and Rogan in the Philippines at the SABR convention, which was by far the best presentation that I saw.

-- MWE

For all the position players who played at least two full seasons in the Negro Leagues and in the major leagues

I'm still trying to understand this, doing research of my own on 1923, and having few direct comparisons except for exhibitions, winter league play, and the occasional Cuban player who faced both top Negro League and white Major League competition, which I consider insufficient for comparison.

One of the efforts I'm making in comparison of 1923 is to eliminate games against the marginal teams, effectively raising the level of competition

So I'd like to know whether .9 is etched in stone or a working theory.

Was there a sufficient sample with this criteria? IOW, how many player/years were in the sample? Did it take aging of each player into account? Did it include the NeL in the early 50s, or did it cut off when the league's talent was obviously depleted? Did it take into account the higher standard deviation in the shorter-season NeL? Did it adjust for the difference in BAs etc. between each league?

Again, if these questions have been asked and put to rest already, then my apologies for asking, but a 10% discount on Charleston's BA (about 36 points) seems a bit steep. I haven't kept up with most of these discussions, so if there's something I can read to catch up, I'd appreciate it.

So I'd like to know whether .9 is etched in stone or a working theory.

Everything here is a working theory.

Was there a sufficient sample with this criteria?

By what definition of "sufficient"? All available data was used. This included all relevant seasons from Boyd, Campanella, Doby, Easter, Irvin, Jethro, Minoso, Robinson, Smith, Thompson, and Thurman, if I recall correctly.

Did it take aging of each player into account?

Yes, as I mentioned above.

Did it include the NeL in the early 50s, or did it cut off when the league's talent was obviously depleted?

The last NeL season used was 1948. I also applied wartime discounts for NeL seasons from the war years.

Did it take into account the higher standard deviation in the shorter-season NeL?

No, because it compared players' cumulative NeL averages to cumulative ML averages, both adjusted for aging patterns, so it did not work directly with seasonal data.

Did it adjust for the difference in BAs etc. between each league?

Yes.

Again, if these questions have been asked and put to rest already, then my apologies for asking,

No apology necessary.

All these questions have been asked and discussed multiple times, as there have been discussions of MLEs on and off here for the last four years, but they are scattered over many discussion threads, so it’s hard to find all the information about the MLEs in one place. If you want to read up on the whole matter, I’d suggest scanning the John Beckwith thread for relevant materials, and reading the Major League Equivalencies thread. The discussion on the Beckwith thread begins before I introduced systematic age adjustments into the MLE calculation: I don’t know where or when I posted the conversion factor study with age adjustments included (which raised the factor from .87 to .9).

http://www.baseballthinkfactory.org/files/hall_of_merit/discussion/john_beckwith/P0/

http://www.baseballthinkfactory.org/files/hall_of_merit/discussion/major_league_equivalencies

but a 10% discount on Charleston's BA (about 36 points) seems a bit steep.

Is that a career BA you are talking about for Charleston? If so, what is your source? Or are you looking at a seasonal total?

The actual reduction of a Negro-League player’s batting average is seldom going to be a full ten percent. The offensive context in the NeL was consistently lower than in the majors, so that usually gives back 2-5%. In some years, NeL players in strong pitchers’ parks actually have a higher MLE average than NeL average. My actual MLEs for Charleston (which you can see above) show him hitting over .400 twice, which isn’t shabby. The HoF data gives Charleston a career BA of .348: my MLEs project that as .324 in the majors. If he hadn't lost his competitive drive in the late 1920s (as discussed above), I'm confident that he could have had a career average 10 points higher.

All these questions have been asked and discussed multiple times

Perhaps we need a FAQ for beginners like me. :-)

By what definition of "sufficient"? All available data was used. This included all relevant seasons from Boyd, Campanella, Doby, Easter, Irvin, Jethro, Minoso, Robinson, Smith, Thompson, and Thurman

Since you were doing that off the top of your head, I'm assuming that Monte Irvin, Elston Howard, and Junior Gilliam were there also, and that Robinson was not.

While I agree it is an excellent method, I don't feel it's a sufficiently large and representative sample upon which to base that .9 adjustment with a lot of confidence. Some of those players had Negro League careers that were about as many games as a single MLB season; they were all from a single era, ignoring the changing styles of play in both MLB and the Negro Leagues, and the fact that those styles did not evolve in tandem.

Is that a career BA you are talking about for Charleston? If so, what is your source? Or are you looking at a seasonal total?

I'm just looking at the .325 in your MLE lifetime and dividing by .9, which leaves .361.

The actual reduction of a Negro-League player’s batting average is seldom going to be a full ten percent.

The further adjustments do indeed give back quite a bit, but it starts with a discount of 10%. Without the BA*.9 adjustment, his lifetime BA would be closer to .360 with those other adjustments. I assume they have a legitimate reason, and that they could deflate some players' numbers even further than 10%. IOW, the 10% reduction is for everyone, while the other adjustments are a bit more fine-tuned.

I am not a mathematician, and my opinions are just that, my opinions, and much less well-grounded than many on this board.

Here's something I found interesting: Without making any 10% discount, I took Charleston's batting average against only the other five teams that played the full 1923 NNL (KC, Chicago, Detroit, St. Louis, and the Cuban Stars); adjusted for park effects and then adjusted strictly for the difference between the NNL and MLB for that year, he came out with a BA of .361. Same as yours for that season. Go figure. Maybe you're right.

Oscar Charleston's Real Stats

Hey, folks,

Please see my latest MLEs for Oscar Charleston. Not much surprising here. He's rating out as one of the four or so players with the most legitimate claims to being the GNLOAT (Greatest Negro Leaguer of All-Time). It's not spoiling anything to reveal the others appear to be Satchel, Pop Lloyd, and Josh Gibson. There are definitely arguments for others as well, especially Martin Dihigo and Turkey Stearnes. As information becomes available, they and others might surge. It seems as though Smokey Joe Williams, Bullet Rogan, Willie Wells, and Pete Hill for now round out the top 10 candidates given the basic results. Once you adjust for schedules and whatnot, YMMV, especially for Monte Irvin. And, if you're an extremely peak-oriented voter, Dobie Moore might rise up for you as well.