The reason I ask this is because that basic model disappears for the 30 years between about 1895 (A/B/C) and 1925 (G/F/etc). (The 1b-man drought.) Is it because there weren't any players like them to be found (even if they weren't quite as good)? Or is it because managers agreed there was no place for them in the deadball game?

I'm betting on the latter.

Having read gadfly's posts carefully, and _most_ of the subsequent discussion, here are few points to toss out.

1) gadfly's seasonal data for the NeL is of great value, and I will need to incorporate it into the conversion factors.

2) Irvin's career alone, even if one accepts every step of gadfly's analysis (I'm not sure about all of it yet -- I need to give more thought especially to the matter of the transition year and the Polo Grounds' effect), needs to be buttressed by other data.

3) It's clear from the seasonal data that the NNL and the NAL should not be lumped together; there were differences in style of play and almost certainly differences in quality, with the NNL being the stronger league. I will try separating the two and see if the differences in conversion factors are consistent.

4) I agree that Eric's point about differing skill sets is true in theory, and it addresses similar issues to those that Andrew Siegel raised. I agree with gadfly for the most part, however, that differences in the leagues small enough that they do not call the whole project of find MLEs into question, especially given that we are trying to find MLEs for the great players who have more than enough talent to make the jump to the majors, not the good-to-very-good players with holes in their game.

From this perspective, knowing exactly what the quality of the league as a whole was is irrelevant; we want only to know what the performance of the great players in the NeL would have been like in the majors. That's what the data we have can show us most readilly, in any case. For overall league quality, the NeL is going to be a "weird" league by white baseball standards because it included a much wider range of talent than the more highly stratified white leagues would have. We are interested in a conversion factor that fits the top stratum of the NeL.

Finally, I think we already know, in a general way, the type of player who could star in the NeL but not in the majors -- the great-field, decent-hit infielder whose batting game relies heavily on speed. The batting conversions drop these players out of the running for the HoM quite consistently. The only NeL great whose skill-set is of questionable conversion value is, as Andrew mentioned, Cool Papa Bell. Most of the guys we're looking at, even Dick Lundy, the lightest hitting of the serious candidates, had very good power by Negro-League standards and, if they had good speed, they didn't rely on it as their primary offensive weapon.

In # 173 Gadfly wrote:

However, I have done two larger studies, that are more on point (and the Monte Irvin study simply confirmed one of them); and was referring myself to them.

In the first study, I analyzed the Major League players who went directly into the Mexican League in 1946. This quickly got out of hand and I ended up doing a study that compared the relative strengths of the Major Leagues, Negro Leagues, Triple-A Leagues, Lower Minor Leagues, and Cuban and Puerto Rican Winter Leagues through the many players that changed between these Leagues.

The conclusions of the study were that the Negro Leagues were comparable to Triple-A (actually slightly better). The Monte Irvin study simply confirmed this study. In fact, the results were so close that they startled me.

This study sounds interesting -- and important! Has it been published yet? We'd be very interested to see a summary.

Or is it because managers agreed there was no place for them in the deadball game?

I'm betting on the latter.

Were the managers correct? Would he have been any better than a bigger left-handed version of Cravath, patroling the tiny RF in Baker Bowl and hitting a few extra HRs?

Karl Magnus-

I know this is useless because you are only interested in arguing and not discussing, but you are wrong. If you're a member of SABR, just go read the Sporting News archives like I did from 1950 to 1954.

Irvin came up in 1950. To begin 1951, Durocher had Irvin batting clean-up. He then lost faith in him after he didn't begin to hit right away. One of the main story lines of the Giants 1951 season was Durocher's manic juggling of the line-up for the first half.

To start the second half, Durocher put Irvin back in the clean-up position and the Giants took off. When Durocher got credit for the comeback, he said that it was nice but he was the guy who had screwed the team up so bad that they needed to comeback.

From that point on until the emergence of Mays as a superstar in 1954 when he got out of the army, Durocher and the Giants obviously considered Irvin their best player. Durocher, by the way, has two autobiographies on the subject you could read too (Dodgers and Me, Nice Guys Finish Last).

Of course, you could research this, eh? But why waste that time when you can just accuse someone of trying to get the starting line-up of the 1926 Harrisburg Giants elected to the Hall of Fame en masse.

Jim D & John Murphy-

I guess I should have made this clear, but I would consider the 1900-1920 deadball/spitball peroid a very extreme outlier of Major League history.

And, of course, these is a total absence of the Gigantic 6 foot 2, 230 pound sluggers in this time peroid from the Majors. But the weird thing is that they are absent from the Minors too. That is something I never understood.

I think Gehrig would have still have been a great hitter at that time, but nowhere near as good as he was in the 1930s. The type of hitter worst affected would be a fly ball hitter. Gehrig (like Honus Wagner) hit line drive rockets.

Oddly enough, there were two (three if you count Santop) of those guys behind the Color line from 1900 to 1920 (Julian Castillo and Bill Pettus) but none before 1900. Maybe it's just an odd demographic result from discrimination.

Brent-
That study is just my own private insanity and it is too endless to post it, but I'll try to summarize it when I get a chance.

Gadfly, in 1951 Bobby Thomson, not Irwin was their best player (150 OPS+ vs 147- Irwin's best season) in 1952 Irwin played only 43 games and had an OPS of only about 120, in 1953 I agree Irwin was the best position player on the Giants, though not by much, and in 1954 Mays was king. If you TREBLE Irwin's black and gray ink you still get only 12 black and 99 gray, nowhere near a HOMer or HOFer. Beckwith, I have now been persuaaded is a borderline case, Foster looks well south of the border to me (I would have ` elected Rixey about 50 before Vance), and Irwin isn't even close.

We'll just have to agree to differ. Unlike Yest, I find the MLE conversions very useful, but they need to be taken with a BIG pinch of salt, and discounted somewhat.

I guess I should have made this clear, but I would consider the 1900-1920 deadball/spitball peroid a very extreme outlier of Major League history.

I don't believe in outliers when it comes to an absence of talent that long (it's actually more like 30 years, BTW) unless there was some kind of restriction against them.

Five years? Ten years? I can see that. But thirty? No way. IMO, management had no use for Gehrig-type players. Whether they were correct is another matter (which jimd hints at in post #104).

Karl Magnus-

I will happily agree to disagree with you. In any argument, it is necessary to have Sceptics, Critics, and Cynics to keep the argument honest. Of course, a Sceptic with an open mind is a beautiful thing.

And I would like to make a correction to a previous statement. You can read "The Dodgers and Me" all you want but it won't shed any light on Monte Irvin since it only covers up to 1948.

John Murphy-

If you limit the argument to just 1B, then it is a 30 year drought. But I was referring to a class of players, 6 feet and up and well over 200 pounds that can really slug, that includes some left fielders and even right fielders.

There is a bunch of these guys in the 1880s: Dan Brouthers, Roger Conner, Sam Thompson, Dave Orr, perhaps Ed Delahanty, and their Minor League counterpart Perry Werden. Without checking, I think Thompson was the last one out of the Majors in 1898 if you don't count Delahanty (though Sam played a little in 1906 and Werden played into the early 1900s in the minors).

In the 1900s and 1910s, the sluggers are all slighter smaller. Honus Wagner, Frank Baker, two guys that got really fat (Charlie Hickman and Harry Lumley), and the prototypes of the period: Buck Freeman and Harry Davis. I think the only really big guy was Tim Jordan (6 foot 2 and 190 or so).

(Freeman, by the way, is listed at 5-9 or 5-11 and 169 or 175 pounds. But, if you ever see any pictures of the early championship Red Sox, you'll notice that he was built like a wall. If he was 5 foot 10, he's gotta weigh about 200.)

All of these guys prove that you could still slug the ball in 1900s and 1910s. But there were no behemoths, not even in the high minors that I know of. If it's true that the Majors were discriminating against these guys, some of them would have had to end up in the high minors; but there's nobody as far as I know.

The first returning over 6 feet and over 200 pound slugger is not Gehrig in 1925. It's Ruth. It's almost like God was saving up to make something really special.

If it's true that the Majors were discriminating against these guys, some of them would have had to end up in the high minors; but there's nobody as far as I know.

But the minors may have had the same mindset as the majors.

I think size in and of itself is not the real issue:

According to bbref

Joe Jackson, 6'1", 200 lbs.
Jack Fournier, 6', 195 lbs.
Hans Wagner, 5'11", 200 lbs.

Lou Gehrig, 6'0", 200 lbs.

Insofar as Gehrig is different from these guys, it's not in body mass, but in footspeed and agility, and Gehrig was a lot like Jack Fournier in both, though he was probably a better fielder, or the standards for first-base defense had dropped considerably by 1925.

Two points from post 210 that strike me.

1) "Gehrig was a lot like J Fournier in both[footspeed and agility]."
Given that Fournier's skill set was very clearly not well liked by the Chisox who farmed him out and the Yanks (IIRC) who purchased him only to farm him out again, we might very well conclude that between the midteens and mid20s something had changed in how 1B were viewed. Which by way of seguing...

2) "...or the standards for 1B defense had dropped considerably by 1925."
With more emphasis on the fly ball, this would make a lot of sense, but I think there's potentially a second reason why 1B defense might have become less important. The ball.

Just as hitters could perhaps not see the ball as well when it was grimey, dirty, and misshapen because it never came out of play, a first baseman would need good reflexes to handle a similarly grimey, dirty, misshapen ball when it came from the left side of the infield against the dark backdrop of empty seats, dark-colored fences, and the convervative colored mens suits in the stands.

With new balls in play more often, by the mid 1920s, managers may have been more willing to take a chance with a slugger at 1B.

Anyway, just a little theory.

With new balls in play more often, by the mid 1920s, managers may have been more willing to take a chance with a slugger at 1B.

Anyway, just a little theory.

At the very least, it probably had some effect on management's thinking in regard to first basemen.

Good "catch," Doc. :-)

John,

I agree with you, but I don't think this is any reason to put guys like Beckley, Sisler, Chance in the HOM. I actually don't recall if you support these guys for the reason of a talent gap, just wanted to put that out there.

If it is true that 00's managers had a bias against Gehrig and Burkett-like sluggers, then we coudl then conclude that guys like Beckley would not have been the best at their position. They were just the best that didn't have a certain bias against them.

Either way, while I acknowledge that it is weird that there are so few outstanding 1B from ABC to Gehrig/Foxx/Terry, I also dont' think we need to elect any of them. There was a lack of outstanding 1B in that period after all.

And if Sisler hadn't contracted that nasty sinus infection, he would most likely have sailed in. But thems the breaks.

I disagree 100%. If this project doesn't put ANY 1st basemen from a 20 year period in the HOM, I think the credibility of the whole project is compromised.

Baseball is a game of 9 positions, and you have to have all 9 positions filled to play a game. If you elect no catchers or no first baseman over such a long time period, then there is something wrong with the methods used.

Either Chance or Beckly or Taylor or someone was the best at their position over that time, and they should be elected.

Same for catchers - Bresnahan or Schang or someone....

I disagree, I see this as analogous to the Joe Sewell situation and I don't buy it. I want to elect the best 200-250 players ever. If none of them happened to play 1B from 1900-1920 then so be it.

I want to add that Chance, Sisler, and Fournier were all HOM talents who had their careers cut short. Had they lasted a few more years then we might not be having this discussion. However, they didn't and I dont' want to elect the next level of 1B.

jschmeagol:

I guess I'm somewhere between yours and KJOK's view of merit. Regardless of their numbers, if two first basemen exceed their position's average by the same amount, they're both helping their teams equally. I know it's a difficult concept to get a hold of and it's only fairly recently that I have to this conclusion, but it seems logical to me.

But that doesn't mean that players should be elected to the HOM by default either. Of all of the first basemen from the Inside Baseball Era, Beckley has only made the middle of my ballot. Of course, I'm revamping my ballot system for the umpteenth time, so who knows where that era's first basemen will wind up on my ballot.

I want to add that Chance, Sisler, and Fournier were all HOM talents who had their careers cut short.

What about Taylor - he had a long career?

I disagree, I see this as analogous to the Joe Sewell situation and I don't buy it. I want to elect the best 200-250 players ever. If none of them happened to play 1B from 1900-1920 then so be it.

Yes, it's EXACTLY the Joe Sewell situation. The problem is that the relative value of positions is NOT FIXED over time (the assumption that WARP3 makes) but CF was less valuable defensively than 1B at one time, 2B less valuable than 3B, Catcher much, much more valuable defensively than any position, etc. during different time periods.

The question is how to define "Best" 200-250 players. If Player A was the BEST at HIS POSITION during his time, and Player B was the 5th best at his position during his time, just because Player B has better "stats" (Wins Shares, WARP3, etc.) overall doesn't mean he should be ranked ahead of Player A...

What about Taylor - he had a long career?

Yes! Vote for Taylor, he will solve this whole dispute. :)

I say that kiddingly, but I do think he is HoM-worthy and deserves a close second (third?) look from the electorate. He certainly seems to have been the best Black 1st baseman of the Teens. And, whether or not you agree with, or like, the Negro League translations, 325 estimated Win Shares is noteworthy.

KJOK,

I almost put Taylor in the list but I was at work and didn't have time to look up and make sure. He should be in that list as well.

I guess it is entirely possibly that the 00's and 10's weren't an era without good 1B or an era that misused the position. Instead it could have been a era where every great player that played the position (Chance, Taylor, Fournier, Sisler) had very bad luck and wasn't able to finish their career as they should have.

As I have stated before, I disagree on Sewell. he was better than his peers but a)it wasnt' a terribly strong field, b) Lundy was better and a peer, and c) he wasn't head and shoulders above them.

The thing with Sewell AND with the 1b is that NOBODY stands out. If there was a standout and an otherwise weak field that would be fine. But there are no standouts. Yes, Beckley was good for a long time but he was almost never the BEST 1B and it wasnt' a strong field. Sisler and Chance were great but had short careers. In fact I think this weak field makes a guy like Konetchy standout a bit.

As I have stated before, I disagree on Sewell.

I also disagree about Sewell, but because there were better players in the Negro Leagues during his time. He wouldn't have stood out as great at short if the major leagues had been integrated at that time.

I guess it is entirely possibly that the 00's and 10's weren't an era without good 1B or an era that misused the position.

Except the span of time is even longer than that. I just can't buy it.

The thing with Sewell AND with the 1b is that NOBODY stands out. If there was a standout and an otherwise weak field that would be fine. But there are no standouts. Yes, Beckley was good for a long time but he was almost never the BEST 1B and it wasnt' a strong field. Sisler and Chance were great but had short careers. In fact I think this weak field makes a guy like Konetchy standout a bit.

I obviously disagree that the quality was inferior, but I do agree that nobody stands out to the same extent as a Gehrig, Foxx, Greenberg, Mize, Anson, Brouthers or Connors.

Yes, it's true nobody stands out, but I think sometimes we're looking at this with the historical perspective that there's an era coming with McGwire, Bagwell, Palmeiro, McGriff, etc. when what a "sabermetric" voter in 1944 would be looking at from a historical perspective would be something like:


CAREER
1876-1938
1B

RCAA RCAA
1 Lou Gehrig 1251
2 Dan Brouthers 967
3 Jimmie Foxx 842
4 Roger Connor 807
5 Cap Anson 730
6 Harry Stovey 485
7 Bill Terry 425
8 Frank Chance 348
9 Jack Fournier 341
10 Jake Beckley 330
11 George Sisler 328
12 Hank Greenberg 325
13 Dave Orr 324
14 Henry Larkin 318
15 Jim Bottomley 282
16 Ed Konetchy 230
</pre>

and I think a voter would look very favorably on inducting some of these guys...

In post # 184 EricC wrote:

Because of selection bias, we arrive at the incorrect conclusion that league A is stronger.

I'm sorry to keep coming back to this argument, but now I see another flaw in it. You're assuming that integration causes players to move from league A (Negro Leagues) to league B (Major Leagues). But when we imagine a hypothetical integrated league, I don't think we should imagine that only the Negro League players would be adjusting. Certainly when integration finally did take place in the 1940s, 50s and 60s, it changed the type of play in the majors. Jackie Robinson, Willie Mays, and others brought a much faster, more aggressive style of play to the majors. While I don't think we can attribute the return of the stolen base entirely to integration, it certainly was a factor.

It seems to me that this implies that when you compare players in leagues A and B to a hypothetical integrated league "AB", selection bias would impact on both groups of players. Just as some players from league A wouldn't have the right mix of skills to succeed, the same would be true of some players from league B. Selection bias would operate in both directions.

John Murphy wrote:

Regardless of their numbers, if two first basemen exceed their position's average by the same amount, they're both helping their teams equally.

Sorry, John, but IMO this argument doesn't hold water. The position's average is irrelevant - what matters is that your team accumulate more value over all nine positions than the other teams accumulate from their players. If the other teams all have great first basemen, you can still beat them if you have a great right fielder, pitcher, or whatever. There is absolutely no requirement that teams compare or match each other position by position. If the league average first baseman is one win above replacement and yours is two wins above replacement, that doesn't make him an "above average" player.

And while KJOK is right that the value of positions hasn't stayed fixed over time, I don't think you can identify the relative value of the positions by the average player at the position at a particular time. Call it a small sample problem or whatever, but just as I believe that the two greatest second basemen played simultaneously from 1916-28 (and, peering into the distant future, the two greatest third basemen will have virtually simultaneous careers during the 1970s and 80s), there also can be periods of two or three decades when there aren't any great players at a position.
Greatness is not uniformly distributed.

Or would a voter in 1944 be asking himself
a) what happened to the days of the thumping first basemen?
b) how come Sisler and Chance had to go and get hurt?

To tag onto the Sewell comments above, the same might be said about Lundy. Once Lloyd moved off shortstop, it was left to Moore, Beckwith, and Lundy to duke it out for the best NgL SS of the era. Moore broke his leg, Beckwith moved to 3B, that leaves Lundy. I'm probably oversimplifying, but in a funny way it kind of works the same for King Richard as it does for Sewell.

I think Sewell's case is somewhat analagous to the 1B situation too in a quriky way. As has been pointed out, Fournier, Chance, Sisler each had some bad luck. And we should mention that Hal Chase, a leading 1B, didn't play it straight and got banned. So too Hollocher, Wright, Chapman. Plus Weaver (who saw a lot of action at short) was banned.

Overall, however, I'm not averse to the idea that we don't elect a 1B from this entire period because while we believe that talent imbalances and position "luck" should trend at least somewhat smoothly, it's possible that several things could conspire (or randomly occur?) to create a long-term talent shortage. We've already identified a bunch of them:
1) Shift away from boppers at first and toward defense/athleticism
2) Possible need for better fielding due to increased bunting and dirty ball
3) Bad injury luck at position
4) Pre-Rickeyian lack of understanding of the relative importance of OBP and SLG as opposed to AVG.

In fact, I think you could make a convincing argument that the late-90s/oughts may have seen somewhat of a talent drought, but that the drought was lengthened and exacerbated in the teens beginning with Chance's retirement, extending to Fournier's minor-league exile and Sisler's sinuses, then ending with a bang when Gehrig, Terry, and Foxx arrived.

But again, I'm just weaving facts together to find a through-story among probably random events. Perhaps the truth lies elsewhere?

Dr Chaleeko wrote:

With more emphasis on the fly ball, this would make a lot of sense, but I think there's potentially a second reason why 1B defense might have become less important. The ball.

Just as hitters could perhaps not see the ball as well when it was grimey, dirty, and misshapen because it never came out of play, a first baseman would need good reflexes to handle a similarly grimey, dirty, misshapen ball when it came from the left side of the infield against the dark backdrop of empty seats, dark-colored fences, and the convervative colored mens suits in the stands.

With new balls in play more often, by the mid 1920s, managers may have been more willing to take a chance with a slugger at 1B.

Interesting comment. I wonder what other changes might have affected fielding conditions. Does anyone happen to know when the modern first baseman's glove came into use?

It also appears that bunting dropped fairly dramatically. Checking the Stats All-Time Handbook, it looks like the modern rule for recording sacrifice hits was used from 1894-1907, and then again from 1931-38 and 1940-present. In 1907 there were 2,574 sacrifices in the two leagues; in 1931 the total was down to 1,437.

The reality is thast Beckley's numbers stack up very well with long career outfielders of his era or later -- he's significantly better than Van Haltren or Ryan (higher OPS+, longer career) and had a much longer career than Duffy. Also stacks up well with Wheat, Terry and Goslin, all of whom we have enshrined or soon will. Therefore, when you add in the difficulties of playing 1B in his era, and the fact that WS and WARP both substantially underrate 1B fielding for his era, he should become a no-brainer. Chance, Sisler and Fournier had the possibility of becoming top tier HOMers taken away by injury; Beckley actually did it, and it's time we recognized this fact.

Interesting comment. I wonder what other changes might have affected fielding conditions. Does anyone happen to know when the modern first baseman's glove came into use?

I don't know about the gloves, but I would betcha that groundskeeping technologies may also have contributed to lowering the error rates in MLB. I wish I had a sense of the timeline, but I do remember reading years ago in Green Cathedrals about how Commiskey Park was built on an abandoned landfill/dump. One day a routine bouncer quoiponged over Luke Appling (IIRC). He asked the groundscrew to inspect the spot where the ball took the funny hop, and sure as sugar they discovered an old tea kettle or pasta pot or something lurking right below the infield dirt.

Of course, that's just one incident, but the first big wave of permanent stadium building hit between 1907ish and 1925ish if I recollect Lowry correctly. With most every team getting a new park, I wonder if infield conditions may well have improved over the previous generation of hastily constructed wood/brick stadia immediately prior.

To state something that's probably obvious...if you combine the possibilities of:
-improving field conditions
-more new balls being put into play
-possible increases in glove technology
-fewer bunts (as a strategic concept, but also, perhaps, because it was more difficult to place a bunt with fewer deadened, misshapen balls in play)
-fewer steals
-more fly balls (due to upper cuts and also possibly due to the denoumont of the spitter)...

and shazzam, it's no wonder error rates fell, corner defense became less important, and DP rates started climbing.

If the other teams all have great first basemen, you can still beat them if you have a great right fielder, pitcher, or whatever. There is absolutely no requirement that teams compare or match each other position by position. If the league average first baseman is one win above replacement and yours is two wins above replacement, that doesn't make him an "above average" player.

If those two first basemen were the best for that season, then the replacement level needs to be adjusted, doesn't it? IMO, each position's replacement level is not at a constant level throughout baseball history. It changes depending on when a position's responsibilities have been transformed or when management's perception of a position has changed.

Interesting question. Do we know the relative decline rate of infield errors versus outfield errors -- should distingiuish between the effect of better gloves (presumably universal) and the effect of fewer teakettles on the playing field (should primarily affect infielders.) This would give a good index of how the relative difficulty of playing certain positions changed over the years.

Error rates by position
(number of errors per 1000 chances)

1B OF 3B 2B SS C P
1871-75 069 186 250 175 210 173 168
1876-80 048 157 162 117 127 128 102
1881-85 041 143 162 108 142 101 125
1886-90 027 099 135 084 124 088 100
1891-95 023 086 120 072 106 068 089
1896-00 022 062 103 060 089 051 084
1901-05 020 053 088 056 079 035 062
1906-10 016 042 072 050 071 032 054
1911-15 014 044 067 048 070 032 056
1916-20 012 037 056 041 063 027 046
1921-25 010 035 052 038 055 026 041
1926-30 010 032 050 034 055 024 041
1931-35 010 029 054 033 052 018 042
1B OF 3B 2B SS C P
1936-40 010 028 055 033 050 020 044
1941-45 010 023 054 033 049 020 039
1946-50 010 022 048 026 039 017 042
1951-55 010 021 047 024 040 015 043
1956-60 010 021 048 023 038 012 040
1961-65 010 023 051 024 038 012 046
1966-70 009 023 049 023 036 012 046
1971-75 009 021 050 023 036 014 051
1976-80 008 021 047 022 035 016 049
1981-85 008 019 050 019 036 014 044
1986-90 008 020 052 019 032 012 045
1991-95 007 018 050 019 031 010 045
1996-00 008 018 049 019 030 010 044
1B OF 3B 2B SS C P
Overall 013 034 066 036 054 026 054</pre>OF breakdowns available by position
1996-00 LF 020 RF 021 CF 014

Jimd, helpful and at first sight indicates that 1B and outfield fielding improved at about the same rate. However, the number of chances per season may well have varied between periods as well, and I think the trend in errors per season is equally important. Or is it -- what's your view?

Actually, that's a slightly odd table. I see why 3B and SS should make most errors -- they get most chances -- but why should they make most errors per 1000 chances?

karl,

because 3B and SS are throwing positions?

why should they make most errors per 1000 chances?

First thought is that the error rate might reflect the difficulty of the chances accepted.

Countering this is the concept that there is an acceptable error rate - that is, perfection is not required - as long as other benefits are gained. (See the constant error rate for 1b-men post 1920 despite the introduction of the big-pocket gloves at some time after that; my guess is that the fielding may have been at the point of diminishing returns and the gloves allowed better hitters with less fielding skills to play the position acceptably instead.)

Another thought is that fielding ground balls requires two distinct plays (at least) - fielding the ball and throwing the ball; an error can be charged on each one. However, baseball accounting counts it as only one chance.

I'm sure other people will come up with more thoughts.

Another interesting thing: the position with the highest rate is 3B. But that isn't true from 1908-1933; SS has the highest rate for this period. (Note that there are also anomalies; SS is higher in 1903 and 1944, while 3B reclaims the distinction in 1913 and 1931. Also in the late 1870's, C in 1876 and OF in 1877 and 1878.)

Offhand, I on't know what any of that means but musings are always welcome.

SS had the highest error rates in the 1920s Negro Leagues, too, by a slight margin over 3B.

Kudos to KJOK for the post of RCAP leaders at 1B. It exactly captures my thinking on 1st baseman, to wit:

I. The guys about 700 RCAP (Gehrig, Foxx, Brouther, Connor, and Anson) are all-time great no-brainers.

II. The next group of guys (all the way down down in the 400's) (Stovey and Terry) are likely selections based on their bats, though not strong enough that you don't want to look at defense, league quality, peak, intangibles, etc. Looking at those factors, Terry's fabulous defense puts him over the in/out line. On Stovey, intelligent minds can disagree after they apply an AA discount, but I'm happy that we picked him.

III. Then there are the seven guys between 318 and 348. Orr and Larkin can be dropped fairly easily based on quality of competition and defense. If you give war credit, Greenberg probably moves up out of this group and has a big enough career number to earn election (particularly given his peak). I think the other four (Chance, Beckley, Sisler, and Fournier) are fairly close and that a case can be made for excluding all, including all, or drawing the line somewhere in between them. This is how I rank them:

(1) Chance--the leader among them in RCAP, he was also an excellent defensive 1B, played a lot of innings at C, has more intangibles than any other player we've seen, was the heart of arguably the greatest team of all-time, has an excellent peak, and put up that RCAP total in a shorter amount of time. Hard for me to see why you would promote any of the others over him.

(2) Beckley--third in RCAP, he only grades out as a good defensive 1B, but he played at a time when 1B had more defensive value overall, numbers somewhat hurt by the stronger 1B competition for his first few years, he's better than his RCAP number.

(3) Sisler--fourth in RCAP, he was a good defensive 1B early in his career, but poor for the second half; he had a nice peak, but not as strong as it looks on raw stats; unless the numbers are wrong, I think he's very much in the discussion but has to rank below both Chance and Beckley.

(4) Fournier--second in RCAP, but a very poor defensive player; when the defense is factored in, he still belongs in the group but has to rank at the bottom of it; the weird thing is that unless the numbers utterly fail to model early 20th-century baseball correctly, his bat was good enough to give him a Hall of Fame career despite his putrid defense if his managers had just held their nose and put him out there.

IV) The next two guys, Bottomely and Konetchy (who brings his great defense to the table), are stars on the Hall of Very Good All-Stars. (I subjectively peg Ben Taylor to Konetchy, but I might be wrong.)

Those are RCAA numbers, not RCAP numbers. So, it have any measure of positional competition... only league-wide competition. (i.e. Beckley was as much hurt by Delahanty as he was Brouthers).

(3) Sisler--fourth in RCAP, he was a good defensive 1B early in his career, but poor for the second half; he had a nice peak, but not as strong as it looks on raw stats; unless the numbers are wrong, I think he's very much in the discussion but has to rank below both Chance and Beckley.

It's possible that career RCAA does not actually show us how good Sisler was at his peak, if he was consistently below average in the second half of his career. I would _guess_ that if you looked at his pre-infection years only, his RCAA would be closer to Terry and above the Beckley/Fournier/Chance group. To be fair, you'd need to drop their decline phases as well, but neither Fournier nor Chance had much of one, and Beckley's was also not long. It's likely, then, that Sisler would pass them on this measure if none were docked for the parts of their careers in which they were below average.

It's possible that career RCAA does not actually show us how good Sisler was at his peak, if he was consistently below average in the second half of his career.

The following numbers are proportional to a context-scaled version of RCAA. The best 7 years of each player's career:

Chance: 78 66 66 52 41 29 27
Terry: 69 68 59 53 47 38 37
Sisler: 70 68 51 46 44 37 35
Beckley: 38 36 34 29 27 24 20</pre>

Chance played in the lowest run environment of all of these, so context-adjusting the RCAA pushes him to the head of the line here. I don't have him in my top 15 because he has only 1200-some games.

Fournier on the same scale as my last post:

66 60 59 57 43 26 17

Definitely not consecutive: the 66 is from 1916, while the next three are 1923-25.

I meant 1915, not 1916. And why are we doing this on the Beckwith thread and not a first baseman thread or Sisler/Chance/Beckley thread?

John Beckwith Fully Regressed Win Share estimates

These win shares use the same playing time and the same fielding win shares as my last estimate, except that a 1935 season of 3 games has been added. Systematic regression lowers Beckwith’s career total slightly from my unsystematic regressed estimate, which was 317.7. His peak seasons are slightly lower, but I think his career value above average is little changed.

John Beckwith’s win shares

Year (games) BWS + FWS = Total
1919 (18) 0.3 + 0.4 = 0.7
1920 (127) 13.6 + 4.3 = 17.9
1921 (149) 19.0 + 5.0 = 24.0
1922 (131) 15.8 + 4.1 = 19.9
1923 (144) 23.4 + 4.3 = 27.7
1924 (146) 26.1 + 4.7 = 30.8
1925 (140) 24.2 + 4.4 = 28.6
1926 (117) 20.0 + 3.2 = 23.2
1927 (151) 22.0 + 4.1 = 26.1
1928 (135) 19.3 + 4.0 = 23.3
1929 (146) 18.5 + 4.0 = 22.5
1930 (100) 15.1 + 2.6 = 17.7
1931 (149) 21.5 + 3.0 = 24.5
1932 (100) 13 + 1.9 = 14.9
1933 (95) 10.8 + 1.3 = 12.1
1934 (54) 0.4 + 0.8 = 1.2
1935 (3) 0.0 + 0.0 = 0.0
17 (1905) 263.0 + 52.1 = 315.1

These full process for calculating these estimated win shares:

1) Assemble fullest, best data available from MacMillan 8th & 10th editions, Holway, and Gary A.’s postings.

2) Apply park factors and offensive-level adjustments to NeL totals. For Beckwith’s seasons with the Chi Am Giants, I treated his park factor as 90. For his 1930 season with the Lincoln Giants, I treated his park factor as 110. For 1931 and 1932, when he played partly with the Lincoln Giants and partly with Newark, I treated his park factor as 105. For all other seasons, I made no adjustment. Since Beckwith played for so many different teams, it seems unlikely that his statistics are skewed by a majority of seasons in pitchers’ or hitters’ parks. As for offensive-level adjustments, I have started with the NeL at .9 of the offensive level of the majors in 1919 and 1920, and then increased the ratio by 1% per year to 1928. From 1929 on I have treated the offensive levels as the same in the NeL as in the majors.

3) Convert Beckwith’s NeL BA and SA to MLEs, using .87 as the BA conversion factor and .82 as the SA factor.

4) Regress Beckwith’s MLEs, using rolling 5-season samples to establish the mean to which each season is regressed, cutting these down to 3- and 4-season samples at the ends of careers.

5) Use the regressed MLEs to find the closest matches among major-league players, paying attention to OBP and SB as far as possible, given the limits of our data. Prorate the major-league players’ win shares to Beckwith’s level (typically adjustments of 5% or less up or down) and to Beckwith’s playing time.

I’ll put the key data for the MLEs themselves in the next post.

John Beckwith career NeL data

Year AB Hits TB BA SA
1919 16 3 4 0.181 0.250
1920 46 10 13 0.217 0.283
1921 134 53 80 0.396 0.597
1922 159 48 72 0.302 0.453
1923 270 82 148 0.304 0.548
1924 119 48 79 0.403 0.664
1925 264 107 205 0.405 0.777
1926 119 43 79 0.361 0.664
1927 309 112 163 0.362 0.528
1928 70 24 34 0.343 0.486
1929 163 62 109 0.380 0.669
1930 67 33 59 0.493 0.881
1931 166 58 113 0.349 0.681
1932 12 4 8 0.333 0.667
1933 9 4 4 0.444 0.444
1934 19 3 4 0.158 0.211
1935 4 0 0 0.000 0.000
17 1946 694 1174 0.357 0.603


John Beckwith MLEs for BA and SA

AB for
Year MLE BA MLE SA Regr. Reg. BA Reg. SA
1919 0.181 0.228 196 0.282 0.384
1920 0.210 0.257 355 0.281 0.389
1921 0.378 0.538 625 0.351 0.510
1922 0.317 0.448 728 0.324 0.480
1923 0.316 0.537 946 0.331 0.550
1924 0.373 0.579 931 0.352 0.573
1925 0.371 0.670 1081 0.355 0.610
1926 0.327 0.567 881 0.338 0.553
1927 0.325 0.446 925 0.331 0.486
1928 0.304 0.406 728 0.323 0.474
1929 0.331 0.548 775 0.326 0.522
1930 0.390 0.656 478 0.346 0.581
1931 0.290 0.532 417 0.302 0.540
1932 0.276 0.521 273 0.300 0.529
1933 0.387 0.364 210 0.297 0.457
1934 0.137 0.173 44 0.164 0.213
1935 0.000 0.000 32 0.123 0.133
17 0.327 0.521 0.330 0.526

</pre>

As you can see, the regression formula I use raises Beckwith’s career rates by about 1% each over the straight MLE values, mostly I believe from the beginning and ending of his career. I don’t think this is a significant source of error, but I’d like to find a way to systematically adjust the ascent and decline phases to avoid it.

I’ve spent the weekend setting up Excel to do all the calculations to get from the raw NeL data to regressed MLEs. Since that is done, I should be able to produce MLEs for other hitters more quickly, and to revise estimates quickly if further study calls for changes in conversion values.

Getting from BA and SA MLEs to win shares still requires work by hand, so that (and getting the raw data into Excel) will be the choke points in the process.

I hope to have new estimates for Dobie Moore and Dick Lundy soon, and have the system ready to put Turkey Stearnes and Mule Suttles through in short order when they become eligible.

Thanks again, Chris.

Nice work, Chris, thanks!

Chris - I wonder if the regressed estimates may be flattening out the career profile a little bit "too" much, especially for peak voters who pay a lot of attention to the player's best seasons. With the old numbers (which already reflected some regressing), his 3 best seasons were 33.5, 29.9, and 28.9, now they 30.8, 28.6, and 27.7. Somehow I don't think Beckwith actually had the consistency of a Sam Rice.

These are probably the best possible estimates for any individual season, but voters who care about peak should probably add a little variation to these numbers to allow for some bigger peaks.

Brent,

Since we're dealing with a hypothetical construction, it's hard to say what is too much or too little "flattening."

Here's how I would put it. The regression certainly underestimates the impact that Beckwith would have had in his best seasons with a direct translation of his play into an equivalent number of big-league games. The unregressed BA and SA numbers show what the direct translation would look like: I think that voters who want the unregressed career can get a pretty good idea of what it looked like.

It is possible that Beckwith would have sustained that higher level of play throughout a full major-league season: it is possible that he would not. Regression to a multi-season baseline will necessarily pull his performance downward from its highest moments.

This will affect Beckwith most, as you note, on isolated top seasons. Beckwith's 5-year consecutive peak is completely unaffected by the regression.

I will note in passing that we have never had published regressed seasonal data for short-season players from the first two decades of professional white baseball, so voters using season-ajdusted win shares to compare Beckwith on peak to, say, Pete Browning or Charley Jones should keep in mind that their peak seasons have not been regressed.

Also, now that you've gotten me thinking about this matter, I am beginning to wonder if there is still a better way to do this regression . . . May post some ideas on the MLE thread in the next day or two as I have time. Lots of papers to grade this week . . .

With 5 minutes more thought and a little number crunching, I'm pretty certain that the regression is too flattening.

In short, I think I need to set up the regression to predict the extent to which Beckwith _would have reverted_ to his 5-year norms over the projected at bats for the rest of the season, which then need to be averaged with his actual at bats, in which his performance level is set.

I'll need to talk things over with the statistics folks, but I'm guessing there's going to be another iteration of the regressed values yet to come. We'll see.

In any case, Beckwith's career value is set, pending any changes in the MLE conversion factors, and we're dealing with the details of how to best estimate the exact height of his peak. The unregressed BA and SA MLEs will also be stable, so they can also provide a more fixed point of reference.

Chris Cobb:

Some comments on your Beckwith Win Share Estimates:

1) 1935 Career End
Previously in this thread, I argued that John Beckwith was still a Major League hitter into the late 1930s. I have been researching the issue and, although the sample size is small enough to be in error, the available evidence confirms that Beckwith did not hit well in 1934 and 1935. I now have no qualms about the early ending of his career.

2) Systematic Regression
As Brent pointed out, systematic regression most definitely flattens out career years and peak seasons (Of course, it also flattens out, in the opposite direction, injury years and off seasons). Perhaps, a study of normal win share variation of Major League players would give a normal range of variance that could be used to give a plus-minus percent to these figures.

3) Beckwith Statistics
In the published 1929 American Negro League Statistics, Beckwith's stats were AB 192, R 57, H 85, 2B 15, 3B 2, HR 15, SB 3, BA .443, SA .776. I would consider these the best statistics for that season. I've done a lot of work on the 1929 ANL and the published figures appear to be quite accurate (with the exception of the Fats Jenkins' mistake in the official averages: he was credited with hitting .317 when his totals equal .371).

4) Negro League Park Factors I
You use a .90 Park factor for Beckwith's 1922-23 years at Schorling Park with the American Giants and a 1.10 Park factor for Beckwith's 1930 year at the Catholic Protectory Oval with the Lincoln Giants. In both cases, these factors are way too conservative.

Although I understand why you wanted to be on the conservative side, I would suggest a BA-SA factor of .90-.75 and 1.10-1.25 for these two parks. These two parks are so incredibly extreme that I still think these figures are too conservative, but that would be much closer to reality.

[This is also something to keep in mind when you do your proposed Mule Suttles MLE.]

5) Negro League Park Factors II
You use a park factor of 1.05 for Beckwith in 1931 and 1932, stating that he played half-seasons with the Lincoln Giants each year. This is simply incorrect. After the 1930 season, Beckwith tried to become the Lincoln Giants' manager at the expense of John Henry Lloyd.

Beckwith lost the power struggle and did not play for the Lincoln Giants (or more accurately the Harlem Stars and Black Yankees) in 1931 and 1932 at all. The 1.05 adjustment is inappropriate.

6) League Offensive Level Adjustments:
I assume that your League offensive level adjustment (.90 in 1919 and rising .01 each year until 1.00 in 1929) is based on the League data posted by Gary A:

1921 NNL .263 .361
1928 NNL .278 .384
1928 ECL .279 .384

While I understand why you did this, I think the straight line incline is not right. The Negro Leagues definitely were behind the Major Leagues in changing from a deadball to powerball offense than the Majors (and never went all the way), but I don't think it was this gradual. I think they caught up quickly.

Gary A has also posted the following Runs Per Game information:

1920 NNL 4.49
1921 NNL 5.05
1922 NNL 5.12
1923 NNL 5.48

The Negro Leagues trailed the Majors, but they were obviously heading in the same direction.

Also, though this is just my opinion, the Negro Leagues were concentrating their talent in the 1920s. By using the straight line incline and ignoring the fact that the League talent level is rising, the data is being skewed upwards.

This is evidenced by the fact the your win shares estimates have Beckwith peaking at ages 23, 24, and 25. Perhaps Gary A has more data that could clarify this.

Also, you must remember that the Negro Leagues seem to always have been a lower slugging enviroment than the Majors, with batting averages and runs scored artificially raised by a greater error rate (i.e. some are called errors, but some are called hits).

For example, here are the full 1928 League stats:

BA-SA-Isolated Slugging Average
1928 .281 .397 .116 AL
1928 .281 .397 .116 NL
1928 .279 .384 .105 ECL
1928 .278 .384 .106 NNL

[Note: weird as it seems, the AL and NL did have exactly the same BA and SA stats in 1928.]

The batting averages are virtually the same, but the slugging in the Negro Leagues is decreased by 10 percent.

If you have to make an adjustment, it probably should be a straight .95 BA-SA (in other words, multiply NegL BA-SA by 1.05) conversion for 1919 to 1922 to make up for the fact that the Negro Leagues lagged behind with an additional consistent .90 conversion factor for the typical decreased Negro League SA (in other words, multiply NegL ISA by 1.10 or SA by 1.03 or so).

7) Conversion
I must admit that I am disappointed that you continue to use the .87-.82 BA-SA conversion formula that your own original study came up with (though I will admit that I have the same typical human resistance to changing my own conclusions once I have reached them).

Meaning no offense, that study was crude and highly biased, relying far too heavily on unadjusted data from Larry Doby and Monte Irvin.
I posted the Monte Irvin study to show you this.

Consider this: My more careful analysis of Monte Irvin's best two Negro League seasons versus his best two Major League seasons came up with conversion factors of BA .933 and SA .877; despite the follwing six biases against the Negro League statistics:

a) Irvin's two Negro League seasons were in his prime, the two Major League seasons post-prime.

b) Irvin had four full previous adjustment or experience years to his two Negro League seasons versus just one full adjustment or experience years to his two Major League seasons.

c) Irvin's Negro League park factor, which surely inflated his Negro League stats, was left out of the equation.

d) Irvin's slugging in both Negro League seasons was artificially inflated by me.

e) Irvin was injured in one of his two Major League seasons, downgrading that year's Major League stats.

f) By equalizing Irvin's Major League park factor's, any normal Home Park advantage was eliminated from his statistics.

There were, as far as I can tell, no biases against the Major Leagues in that study, only the Negro Leagues. And, although I didn't add it in, these is the additional bias against the 1940s Negro Leagues that the Major Leagues of Irvin's time in the 1950s were of obviously higher quality simply becuase of integration.

I hope you understand that I think that it is great that you have taken the time to work through these MLEs. You seem to be about as idiosyncratic as I am. So I would like to ask you a favor.

Could you run a Beckwith MLE with the all the above modifications and a .95 BA-.90 SA Conversion Factor?

I would be very interested in seeing those results and if my theory that Beckwith was about equal to Hornsby was correct.

Of course, if it's not too much trouble, you could do both sets of equations for all the Negro League players; and, just so that it was clear which was which, you could name the .95-.90 and .87-.82 Conversions the following:

1) The Chris Cobb Best Possible Estimate Negro League MLE Conversion.

2) The Karl Magnus Double Secret Probation Triple Penalty For Playing Baseball While Black Conversion.

I would appreciate it.

Gadfly,

Thank you for the comments.

Quick responses as needed:

2) I hope that a different approach to regression will substantially solve the flattening problem. I think that regressing the additional projected at bats while giving credit for actual at bats exactly as they translate should work, though I'll have to retool substantially to run the numbers. It that doesn't work, some study of major-league seasonal variance would be the next step.

3) I'll work in the 1929 stats. Without looking at my spreadsheet, I think they're not very differnt from the ones I have used for that year.

4) I'm hoping for more data from Gary A., at least on the Chicago park factors before I commit to a park factor larger than 10% in one direction.

I know this will be a big issue for Suttles, Bell, and Wells. (I also wonder if there were hitters on the Chi Am Giants who were really great, but who are not remembered that way because of the park effects. I don't worry too much about Bingo DeMoss, because I think you could boost his offense by 25% and it still wouldn't be major-league average. The experts have never fingered anybody as having suffered as a result of the park, but I do wonder . . . )

5. On the NY Lincoln Giants: Saying that he was with the NY Lincoln Giants those seasons and that I was adjusting for that was an error on my part. I have him with Newark for part of those seasons, and had assumed the Newark team played in the same park that was later a strong hitter's park. That was the basis for that adjustment. If that team was not in a hitter's park in those years, then that adjustment is indeed incorrect.

6. I'll look into this: I don't trust the run levels on their own, because of the error factors.

7. I'm not set on the .87/.82 conversion factor, but I simply haven't had time to re-examine all the evidence you've provided, which I need to do before committing or not committing to make a change. Until I can myself offer a clear explanation for any change I make, I can't offer it as my interpretation of the data.

That said, while many changes in the system will be laborious to implement, changing the conversion factor itself is quite simple, so I could easily produce data for a .95/.90 conversion factor and post it, and have no objection to doing so. The numbers I can produce quickly would still have the regression issues and the offense-level issues and the park factor issues, of course, until I can address those as well.

I'll do what I can to get more data out. I'm working on 1924 PFs, but it might be a little while before I get them done (I need Birmingham and Memphis papers). I will try to finish a closer analysis of 1920-23, one that takes into account road parks, etc. Maybe within a week or so.

As for American Gts players who might have been underrated due to park effects: offhand, I'd say Walter (Steel Arm) Davis, Pythias Russ, and maybe Alex Radcliff. Jelly Gardner's career is interesting, too--he's actually one of the NeL career leaders in walks. However, he wasn't a power hitter, the category that would have suffered most in Schorling.

I agree about DeMoss, though it may be that his offense was closer to Newt Allen than you might think looking at his raw stats. Torriente was certainly much better than his unadjusted stats make him look (and they make him look pretty good)--during his prime, he rivalled Charleston as one of the very best NeL hitters.

Just to recap from earlier in this thread: raw park factors I found for the Am Gts:

1920: 85
1921: 54
1922: 66
1923: 81
1928: 57

Mean PF: 68.6

For I think 3 of these seasons, the Am Gts played many more games at home than on the road, so the effect would be much stronger than in an ordinary, balanced schedule. I will have more details in a few days.

Since Chris has used a PF of only 90 for Am Gts players to this point, this should make it clear that there's no "rounding up" going on. Quite the opposite.

This is evidenced by the fact the your win shares estimates have Beckwith peaking at ages 23, 24, and 25.

My interpretation of the data is that Beckwith did peak early. If we use a pf of 80, say, for the Chi Am Giants in 1922 and 1923, I think Beckwith's 4 best years are going to be 1922-25. He was still a dominant player after that with a fabulous year in 1929 (the season that's _really_ flattened by the regression), but I think his years with the American Giants and the Black Sox were, as a group, his best, unless there are park factors for Harrisburg we don't know about.

raw park factors

One thing that is very confusing when discussing "park factors" is that people use the same term to refer to different quantities.

My guess is that "raw park factor" refers to what I call the "home/away ratio"? (though "raw park factor" is also descriptive as long as I know they are the same thing).

In Win Shares, James uses the term to refer to what I call a "schedule adjusted park factor" (though his calculations only approximate this for unbalanced schedules, assuming that they were balanced). IIRC, this is also what BP uses, though they average over a different time period to calculate theirs. I assume that this is what Chris is using (being Win Shares), though I could easily be wrong.

Total Baseball/Baseball-Reference.com uses the term to refer to what I call a "opponent adjusted park factor" where they also compensate for the Yankee hitters not having to face their own pitchers and vice versa. (How they do this for unbalanced schedules is a good question.)

This discussion would be easier to follow if I was sure which "park factor" people were talking about.

Thanks.

Chris Cobb-

The Newark Browns' home stadium was Sprague Field. Off the top of my head, I am pretty sure it was a hitter's park; but I doubt it was to the extent of the Catholic Protectory Oval or even Ruppert Stadium.

From my research, I am pretty sure Beckwith peaked in 1929 and 1930, although the latter season was unfortunately cut short by injury. In both of these seasons, Beckwith was absolutely crushing the ball. And I think I know why.

For some reason (perhaps his Chicago pool hall), Beckwith usually didn't play winter ball, taking his winters off (and, as one of the highest paid players in the Negro Leagues, he could actually afford to do this).

[He did play in Cuba one winter; but went late, left early, and apparently didn't enjoy himself.]

But in 1928-29 and 1929-30, Beckwith played in the California Winter League. The Winter League had Negro League stars, Pacific Coast League Stars, and even Major League Stars. Beckwith was absolutely the best hitter out there.

John Beckwith, California Winter League
YEARS-G-AB-H-2B-3B-HR-BA-SA-TB
1928-29 019 071 022 04 01 05 .310 .606 043
1929-30 027 101 049 09 00 14 .485 .990 100
2 YEARS 046 172 071 13 01 19 .413 .831 143

From: The California Winter League, William McNeil

I think Beckwith peaked in 1929 and 1930 because he was playing year round and had no time to get out of shape or off-form.

Of course, Beckwith can be directly compared with a bunch of known white Major and Minor League hitters in the California League; but there is the obvious sample size problem.

However, with that said, only two white players can hold a candle to his statistics: Earl Averill and Wally Berger. And both of these players compiled their statistics in much smaller sample sizes.

In other words, regression to the mean would have taken them out of Beckwith's class if they had the same number of at bats as Beckwith (172).

AB-H-TB-BA-SA Player
054 027 048 .500 .889 Earl Averill
051 022 057 .431 1118 Wally Berger
131 052 088 .397 .672 Smead Jolley
095 035 057 .368 .600 Buzz Arlett
268 094 148 .351 .552 Babe Herman
182 059 091 .324 .500 Bob Meusel
399 125 207 .313 .519 Irish Meusel
239 074 112 .310 .469 Ping Bodie
048 014 020 .292 .417 Heine Manush
044 012 012 .273 .273 Max Carey

I look forward to the .95-.90 Beckwith MLE.

Gary A:
Interestingly, there is a way to compare Torriente directly to Charleston. They both played in Cuba, pretty much same park, same league in 1920-21, 22-23, 23-24, 24-25, and 25-26. Their statistics are:

500 129 177 33 14 10 .354 .536 Charleston
719 131 251 41 17 12 .349 .503 Torriente

That's damn close. I always rate Beckwith and Charleston as the best Negro League hitters in the 1920s. Apparently Torriente is closer to them than I thought.

Two Other Random Notes:

Note#1
Some Final ECL Stats were actually published in the 1925 and 1926 Reach Guides. Beckwith finished second in each batting race behind John Henry Lloyd (.444) in 1924 and Oscar Charleston (.438) in 1925.

John Beckwith
YEAR-G-AB-R-H-BA
1924 30 108 31 45 .417
1925 42 155 38 65 .419

Note#2
Bingo DeMoss has always fascinated me. Usually, if the qualified contemporary observers of the Negro Leagues said a player was a .350 or .300 or .250 hitter, that player would turn out to be, when researched, a .350 or .300 or .250 hitter.

DeMoss is the exception to this rule. Everyone said he was a .300 hitter but he sure looks like a .200 hitter when you study him.

I've always wondered if there were three factors consiring to destroy his reputation: park factor, walks, and sacrifices.

In his best known years, DeMoss, who was known as the Negro Leagues' greatest bunter, was the number two hitter for Rube Foster's Giants, a team that played a walk-sacrifice-bunt offense to an absurd degree, and played in Schorling Park, a pitcher's paradise.

Walks and Sacrifices are two things that analysis of Negro League box scores certainly has trouble picking up. Could this effect be so large that DeMoss's career stats are being unfairly ruined by it?

One interesting thing to note is that DeMoss' statistics pick way up in 1926 when he goes to play and Manage Indianapolis. If I remember right (though I could be wrong), he was usually the number 3 hitter for the ABCs, who had a woeful hitting attack.

Another thing to note is that DeMoss was really born in 1885 and actually over 40 in 1926.

To return to an old issue regarding Beckwith: can anybody supply the sources that tell us Beckwith was a poor fielder? I'd always thought so, too, and assumed that Riley must have savaged his defense along with his character. Then I checked Riley's entry again, and all he says that he was versatile enough to play any position (which he says more than once), then remarks at the end that "he did not excel at any position."

As I've said before, the only remarks about his defense I've found in contemporary papers are positive ones (about his play at shortstop). I've only checked 1921, 1924, and 1928 (I'm currently looking at 1922 and 23), so it's not exactly a comprehensive look. Still, I think it would be a mistake to assume Beckwith was a bad fielder without better evidence. He played nearly his whole career at either third or shortstop, with catcher and first base way behind in terms of games played.

So--does Holway say Beckwith was a bad fielder? Does it appear somewhere in an oral history or interview?

Btw--despite misunderstandings on the part of some, I'm not trying to push Beckwith's candidacy--I'm just interested to finding out the truth about him.

Gary A.,

I don't have any evidence for you. I think people have that idea because it's implicit in the picture painted of Beckwith in Riley and, following Riley, the NBJHBA.

Without saying directly that Beckwith was a poor fielder, Riley certainly creates that impression:

"Combined with a severe drinking problem and an often lazy, unconcerned attitude about playing, his characer deficiencies often negated his performance value. Sometime he would play in an inebriated condition or exihibit meanness on the field. On one occasion, when Beckwith's error cost a game, pitcher Bill Holland tossed his glove in disgust and, in the clubhouse, Beckwith responded by knocking the pitcher unconscious."

And, of course, the passage you referred to:

"He possessed sufficient versatility afield to play almost any position. However, he did not excel at any position."

I'd say that the first comment paints a picture of Beckwith as a talented athlete who didn't consistently play up to his abilities, especially on defense, but it offers only one piece of anecdotal evidence.

The second seems frankly contradictory. How many players are there who are talented enough to play any who aren't very good at any of them?

Now James:

"Most of the Negro League third basemen were like major league third basemen pre-1930 -- that is, they were defensive players first, closer to shortstops than to modern third basemen, who tend to be slow power hitters. Negro League third basemen were small and fast, chosen first of all for their ability to field a bunt."

"Beckwith and Boojum Wilson were exceptions to this rule. Beckwith was a real boomer. He hit .480 in 1930, with 19 homers in 50 games; also hit 16 homers in 51 games in 1931, which still led the league [note that James does not follow the corrected stats for Beckwith in 1930 from Macmillan 10th]. Built about like Bobby Bonilla, he was by far the biggest third baseman, adn one of the biggest and strongest men in the league."

James, by contrasting Beckwith to players who were fielders first, creates the impression that Beckwith was not a particularly good fielder. By comparing him to Bobby Bonilla, whom we know to have been a big man who was a bad (though determined) third baseman, he extends that impression.

He goes on to quote the first part of the first passage I quoted from Riley, cementing the impression with a bit of what is apparently more reliable evidence. And thus a view of Beckwith as a fielder is established.

For the HoM, I expect that James's comments have done the most to shape the electorate's view of Beckwith, at least prior to the presentation of additional data and biographical evidence on this thread.

Chris--very interesting look at how Riley and James shaped an image of Beckwith. I hadn't quite thought of it that way, but of course you're right.

James in particular I had forgotten about. One thing to think about, though, is that, as far as I can tell, he relies entirely on Riley, Holway, a few other historians (Phil Dixon, Donn Rogosin) and some player biographies--basically, the same material that's available to all of us. It's not apparent that he consulted any contemporary newspapers, so it's safe to say he didn't do any original research.

James is quite a good storyteller, so he's able to create some powerful images. It just bears remembering that the material he fashions them out of is secondhand.

An interesting comparison would be to Martin Dihigo. Dihigo's defensive profile, in terms of positioning, looks very similar to Beckwith's: 3B appears to be his most frequent position through the mid-30s, followed by SS. He also played a good deal at 1B and in the OF during these years. (My sources don't have actual games played at each position, so it's hard to tell.) (Beckwith, btw, played very little in the OF--I've actually never seen a box score with him in the OF).

They played side-by-side on the '28 Homestead Grays (Dihigo was 23, Beckwith 28, fwiw).

In the few games I have against ECL competition, here are their games, innings played, and stats at each position:

Dihigo: g/di/rf/fpct
ss-9/60/4.50/.968
2b-5/41/4.39/.952
3b-4/31/2.03/.778
p-2/17/0.53/.500
1b-1/9/---/1.000

Beckwith:
ss-12/98/5.51/.923
3b-6/47/1.15/.857
c-1/8/---/1.000 (1 sb allowed)

Very small samples, but at least it gives a decent idea of how they were used on the same team.

Gary A-

If I remember correctly, the Homestead Grays were not really happy with their middle infield defense in either 1928 and 1929. Of course, Beckwith was playing SS and 3B for them in those two years.

Basically, there was some talk that the team needed a defensive wizard like Dick Seay for a middle infield that was composed mostly of sluggers (Beckwith, Britt,Cannady, Dihigo, Scales). However, whether this need was real or just one of appearances is debatable.

In the 1929 American Negro League, the Grays had a disappointing season. They expected to contend but did not. There was more talk of the team's infield problems and the Grays traded Beckwith early in September of 1929.

As far as I can remember, that is the only time that I saw Beckwith's defense questioned. Usually, it was lauded and his versatility praised to the skies (the Negro Leagues had an absolute fetish for versatility because of roster size).

All this being said, I'm not sure that it qualifies as any proof at all that Beckwith wasn't good defensively. The Grays' problems in 1929, as far as I can tell, were injuries and their ancient pitching staff; not the infield defense. It's an old story but often true that teams incorrectly identify their weaknesses and blame their best players.

Also, Cum Posey was a world-class sore loser. If I remember correctly, when the Grays didn't do as well as he thought they should have in 1929, he expressed his unhappiness. Although it was not said outright, I remember that the impression I got was that Beckwith told Posey where to shove it and was traded away.

Beckwith was not a classic Dick Seay type SS (small quick guy - Rey Ordonez is a good comp for Seay). Beckwith was a Cal Ripken type with the ability to play back because he had a cannon for an arm. Sometimes in baseball, images are more important than reality.

Beckwith surely was not the prototypical SS or 3B of his day.

Also (on Beckwith not playing OF):

In 1931, Beckwith played virtually the whole season for the Baltimore Black Sox as their RF (more proof of his arm strength). In September of that year he joined the Newark Browns and then barnstormed with Hilldale in October if my notes are correct.

I see this a lot: Beckwith = Dick Allen. Or something like that.

In #242 above, Chris Cobb gave us his latest WS estimates for Beckwith. The peak seasons are:

31..29..28..26..24..24..23..23

The discussion points out this probably understates his peak, while hitting the career number accurately. If we take these numbers, adjust to 162 games AND add 5%, I get this:

34..32..31..29..27..26..26..25

Career total goes from 315 to 331.

Is this equal to Dick Allen?

Adjusting 1972 to 162 games, these are Allen's peak years in WS:

42..41..35..33..32..29..29..24

His adjusted career total is 344.

Beckwith looks good, but he falls short of Allen.

Dan G -

Your point is, of course, a good one. In fact, if Beckwith was as good a hitter as Dick Allen, his WS should be higher than Allen's for two reasons:

1) The overall quality of the Majors was almost indisputably higher in Allen's time.
2) Beckwith generally played more demanding defensive positions than Allen.

In my opinion, there are two reasons that Chris Cobb's WS projections underrate Beckwith:

1) Chris Cobb's conversion rates are about 10 percent two low; and
2) As I understand the math, a ten percent reduction in BA and SA is geometric, not arithmatic increase.

In other words, by increasing the offense and reducing the outs, a 10 percent increase in BA and SA would create something like a 21 percent increase in WS. If anyone could figure out the actual math, I would be interested.

I also think that other adjustments that Chris Cobb made (outlined by me above) further reduced this figure by at least another 10 percent (in particular, his conversions lowered the Negro League SA conversion by 10 percent more).

My personal opinion is that Chris Cobb's WS conversions are off by a significant margin, greater than 30 percent.

If you mutiply Beckwith's WS by JUST 25 percent, Beckwith looks like this (in a 154 G season):

39, 36, 35, 33, 31, 30.

Two notes on this:
1) the systematic regression flattens out his peak seasons.
2) Beckwith's real peak season, 1929, is not even included in the above.

In addition to all this, I have come to realize that a point made much earlier in this thread by Eric C (I think w/o checking), and wrongfully dismissed out of hand by me, is entirely valid:

Different skill sets cause different values in different Leagues.

Beckwith would have been more valuable in the Majors Leagues than the Negro Leagues. The Negro Leagues played a much more speed-oriented, less power dependent brand of baseball than the Major Leagues.

Beckwith was the greatest power hitter of his time in the Negro Leagues. He would have been more valuable, much more valuable in fact, in the Majors. Because he was the best power hitter of his time, he is more affected by this phenomenon than any other 1920s' Negro League player.

In my opinion, Beckwith would have and put up, in the Major Leagues of his time, WS peak seasons in the 40s.

Once again, Hornsby is a good comp (in the 1920s):
47, 42, 41, 40, 38, 38.

I still think that Beckwith was, in fact, a greater hitter than Hornsby.

One thing I would be interested in seeing, though Chris Cobb hasn't done it yet, would be WS conversions for Oscar Charleston. Simply by reputation, Charleston was the second greatest player in the 1920s behind Ruth.

But I'll bet Chris Cobb's conversions do not show that.

Note to Chris Cobb:

I don't mean to slam your conversions, which I think are great. Your methodology is sound, only your rates are (in my opinion, of course) incorrect.

Oscar Charleston and John Beckwith were the two greatest Negro League hitters of the 1920s. This, I think, is pretty indisputable. Once again, I point out that a conversion system that ranks them as just run of the mill star hitters has something wrong with it.

Perhaps also, as a check on the conversion system, it would be useful to reverse-convert a superstar like Hornsby to what his stats would have been in the Negro leagues.

I don't mean to slam your conversions, which I think are great. Your methodology is sound, only your rates are (in my opinion, of course) incorrect.

Just so you know, I am working today on reviewing the conversion factor, taking into account the data you provided on levels of league offense and applying a wartime discount to 1944 and 1945 numbers equivalent to the WARP view of those seasons' strength relative to 1946. This study will also treat the NNL and the NAL separately to see if there are significant differences there, leave out the transition year (though any career projection needs to target a rookie year and apply a heavier discount to that season) and look only at seasons between ages 23 and 32. We will see what happens.

In addition to all this, I have come to realize that a point made much earlier in this thread by Eric C (I think w/o checking), and wrongfully dismissed out of hand by me, is entirely valid:

Different skill sets cause different values in different Leagues.

I have been thinking about this point as well. KJOK and EricC raised the point to suggest that the conversions will necessarily be slightly too high, because the players whose skill-set is right for success will make the transition to the higher league.

I agree that they are right in principle, but I don't think we can assume that this would lead to a conversion rate that is too high.

Here's the thing. We are pretty clear on the fact that speed was more valuable in the NeL than in the majors (gadfly argues above that Beckwith's conversions would be higher because he was not a speedster but a slugger), so that speed would give players more advantage in the NeL than in the majors, so the sluggers would succeed better in the majors than in the NeL.

But who did the major leagues want from the NeL? In large part, they wanted speedsters: Doby, Irvin, Jethro, Robinson. The majority of the players first pulled were the players whose game was well-suited to the NeL in this respect. Campanella, an exception to this pattern, was brought across early, but Bob Thurman, for instance, had a tough time breaking in, even though he was a tremendous power hitter. So I don't think we have any reason to assume that the sample of players we have is biased towards those who would convert at the highest rate.

I think part of the reason these players were attractive to some scouts was because their speed-based style of play went counter to the ML style of station-to-station baseball, so they could see that this players would add something that teams didn't already have, and they were clearly dominating players in the NeL. The scouts recognized the value of a skill set that was different from ML-standard, perhaps, more than they recognized the value of the players whose skill-set was more tailored to the ML game as it was played. Doby, Irvin, and Robinson were, of course, five-tool players, so they were going to succeed no matter what, but I believe they all felt the "speed-penalty" in the conversion quite fully. Jethro, whose game was more purely speed, probably felt it the most.

Chris Cobb-

1) ON SKILL SETS

Exactly.

I have been working whenever I can on my Ramon Herrera-Eastern League study (and I'll post it on the MLE thread when I get sick of working on it). But, as is usual with me, the study has mutated into something else: the differences between an extreme deadball and extreme powerball offenses and the different influence of speed on BA and SA under each condition.

But the thing to remember is that speed is always valuable. Speed works on both offense and defense. In an extreme powerball offense, speed is still valuable. It's value is just reduced some by the emphasis on slugging.

But power is the great variable. The difference in the value of power between deadball and powerball is extreme and geometric. The Major Leagues are (perhaps were) in an extreme powerball offense right now and have been for the past decade.

Mark McGwire (especially) would have never been as valuable as he was during his career at any other time in baseball history.

I hate to say this because it borders on racist territory, but it also seems pretty obvious that the Negro Leaguers, as a group, were a lot faster than their Major League counterparts.

I don't know why this is and am not interested in getting into some nature versus nuture argument, but better minds than me have already pointed out this difference. Bill Veeck, in his autobiography, simply came right out and said that black players were faster. Bill James, in a famous study of rookie players and career paths, found the same ghost in his data without even looking for it.

For different reasons above in the thread, I printed out these 1928 statistics:

BA-SA-Isolated Slugging Average
1928 .281 .397 .116 AL
1928 .281 .397 .116 NL
1928 .279 .384 .105 ECL
1928 .278 .384 .106 NNL

All four Leagues hit just around .280, but the Negro Leagues isolated slugging was reduced by 10 percent.

[Which is why your equating 1928 NL and 1928 ML with no adjustments is inappropriate.]

There is an obvious reason for the slugging reduction: The Negro Leagues played baseball with inferior quality baseballs and also were much more lax on policing doctored baseballs, i.e. spitballs. But the question then is: Why aren't the Negro League BAs also lower?

I think the answer is simply speed.

The speed factor is pumping up the Negro Leagues BA and also, to a lesser extent, the SA. If you could go back and hobble the Negro Leaguers to bring their speed down to contempory Major League quickness, those 1928 Negro League BA-SA wouldn't be .280-.385 or so. I think they would be more like .270-.370 or so.

In other words, Negro League players conversions are not going to be, as argued, slightly too high; they are going to be too low becuase the skill set the Negro Leaguers brought to the Majors is MORE VALUABLE than the contemporary Major League skill set and the conversions are eliminating the Negro Leaguers' speed advantage.

And the proof of this is in the pudding. The Major Leagues, after integration, became much more like the Negro Leagues. You could even say that integration forced the Major Leagues to adopt the Negro League style of play.

But the most important thing to remember is this: the Negro Leaguers as a group, while having a speed advantage, did not have a corresponding power disadvantage. They had just as much power, maybe even more, it was simply being masked by the inferior baseballs and continued use of spitballs.

When the Negro League players migrated to the Majors, their speed stayed intact and their power blossomed.

Of course, SOME Negro League players would have been hurt by the transition to a more powerball offense. You mention Sam Jethroe above and someone mentioned Cool Papa Bell before; but neither of these men would have been hurt by the switch at all.

Jethroe had better than average power (he is quite comparable to Lou Brock, only better). In his two full, injury-free Major League seasons, Jethroe hit 18 HRs in both years. In 1953, at the age of 36, Jethroe hit 28 HRs in Triple-A.

Bell had average (and, if he hadn't taken up switch-hitting, above average) power. His Mexican League power stats from 1938 to 1941 are excellent and he even led the Mexican Leagues (though it was actually a fluke) in HR in 1940.

Bell, by the way, is quite comparable to Jethroe, only significantly better.

A player with about average power is going to be little affected by the switch from a deadball to powerball (or vice versa) offense. Who is going to be affected the most is the extreme ends of the spectrum: sluggers and defensive specialists.

One Negro Leaguer (already mentioned by me above) who would have had a much shorter, if any at all, Major League career would have been Dick Seay. He played over 20 years as a defensive specialist in the Negro Leagues and is, as already noted, quite comparable to the present day Rey Ordonez.

But, if Seay had played in today's Majors, his career would have been much more like Ordonez' career: Cut short by his offensive short-comings. Seay was a much more valuable player in the context of the Negro Leagues.

And, of course, the opposite end of this argument is John Beckwith and (frighteningly) Josh Gibson. Negro League players with extreme power would be much more valuable in the Major Leagues.

[By the way, as you note, the initial black players that integrated the Major Leagues all had good or great speed, but they also had good or even great power. The two exceptions you point out, Campy and Thurman, don't even fall outside this category. Campy was fast for a catcher and Thurman was also quite fast. His career was derailed by injuries - a severe wrist injury if I remember correctly - and age, not talent or speed.]

ON CONVERSION FACTORS:

I have been running conversion factors for Triple-A