There are no pages online specifically about Clarkson, so that explains the summary above.

I've got background info, then I'll post MLEs. First the real stats



NEGRO LEAGUES
YEAR LG TM AGE POS TMG G AB H TB 2B 3B HR SB BB K AVG SLG
---------------------------------------------------------------------
1937 NNL CRW 19 SS 28
1938 NNL CRW 20 SS 42
1939 NAL TOL 21 SS 19 5 19 8 8 0 0 0 .421 .421
1940 NAL IND 22 SS 8 16 3 6 0 0 1 .188 .375
1940 NNL NWK 22 SS 46 89 39 68 3 1 8 .438 .764
1942 NNL NWK 24
1942 NNL PHI 24 SS 31 114 41 58 11 0 2 2 .360 .509
1946 NNL PHI 28 3B/SS 46 38 146 45 62 7 2 2 8 .308 .425
1949 NNL PHI 31 3B/SS 56 192 60 .313 .313
1950 NNL PHI 32 SS/3b 44 33 108 32 56 10 1 4 8 .296 .519

MILITARY SERVICE
1943
1944
1945

MEXICAN LEAGUE
YEAR LG TM AGE POS TMG G AB H TB 2B 3B HR SB BB K AVG SLG
----------------------------------------------------------------------
1940 MEX NUE 22 3b/SS 91 19 80 27 40 4 3 1 1 5 19 .338 .500
1941 MEX TAM 23 3b/SS 103 82 326 109 195 23 3 19 7 43 53 .334 .598
1946 MEX VER 28 3b/SS 84 37 131 39 76 8 1 9 7 24 25 .298 .580
1947 MEX VER 29 3b/SS 119 112 390 118 202 19 7 17 20 79 65 .303 .518

CANADIAN PROVINCIAL LEAGUE
YEAR LG TM AGE POS TMG G AB H TB 2B 3B HR SB BB K AVG SLG
----------------------------------------------------------------------
1948 CPL STJ 30 3b/LF 100 84 318 110 219 14 1 31 17 .346 .689
1b

AMERICAN ASSOCIATION
YEAR LG TM AGE POS TMG G AB H TB 2B 3B HR SB BB K AVG SLG
----------------------------------------------------------------------
1950 AA MIL 32 3b 154 59 205 62 96 11 1 7 0 26 30 .302 .468
1951 AA MIL 33 SS/3b 154 97 283 97 132 12 4 5 6 43 29 .343 .466
1952 AA MIL 34 SS/3b 154 74 242 77 131 14 2 12 10 56 27 .318 .541

NATIONAL LEAGUE
YEAR LG TM AGE POS TMG G AB H TB 2B 3B HR SB BB K AVG SLG
----------------------------------------------------------------------
1952 NL BOS 34 ss/3b 154 14 25 5 5 0 0 0 0 3 3 .200 .200

TEXAS LEAGUE
YEAR LG TM AGE POS TMG G AB H TB 2B 3B HR SB BB K AVG SLG
----------------------------------------------------------------------
1953 TL DAL 35 3b 154 137 445 147 235 32 1 18 11 100 51 .330 .528
1954 TL DL/BU 36 SS/3b 162 157 543 176 327 21 2 42 7 104 48 .324 .602
1956 TXL TUL 38 3B 154 48 137 35 54 4 0 5 1 38 14 .255 .394

PACIFIC COAST LEAGUE
YEAR LG TM AGE POS TMG G AB H TB 2B 3B HR SB BB K AVG SLG
----------------------------------------------------------------------
1955 PCL LA 37 3B 172 100 316 93 140 8 0 13 1 34 28 .294 .443
1956 PCL LA 38 3B 168 8 11 3 4 1 0 0 0 4 4 .273 .364

WESTERN LEAGUE
YEAR LG TM AGE POS TMG G AB H TB 2B 3B HR SB BB K AVG SLG
----------------------------------------------------------------------
1956 WES DES 38 3B 140 60 205 57 107 11 0 13 3 .278 .522

CUBAN WINTER LEAGUE
YEAR LG TM AGE POS TMG G AB H TB 2B 3B HR SB BB K AVG SLG
----------------------------------------------------------------------
1947 PFL ALC 29 IF 95 106 21 29 5 0 1 1 .198 .274


CARRIBEAN SERIES
YEAR LG TM AGE POS TMG G AB H TB 2B 3B HR SB BB K AVG SLG
----------------------------------------------------------------------
1951 PR 33 3B 20 4 7 0 0 1 0 .200 .350
1952 PR 34 3B 21 4 4 0 0 0 0 .190 .190
1953 PR 35 SS 15 7 9 0 1 0 2 .467 .600

PUERTO RICAN WINTER LEAGUE
YEAR LG TM AGE POS TMG G AB H TB 2B 3B HR SB BB K AVG SLG
----------------------------------------------------------------------
1940 PRWL MAY 22 UT 42 18
1941 PRWL MAY 23 UT 44
1946 PRWL C/G 28 UT 55 172 60 13 .349 .576
1948 PRWL PON 30 UT 80 149 46 79 7 1 8 8 .309 .530
1949 PRWL PON 31 UT 80 269 77 144 16 6 13 .286 .535
1950 PRWL SAT 32 UT 78
1951 PRWL SAT 32 UT 72 4
1952 PRWL SAT 34 3B 72 18
1953 PRWL SAT 35 UT 80 175 49 58 3 .280 .331
1954 PRWL SAT 36 UT 72 240 76 121 15 .317 .504
</PRE>
Have bat, will travel.

For Clarkson's MLEs I did four things that should be noted:
1) I decided that he was a SS/3B (75%/25%) until age 30, then exclusively a 3B thereafter. This seemed to fit his usage since he seems to have never really settled at one position, suggesting that he was athletic enough to handle anyplace, but perhaps not good enough to stick.
2) I used surrounding seasons to determine his wartime credit.
3) I shaved off the first two years of his NNL career (ages 19-20), mostly because there's no data available for them, suggesting he was a greenhorn who was on the roster but getting seasoning and so not playing.
4) 1948 in the Canadian Provincial League was converted at a rate slightly higher than I would normally use. Typically I would convert this league as a Hi-A league at .85/.72. However, on Christien's counsel, I went with a AA rating due to the presence of numerous NgL stars.

Bus Clarkson MLEs
YEAR LG AGE POS AVG OBP SLG G PA AB H TB BB ops+ sfws
-----------------------------------------------------------------------
1939 NL 21 3B/SS .400 .455 .400 5 22 20 8 8 2 131 1.0
1940 NL 22 3B/SS .341 .416 .529 133 571 506 173 268 65 159 32.0
1941 NL 23 3B/SS .278 .341 .464 123 511 466 129 216 45 125 19.1
1942 NL 24 3B/SS .324 .421 .417 114 504 432 140 180 72 145 20.7
1943 NL 25 3B/SS .297 .380 .447 127 549 484 144 216 65 138 21.9
1944 NL 26 3B/SS .290 .376 .442 121 523 459 133 203 63 130 20.1
1945 NL 27 3B/SS .291 .383 .435 119 519 452 131 196 68 127 19.9
1946 NL 28 3B/SS .293 .379 .455 146 631 554 162 252 77 136 25.4
1947 NL 29 3B/SS .249 .345 .426 97 423 369 92 157 54 104 13.4
1948 NL 30 3B/SS .284 .377 .483 106 461 401 114 194 60 131 19.7
1949 NL 31 3B .262 .350 .402 139 601 529 139 213 72 101 18.0
1950 NL 32 3B .266 .350 .399 154 659 584 155 233 75 96 19.6
1951 NL 33 3B .302 .375 .380 97 412 369 111 140 43 104 13.1
1952 NL 34 3B .270 .368 .455 105 447 386 104 176 60 127 16.8
1953 NL 35 3B .274 .367 .363 117 508 443 121 161 65 91 14.1
1954 NL 36 3B .287 .373 .469 148 639 562 162 263 77 122 25.9
1955 NL 37 3B .261 .332 .396 90 376 340 89 135 36 94 10.5
1956 NL 38 3B .231 .337 .347 40 178 154 36 53 25 85 3.9
=======================================================================
TOTAL .285 .371 .435 1980 8533 7510 2143 3265 1023 120 314.9

</pre>
The games played would put him around #20 all time among shorstops, between Guillen and Speier.

Among 3b it would put him roughly #15 all time between Aurelio Rodriguez and Pie "Potty" Traynor. (I'm saying roughly because I'm using the 2001 version of the SBE, and so players climbing up those career leaderboards might not be on these lists).

I'm glad that I was asked to post this thread for Clarkson. I didn't realize he was that good.

1. I would discount his 1940/age 22 season, it is too much apart from the rest of his curve.

2. Why did he play partials in '50-'51-'52? I see you have him playing 200 games over the latter 2 years, but 154 games in '50. Was he injured or something? Should it reflect in his MLE?

3. And a 26 WS year in 1954 seems generous in context of his career MLEs. He had a decent year in PRWL that year, but .324 in the TxL is not a really big deal.

Still, he looks bona fide, though I'd be inclined to knock off about 25 WS = about 290.

Before I answer SDY2's question, let me give a big shout-out to Gary A. and KJOK for a lot help in assembling this data.

1. Sounds good, could be a park thing here.

2. The Partials:
-1952 has to do with his getting a call-up to the Braves, where it appears he did a lot of bench warming and didn't get much of a shot. I chose not to manipulate his playing time that year until discussion took place, but I think a reasonable case can be made that he should have his season prorated toward his usual levels of durability. Is that a reasonable approach, or better to leave it alone?

-1951: I don't have an answer about this one. He was with the Braves affiliate in Milwaukee all year as far as I can tell. Could be an injury, could be racism at the local team level. I dunno.

-1950: I've credited him for playing the entire season due to a combination of his playing 3/4s of PHI's games (NNL) as well as 1/3 of Milwaukee's games in the AA. Caveats apply, as noted elsewhere by Chris Cobb concerning actual decisions versus reported results.

3. My understanding of the TXL of this time was that it's what we would call a AA-level league today, so I converted it at .875/.76. in 1954, Clarkson absolutely mashed, going .325/.602 with 42 dingers and 104 walks. Combine that with the same conversion rate for an estimated .317/.569 year in the PRWL, and that's how it comes out that way.

The interesting thing, however, is that 1954 appears to be a genuine last-hurrah kind of season for Buster. His surrounding years in the PCL, TxL, PRWL, and the Western League aren't anywhere close to EITHER his TXL or PRWL totals for the year. For whatever reason he just had one more good year in him before sputtering out in his late 30s. To this end, here's a note about his TxL parks that Gary A. has shared with me.

Btw, Clarkson and Willard Brown played on the same team in the 1953 Texas League (and for part of 1954). It's interesting to look at their stats side by side--at the time, Clarkson was clearly a much superior offensive player (and, given that he was an infielder, even better as an all-around player). If they were both born in 1915, then this would be an awfully good indication that Clarkson was a much better player over his career.

As for park effects: Dallas and Beaumont in the Texas League and Milwaukee in the American Association seem fairly neutral, as far as I can tell. Milwaukee *possibly* cut down on home runs a bit, but not very noticeably; and overall offense seems normal, considering that the team won the pennant in 1951-52.

How often does a hitter have his best season at age 22?

Among all hitters born before 1980 who earned at least one win share 345 of 6610 had their best season at age 22, or 5.2%

Of those with at least 10 win shares for their career 119 of 3810 had their best season at age 22, or 3.1%

The percentage is basically the same, 3.0%, for those with at least 100 career win shares.


The best players that had their max win shares at age 22 (includes ties for Max Season).

As long as I have teh spreadsheet out, how many hitters have had a 30+ win share season at age 22?

22 of them.


That age 22 season for Ruth is really pitching season.

Please do not take this as me endorsing The Bus, I just was curious, so I looked it up.

What I think this really shows is the 32 win share season is likely a sample size or park effects or bookkeeping Sasquatch.

It's also worth noting that Clarkson lost a good chunk of his peak/prime (ages 25-27) to the war. Doc's MLEs have him pretty flat, win-share-wise, during those years, leveling out at 20 WS a season. His first year back, 1946, he gets 25; it could be he lost another couple of 25 win share seasons, or better.

There's no way to know, of course, but this bears keeping in mind, particularly if you're considering discounting the value of his 1940 and 1954 seasons.

Mike,

Thanks for the feedback. I'm going to take a look back over my computations, I have a feeling any sasquatching is mine. He was with three teams that season in two different leagues, and I think it's possible my own calculus is doing something funky as I try to balance out his playing time.

More later....

OK,

I don't think the sasquatching is mine. I think it's the record keepers. I've tried a bunch of different ways of figuring it, and it always comes out within a few points of what you see above. Of course, hitting .438/.744 will do that, I guess!

Thinking empirically about his 1940 season, if you knocked his BWS down by 10% because you thought the numbers were park-induced, then he would be at 29.2 for the year, still his best single season. Of course, he only played a single year in that NgL park, so it's impossible to say whether or not his numbers would have been consistently inflated by it.

I wish I had a better answer, but I think it's also entirely possible that this is just a random fluctuation over a small sample (only 89 ABs). If so, regressoin would smooth it out.

I think the ages are incorrect on the MLEs. Clarkson was born in 1915 (although he was called a '29 year old SS' when called up to the Braves in 1952') which would make him 24 in 1939, not 21...

Also, Clarkson was definitely a SS at Milwaukee in 1951, although looks like he moved to 3B in '52...

BB-REF gives the 1915 birthdate for Clarkson; Riley lists 1918; and The Negro Leagues Book has 1913 (all have March 13 as the day). I don't know which is now accepted--I'd think 1915 is most likely. (This would make his war years ages 28-30.) Maybe Gadfly knows?

In 1951, Clarkson played 97 games at Milwaukee; the source I have lists him with 73 games at shortstop, but doesn't give game totals for his other two positions (3b and of).

If either 1915 or 1913 is correct, that would radically change how I am viewing him. First off it would make the big 1940 season all the more likely, second, it would make the end of his career much more impressive since he'd be a 40 or 43 year old racking up 15+ WS a year as an IF.

bump

Clarkson spent 8 weeks with the Braves in 1952 (April 30-June 22), if that helps any.
6 SS, 2 3B

Hey everyone, quick update on Clarkson. When applying the more appropriate 93% conversion rate to Clarkson's AAA-level numbers, his career line changes from

AVG OBP SLG OPS+ SFWS
.284 .369 .433 119 311.2</pre>
to
.286 .371 .437 120 316.1</pre>

Not vastly different, but enough so that I thought it might be worth mentioning.

Doc,

Is he better than Vern Stephens?

I've been meaning to make this comparison:

Clarkson* 315 32-26-25/115
Sewell 277 29-29-26/125

(* per post #3)

And then:

Stephens 265 34-32-27/129

Clarkson .285-.371-.435/120 in 8500 PAs
Sewell .312-.391-.413/109 in 8000 AB+BB
Stephens .286-.355-.460/118 in 7200 AB+BB

Clarkson had a longer career than Stephens but then everything hinges on whether he would have had a longer career in the MLs or not. At a glance they seem similar otherwise.

But note that while Stephens is the one whose defense is often questioned, he is the only one of the three who played his entire prime at SS, and even moved Johnny Pesky off the spot at one time. My ######## dump says that Clarkson's ML career wouldn't have been 8500 PAs and that Stephens' defense was better than reputed. So I would rank the three Stephens, Clarkson, Sewell.

I have Clarkson ranked identically to Sunnyday2.

and even moved Johnny Pesky off the spot at one time.

I suspect someone's brought this up before, but is there evidence that this was because the Red Sox thought Stephens was a better SS defensively? As the A-Rod/Jeter situation in New York now proves, there can be other reasons...

I confess great ignorance of Clarkson. Do we have ANY info that knowledgable people who watched him play considered him a great player; I mean, other than one-line throwaway quotes that you could find today on the 'worthiness' of Neifi Perez? Can I justify ranking him near Dobie Moore or Bill Monroe?

Sunny,

While I haven't ranked Clarkson yet, I would rank Sewell above Stephens because some of Stephens best years came in the watered down wartime AL. This shaves a few WS off of his top five seasons, enough so that Sewell probably moves ahead of him.

Clearkson does seem to fit this mold, however. Right now I have Lundy, Rizzuto, Sewell, Stephens. None are in my top 30.

j,

It's true I didn't give Stephens the WWII discount in the numbers above, and in making my ballot I do give a 10 percent discount. And yes, "some" of his best seasons came during the war. But like Newhouser, his years outside WWII give pretty powerful evidence of what kind of player he really was. He was not Snuffy Stirnweiss.

What is a killer for Sewell IMO is that only 8 of his 13 seasons were at SS. That is more of an issue to me than Stephens play in '43-'44-'45.

But if you seriously have Lundy and Rizzuto ahead of Sewell and Stephens, I can't believe Dave Beauty Bancroft doesn't factor in there somewhere.

TomH,

I think one of the reasons there's not much awareness of Clarkson is the fragmentary nature of his career. He moved around a lot, from one team to another, from one league to another, from one country to another, from one level of the minors to another, from the ballplaying to the war and back. He didn't get a real shot in MLB, and by then he was already 32 and past his salad days. I just don't think he was ever in one place long enough for anyone to fix his image in their mind.

I believe this is one of the subtler effects of the color barrier and of the slow pace of integration.

Another player we'll experience this with is Marvin Williams, a second baseman whose numbers will look a lot like Clarkson's. Alonzo Perry will for this (and other reasons) be yet another.

my rankings right now are

35. Lundy
36. Rizzuto
41. Sewell
47. Stephens
58. Bancroft


Beauty is definitely better than Tinker, Long, and Maranville. I just don't like his offense or his peak/prime enough to place him with the others. He is jsut behidn Sewell in WS, but in WARP Sewell kills him.and I guess 35 to 58 isn't as far as it used to be. I should have included him.

What is a killer for Sewell IMO is that only 8 of his 13 seasons were at SS. That is more of an issue to me than Stephens play in '43-'44-'45

Not sure this is a "good" reason. 3rd base in Sewell's time was not much different than SS. It was a primarily defensive oriented position, and the reason Sewell was so outstanding is that he could ALSO HIT.

True but.

I don't think there's any question that Sewell's demise in these elections came about the time that most voters realized that he hadn't played SS for 13 years.

Still "killer" was too strong. Joe is in the mix. He is #12 on my list (not my ballot, but my list) of post-1920 lively ball candidates. Along with several SSs.

bump

An update. Based on the league averages reported in the Hall's .pdf for the last 37 NgL candidates earlier this year, here's I've updated Bus's translations. Like for Trouppe, the news is that it slightly improves his SLG and OPS+.

Old translation (see post #18): .286/.371/.437/120/316.1

New translation: .285/.370/.439/121/318.3

It's not too big a change, but I figured I'd better let everyone know anyway.

any word at all about Clarkson as a gloveman?

Once again, are the MLEs in #3 not credible? Or is a 315 WS SS not ballot-worthy?

Once again, are the MLEs in #3 not credible? Or is a 315 WS SS not ballot-worthy?

He's a better player without imagination, than Aparicio was with reality, now if he'd only had a stable league envrionrment....

My main knock on Clarkson is that his peak is a bit soft.

.285/.370/.440 looks a lot like Barry Larkin with the bat. Right now I've got Clarkson about even with Vern Stephens although I have more career offensive value for Clarkson and less defensive value (I have Clarkson at exactly average). That gets him into the top 100 but not the top 50.

I agree that Larkin looks like a pretty good offensive comp for Clarkson, both in level of production and (if our guesses on Clarkson's playing time are accurate) in-season durability.

Because Larkin was an excellent defensive player, his offense/defense combination will make him a clear HoMer. Clarkson's case is hard because we have so little information about his defensive performance. If he was average (and Dr. Chaleeko has built the assumption of average fielding value into his estimates, I think?), then Clarkson is a borderline candidate. If his fielding value is overestimated by a projection of average, then he's on the outside looking in. If his fielding value has been underestimated, then he ought to be elected pretty much immediately, as my system sees it.

This is a point at which non-statistical documentary evidence would be really valuable!

His cup of coffee in Boston in 1952 suggests that he was no longer capable of playing shortstop at major-league quality at age 37 (the sample size is only 6 games, but his performance is _really_ poor, both in terms of range and error-rate) but that he was fine at third base.

It looks like he was competing with Sibby Sisti for the utility infielder spot on the Braves' roster. Sisti couldn't hit at all, and his defense was shaky at 2B and SS, but Clarkson didn't get enough playing time to come around with the bat, and he didn't show sufficient leather at shortstop, so that, it appears, was it for him in the majors.

Without some more definite evidence on Clarkson's fielding, I think he's going to remain a borderline candidate, but as we are about to _elect_ a bunch of borderline candidates, we ought to look at him head-to-head with Sewell, Childs, Boyer, Fox, Elliott, Stephens, Rizzuto, et al. He's surely good enough to place somewhere in this group.

>It looks like he was competing with Sibby Sisti for the utility infielder spot on the Braves' roster. Sisti couldn't hit at all, and his defense was shaky at 2B and SS, but Clarkson didn't get enough playing time to come around with the bat, and he didn't show sufficient leather at shortstop, so that, it appears, was it for him in the majors.

And this process was repeated through the MLs during this period--the black player getting a look, but not much of one--and possibly some legitimately past their prime anyway. But of course he got a look in 1952, not 1947. So, ergo you get your lost generation.

That is not to say he is a HoMer. I'm still sittin' and thinkin' on that one (and Williams and Artie Wilson and Scales). Moving to 3B fairly early in life suggests he might not have been the average fielder that he is in Doc's MLEs, and his offense while excellent for a SS is merely above average for a 3B. I'm slightly more enamored of Scales and Williams right now, though Clarkson is arguably top 50 as well.

Pursuant to Chris's comments viz Larkin/Clarkson, Larkin's ability to stay at SS for so long and play good defense there the majority of that time is a big point of differentiation with Clarkson. It's not at all clear to me how long Clarkson would have been a SS. In no full season after returning from the war did he play in any league where was he a SS-only type player. He always has the slash after 1946. I strongly suspect that means he would have moved to 3B or OF at some around age 30. Combined with his lack of clippings about defense, I have to imagine that's a strong indication of his defensive skills. It's not perfect, of course, because there's a lot of leagues where reports on his defense would have been hard to come by, and also he'd only been a regular a few years before the war, so any public defense reputation may have gone by the boards in that time.

I gotta think he's a SS til about age 30, then a 3B, and probably around average in each instance.

Strangely, as a modern comp, I usually think not of Barry Larkin, but rather of a less discontinuous version of Julio Franco. And while we don't necessarily see Franco as a HOM kind of guy right now, using MLE technology, Franco's career totals for his exiles bring him friggin close to the kinds of career totals that we associate with very strong candidates. Throw in better defense for Clarkson, I think, and you're talking about a very compelling guy.

One overlooked thing that is really in Clarkson's favor is the leagues he was actually in.

Unlike Williams who played in several lower-classification leagues, Clarkson played pretty much his entire post-NgL prime in really good leagues: MxL (during the Pasquel raids of MLB and before integration), AA for three years, PCL for a couple, and then declining in the TxL. He played a lot of games in very high minor leagues during his post-peak period and did uniformly well in them. That speaks well to him in a similar way to how Easter's minor league record reflects well on him. What's different is that Clarkson played tougher positions and doesn't have the slow-minor-league-slugger-on-chronically-bad-legs thing that Easter does.

> as a modern comp

I've been looking around at playing time for various infielders at different positions and I think Jay Bell is probably a good comp for Clarkson defensively.

As I've been mentioning lately, I've been doing a lot of work with the MLEs for NgL players. A few weeks ago, I posted Artie Wilson's new MLEs. My WS estimate came out about 15-20% higher than my previous estimates had. Now Clarkson has come out 20-25% higher than the previous estimation. I've been through it a bunch of times to see if there's something wrong. I don't think there is, though I reserve the right to be wrong. I'm certain a few of you will raise an eyebrow over the result that has come through for Clarkson, and so I do invite feedback, and I'll happily explain, as ever, since it's always possible I've just scrooed up.

As for the how and whys....

1) Same method as elaborated in Charleston and Johnson (with many of the modifcations suggested by the group).

2) For the war years I took a simple average of the MLE PAs Clarkson had in the four surrounding years and used the same basline that I regress partial seasons from. This is 1943-1945.

3) Yes, I do notice that the late TxL seasons stick out somewhat. I've translated them at .825 on runs. Those leagues were normal for R/G at 4.42 and 4.66 (the NL, which the projection is based on, was 4.75 and 4.56), and I've estimated his pfs and accounted for them (.97 and 1.00). I just think he had two big, healthy years left in him---it happens. The OPS+ both years is well within the range based on his career norms. Which is bringing me to a point. How did I discount his leagues?
-NgL before 1946 and the Open PCL: .90
-NgL 1947-1948 and the AA: .88
-NgL 1949-1950, MxL, and TxL: .825
-Western League: .75
-Canadian Provincial League: .70

I think this discount schedule is working well since it comes up with values for his seasons that are within normal for him. No crazy high years that would indicate an under-discounting nor the opposite.

4) I had some MxL information on HPB and SH, so I used them to create estimates which are included below. I had to estiamte his walks a few years, but there was ample data to work from. In the couple places I was missing vitals (xBH in 1950 for instance), I used career norms. All CS are as if the player were a 70% stealer.

5) I've got good information on the r/g of most of his leagues. Not the Western League. However, the league leader in HR that year was none other than Dick Stuart and his famous 66. So I gave that league a very high R/G at 5.50 during my translation routine. Usually in doubt i just go with 4.5, but I did a little wiki bullpen search and the article revealed that this league was at altitude, so I just gave it a really high r/g. Since I also didn't have a pf, I figured this made things conservative.

6) For defense, I made Clarkson a SS through age 30, then a 3B thereafter. I made him average at both positions.

7) I ran a total SFWS analysis too, using the same defensive scheme as mentioned just above. It comes in 7 total WS lower. So I don't think that the increase in WS over my previous MLE is due to my way of calculating WS.

Here goes....

402 WS?!!

Yes, I have raised an eyebrow, Eric :-), but because of the huge difference between the before and after MLEs, not because the new one is necessarily unreasonable.

It would be nice if Chris or Brent went over your new numbers, though, just to make sure everything is okay before I add him to my ballot.

BTW, if these numbers are deemed creditable, he would be #1 on my next ballot.

I'll post a fuller response to these MLEs when I am at home and can compare them to what I have, but here are a few quick thoughts:

1) I am skeptical about the playing time, if not necessarily the quality, of Clarkson's last four years. I think it is accepted now that Clarkson was born in 1915 (that's the date BP uses), so he is three years older than these projections indicate, though if he was really playing every day in the PCL in 1953 and 1954, then maybe these numbers are right. In any case, I'd like to know a bit more about how playing time was estimated, and whether a different age would affect the estimates.

If I am remembering rightly that he was in the PCL in those years, it would also be helful, if possible, to see Clarkson's ranking relative to league leaders in those seasons. He really ought to have been among the leagues big stars.

2) The WW2 win shares estimates look a little high, in that they are higher than the average of the 4 surrounding seasons (which is 24 -- so they're not _way_ out of line). I'd like to know a little bit more about how they are being estimated.

3) I have long contended that Clarkson's numbers are very strong, so I'm not shocked that Dr. Chaleeko's new MLEs for him look very good. Even if these are 10% too high, he would still be a clear HoMer.

I'll go point by point here.

Age: I'll definitely look back over for age then. I should have corrected it, my bad. That puts his last year at age 41, not 38.

In the TxL in 1953 and 1954, he played 137 of 154 and 157 of 162, respectively. I don't have the Wright book here, but I can look up what his performance was like relative to league. I'm certain it was of league's-biggest-star quality:

lg R/G BC RC/G* MLB equiv comps
-----------------------------------------
1953 4.42 10.44 11.21 Musial (11.01) Snider (10.50) Mathews (10.17)
1954 4.66 9.90 9.48 Klu (9.87) Mathews (9.68) Musial (9.36)
*park adjusted

So in the TxL, his level of dominance was comparable to that of the biggest stars of the NL in both of these seasons. He translates to a the tier below those big names, with rc/g in the 8.25 area.

In any case, I'd like to know a bit more about how playing time was estimated, and whether a different age would affect the estimates.

I haven't accounted for age at this time. I estimate the number of PAs Clarkson would have in a season based on this formula:

Games / Team Games * 154 * 4.1 PA/G

The 4.1 PA/G was figured based on the fact that MLB had about 4.3 BFP/inning over his career span. This means I have him hitting seventh in general. But actually I wanted to come in a little conservative because in general he wasn't consistently durable. I could easily scale up or down, whatever anyone thinks is wise. I'd love any ideas anyone has for working with his PAs. That's admittedly the part of the process with which I have the least facility. That said, Clarkson's possibile brittleness results in PA totals that are Barry Larkin esque in some ways. He's got three years over 590 PA and a decent amount below 500.

2) The WW2 win shares estimates look a little high, in that they are higher than the average of the 4 surrounding seasons (which is 24 -- so they're not _way_ out of line). I'd like to know a little bit more about how they are being estimated.

It's initially dependent on the PAs. I project them from the neutralized 4.50 r/g numbers. I used the surrounding four seasons. In the 4.5 environment, the PAs for 1941-1942 and 1946-1947 are 503, 515, 525, 594 = average of 534 (i used 535). Then I used the neutralized career norms to figure the actual performance. Each of those three seasons looks like this in the un-projected 4.5 r/g ether:

535 PA 89 RC 463 AB 131 H 216 TB 69 BB .283 .374 .466

Once I project into the actual MLB level, they all look different. The NL of 1943-1945 has R/G of 3.94, 4.25, 4.46 (average 4.22), which is how he ends up instead creating 78, 84, 88 runs instead of 89 89 89. So something similar happens with the PAs. Once you project the surrounding PAs into the actual MLB r/g environments you get 495, 494, 508, 595 = average of 523. The NL r/g for those years is 4.23, 3.90, 3.96, and 4.57, which averages to 4.17, a bit lower than the 1943-1945 seasons, and possibly just enough to send the BWS up. The illusions created by the r/g changes might be causing the distortion.


Chris, or anyone, I'm totally open on the age 40 and 41 years. He does decline in the MLE, but if you think further shaping is necessary, I'm fine with doing so.

Just an FYI to everyone, I've virtually completed re-evals on:
Marv Williams
Bobby Estalella
Ray Dandridge
Bobby Avila

I'll likely post these for group inspection over the next week. Avila, in particular will require some attention due to the strange circumstances surrounding his 1949 season. Estalella's case is just plain weird, the two-three sources I have just don't line up.

And I do take requests! In the hopper I've got:
Jackie Robinson (for Dan R, and, yes, Dan, this one will be more reasonable with the new method)
Bill Wright (for KJOK I think)
Jim Gilliam (for me, hey i count too!)

I'm very happy to look into any hitter that's still out there who Chris doesn't already have queued up. Please let me know.

Also, about the war credit. That's just my solution via the MLE process. Everyone has their own vision of how to do it, and they should feel free to accept or reject my technique in favor of their own as they see fit. I'm more interested in getting the rest of the seasons "correct" because they are the ones whose interpretations will ultimately carry the most weight.

Also, also, this reprsents an increase of 27%, not 20-25% as I wrote previously.

The problem I have is this (and of course I could be completely missing something):

These MLEs answer the question, if he had been in the MLs, what would his value have been?

Meanwhile, for white ML players (contemporaries of Clarkson) and for ML players generally, the actual record answers the question: What was his value when he was in the MLs? Occasionally we add an MLE year for this or that ML player if we believe that player to have been "held back unreasonably." But generally, no, what you see is what you get. NeL MLEs, then, are more like taking the entire record--ML and MiL--of a ML player as if he had been in the MLs the entire time. And what would happen here? You'd have lots of players jumping from 300 to 400 WSs, as Clarkson has done.

Or taking it the other way around, what if Clarkson had been white or if he had come along in the 1960s. Well, then, his 21-23 year seasons would probably be in the MiLs, not the MLs. He would have needed those 2 years to prove that he was ready for the big time. And given his age, it would be hard to say it was unreasonable. And then, at the tail end of his career, who is to say he would have had the chance to post those 60 WS in 1953-54. More likely he would already have been judged to be over the hill and out of the MLs.

Just because a player woulda/coulda got 20-25-even 30 WS in the MLs doesn't mean he would or should be there. There are probably more 20 WS players in any given year of NeL play than there are WS available to be won.

So anyway, the MLEs are terrific, great info, I love them. But to compare Clarkson with, say, Vern Stephens on the basis of these numbers is apples and oranges.

Or taking it the other way around, what if Clarkson had been white or if he had come along in the 1960s. Well, then, his 21-23 year seasons would probably be in the MiLs, not the MLs. He would have needed those 2 years to prove that he was ready for the big time.

Please note that Dr. C's ages are incorrect. Clarkson was _24_ in 1939. I think we can safely assume that a player of his calibre would have had a chance to establish himself as a big-league regular by this age. We know of a few cases where players had to wait longer for their chance (e.g. Al Rosen, Hank Sauer), but those are the exceptional cases, not the norms.

This means, of course, that he might lose more of the tail end of his career. But how many cases are there of players who were still good players who were pushed out because of age? Veteran players almost always stick around until they can't compete effectively any more. A few walk away, and a few are pushed out, but again, those are the exceptions.

My bad on the age stuff, I should have rechecked the thread more carefully. I'm very sorry for any confusion.

NeL MLEs, then, are more like taking the entire record--ML and MiL--of a ML player as if he had been in the MLs the entire time. And what would happen here? You'd have lots of players jumping from 300 to 400 WSs, as Clarkson has done.

I've actually used the exact same source data as in the previous round of MLEs, there's nothing been added to Clarkson's record to expand it on either end. The jump results from differences in my current translation method (see Charleston/Johnson threads) and my previous one. The new one yields "higher" numbers because it's based in a different comparison than the previous MLEs were. I believe it's an improvement over my previous efforts, but it could not be too. I'm very much open to feedback.

So anyway, the MLEs are terrific, great info, I love them. But to compare Clarkson with, say, Vern Stephens on the basis of these numbers is apples and oranges.

This has always been true, and for reasons like this, anyone who says MLEs are 100% correct is deluding themselves. But that said, every MLE is conceived in this same way: what does this career look like in an MLB setting? I'd hope the MLEs are the best we can do in the moment, which is why I'm saying that if the group thinks that more shaping needs to be done, I'm happy to do so in whatever direction it identifies.

In Clarkson's case, he is listed by Lester and Clark on the Philadelphia NNL roster for 1937-1938 but appears to have no record of league play. This may have been his woodshedding period or something like that, maybe seeing a lot of time in non-league contests after the regulars had played a few innings. I don't know. Someone like Gary A could probably tell us what that means.

Chris, since you asked, I looked up what Clarkson ranked in several categories in the TxL in his two big years:



The league played a 154 sked, and 1953 is in 137 games, which is interesting---missing 15 games he's still placing very highly. A down-ballot MVP candidate at best, but probably an All-Star level player.

1954 is gold. Nearly a triple crown, played 150+ games, clearly one of the big stars in the league, and an obvious MVP candidate.

Eric,

Thanks for that data on Clarkson's position among league leaders. It's a lot easier for me to accept a translation that a player in the Texas League would have been a major league all star when I can see that he was possibly the best hitter in the Texas League, and undoubtedly in the top 3.

Eric,

Comparing the MLEs you did a year and a half ago with these latest ones, the results are quite different and many of us would like to better understand the reasons for the difference. Since you're doing a lot of things differently, it's hard for us to sort out which new data or assumptions made a difference and which ones haven't. For example, I'd expect regression to flatten out a career, but not change (much) the career totals or rates. New data on league environment or park effects could affect some seasons and not others. Changes in assumptions about quality of competition could have a more pervasive impact. Can you shed some light on which of these factors contributed the most to Clarkson's revised MLE career level and patterns?

Comparing # 3 with # 41, I notice that Clarkson's MLE average actually drops slightly (from .285 to .280), OBP is unchanged, and SLG increases quite a bit (from .435 to .461). OPS+ goes up from 120 to 126. He's also assumed to have about 270 more plate appearances.

The increases in OPS+ and in plate appearances don't seem sufficiently large to explain the addition of 87 win shares. Did your assumptions about his fielding also change?

Your work is appreciated, but I'm still trying to wrap my head around this.

The increases in OPS+ and in plate appearances don't seem sufficiently large to explain the addition of 87 win shares. Did your assumptions about his fielding also change?

Well, actually, it's a systemic thing, a sea change.

Here's a Bill James tax-return schematic of the old system. It's components based, and hinges on batting average (BTW, projecting from, say, NgL into the NL):

1) AVG / lgAVG (if known)
2) Take 1 and multiply by NL avg
3) Take 2 and multiply by .90 (average factor I used based on Chris's research)
4) Repeat for slugging using square of avg factor (also per Chris’ research).
5) Estimate walk rates if there's any idea about them.
6) Prorate playing time as g/tm g * 3.5 AB/g.
7) Figure out average, slg, and walk rates in the playing time from 6.
8) Season to taste with OPS+ and/or SFWS.

Here's the current routine for a NgL to NL translation.
1) RC/g / lg R/G (known or estimated)
2) Multiply 2 by pf (known or estimated)
3) Divide league r/g by 4.50, then multiply by player's rc/g to get neutral RC/G
4) Multiply 4 by .90, the NgL runs conversion factor.
5) Use the Willie Davis Comment Method (aka just Willie Davis) to figure out how the player's original stats translate into the new environment.
6) Sum all seasons to get neutral career averages.
7) Figure playing time as g/tm g * 154 * ~4.2 PA/G (variable based on the era, league, player's likely slot in the order, etc....)
8) Subtract the PAs figured in step 5 from step 7: this is how many PAs you will apply his career average to in building his full season.
9) Figure how many runs he needs to create in the PAs from step 8 based on his career RC rates (and his nearest three, if it's possible).
10) Use Willie Davis to figure his batting line based on the RC in step 9 and the PAs in step 8.
11) Add the batting lines from steps 5 and 10 (the neutral and "remaining" batting lines) to get the total neutral season line.
12) Divide 4.50 to NL season's R/G.
13) Multiply 12 by the player's neutral rc/g to get his projected NL RC/G.
14) Use Willie Davis to figure his component line.
15) Figure his WS as though he were on a .500 team scoring at the league average and figure OPS+ the usual way.

It's a little more complicated, and I've truncated my explanations a bit. If you can make sense of my hyeroglyphs, you'll see that the crux of the difference comes down to assessing runs versus components, something only made possible for the NgLs within the last year.

What I discovered was that I was likely misapplying Chris's conversion factors. As a result, I undercut player's skills/performance. For example, take a player in AAA (.88 on runs), Joe Minors:

664 PA, 600 AB, 175 H, 64 BB, 279 TB
.292/.360/.465 100 runs created (basic version)

My current method uses the Willie Davis comment. To get to 88 RC, it says Joe Minors will have this line in the majors:

646 PA, 587 AB, 162 H, 59 BB, 258 TB
.276/.342/.439 88 runs created (basic version)

The MLE batting average is 94.5% of the original. The MLE slugging is 94.4% of the original.

If we use my old system with the NgL converters I was using, it would yield averages of .262/.377. When I reverse engineer a .262/.377 player in Willie Davis:

631 PA, 576 AB, 151 H, 55 BB, 217 TB
.262/.327/.377 71 runs created (basic version)

So a .71 discount on runs via the old method. Which Clay Davenport, for example, claims is something like A Ball. So the discount structure I had been using was severly underrating players' performance. Which is part of why there's the big difference.

I can only apologize and say that I didn't know it at the time, but I think the assumptions that go into this new system are more robust than the ones that informed the old one. The Willie Davis technique does a better job of maintaining the relationship of players’ component skills, and by using runs created I’m assessing the player’s entire contribution rather than patching it together from one component to the next. Better yet, it’s more sensitive to fluctuations in runs scored between leagues, which is a huge issue for these purposes.

I’m not saying this new system is bullet proof, nor that I haven’t made errors in either input, formulae, assumptions, whatever. And I do have to make some estimates that aren’t ideal, but this system is at least grounded in something that I feel confident is giving me better answers. I don’t have time this afternoon, but hopefully tonight I’ll run Clarkson’s 1954 season through each version of my MLEs to show you where the differences crop up.

1954 is gold. Nearly a triple crown, played 150+ games, clearly one of the big stars in the league, and an obvious MVP candidate.

I'm a bit skeptic on the 60 WS these two years. 1954 was a longer season (162G) no? The Texas League was AA and he regressed a bit when he went up to AAA in 1955. I know he was getting older and I know he shouldn't have been in the minors at all, but an "MVP candidate" deserves a second look.

Within the Texas League, Clarkson was certainly an MVP candidate, and I believe that's what Eric was describing when he used the phrase.

I'm not certain that the 31.3 win shares the new MLEs assign to Clarkson is a correct estimate of the MLE equivalent for that season, and I'm not certain that it's incorrect, either. The key point is that Clarkson was playing third base and was the best hitter in the Texas League, so he was very likely the most valuable player in that league.

That is a key point because it is obviously counterintuitive that the Texas League would have had a bunch of players who were capable of putting up all-star seasons in the majors at that time. It is not counterintuitive that the best player in the league would have been an all-star in the majors.

Put another way, this result passes the sniff test. Doesn't mean it's right, but it's not necessarily incorrect.

1951 33 13.9 2.5 16.4
1952 34 14.1 2.4 16.5
1953 35 25.3 3.7 28.9
1954 36 27.4 3.9 31.3
1955 37 12.8 2.4 15.2
1956 38 12.2 3.0 15.2

The 1953 and 1954 MLE seasons look out of place in that career summary.

I think it's fair to ask who the other best 3-4 hitters were in those leagues, and if possible, what were their MLE projections for those seasons. (Assuming that data is available. Or is the source only data for Bus Clarkson?)

Brent, you are correct that TxL in 1954 was a 162 season, not 154 as I had said. I had that right in my original data sourcing, so that was an error strictly in post #50, not in the MLE itself.

His 1953-1954 seasons do bring us back to the question of What about Barry Bonds in AA? That is do great players look that much better in lower leagues, and if so, what to do about it? No answer is forthcoming from these quarters, because I just don't know. I agree the seasons look out of place, but look at the PAs and OPS+

396 118
382 124
570 140
613 139
367 114
470 97

While those do stick out, they look more like part of what could be a single individual's career. If said this

396 118
382 124
570 175
613 150
367 114
470 97

Then I would be much more concerned. Particularly because earlier in his career he is in that neighborhood a few times. Granted these are his top two OPS+ seasons. One thing I'd like to suggest is that everyone take a look at post #41, point three. That's where I go through my discount structure. I'm very anxious to get feedback about whether that feels OK to people. I've gone with .88 for AAA, and I feel pretty solid on that (see the estimating league quality thread for more about that), and I cascaded downward from there.

That said, I'll take a look at who the other big guys were that year and see what I can tell you. But not til later tonight.

Bus Clarkson, 1954, step by step.

As I’ve written this, I realized that my description of my previous system was not very easy to read. Worse, I completely misdescribed the first two steps. I’ve rectified this and tried to flesh it out a little more here to make it easier to see what I’m doing.

Old system
1) AVG / lgAVG (if known)
.324 / .261 = 1.24

2) Take 1 and multiply by NL avg

1.24 * .274 = .340

3) Take 2 and multiply by .88 (average factor for AA level that I guesstimated based on discussions surrounding Chris's research)

.340 * .88 = .299

4) Repeat for slugging using square of avg factor (also per Chris’ research).

.602 / .388 = 1.55

1.55 * .427 = .662

.662 * (.88^2) = .513

5) Estimate walk rates if there's any idea about them. [I estimated these by bb/h in the old system because there situations in which it was more helpful, though I don’t now remember why. But I kept on doing so even when it was obvious I could have used better information. Shirk pointed this out at least once.]

Clarkson walked 104 times with 176 hits, or .59 per H.

TxL walked .47 times for every hit.

NL walked .377 times per hit.

.59 / .47 * .377 = .473 BB/H

6) Prorate playing time as g/tm g * 3.8 AB/g. [I early wrote 3.5, should be 3.8, sorry for any confusion.]

157 / 162 * 154 * 3.8 = 567 AB

7) Figure out average, slg, and walk rates in the playing time from 6.

.306 * 567 = 174 H

.536 * 567 = 304 TB

.473 * 174 = 82 BB

8) Season to taste with OPS+ and/or SFWS.

117 and 24.2


This is how that season looks in the current setup. [I explained the first two steps incorrectly when I previously posted. Should be correct below. I’m sorry, this stuff doesn’t trip off the tongue.]

1) Find his RC.

Clarkson created about 141 runs in Texas.

2) Find his offensive context by multiplying his games played times the league’s R/G times his pf.

157 * 4.66 * 1.00 = 732 runs

3) Find neutral context by multiplying his games times 4.50

157 * 4.50 = 707 runs

4) Figure his neutralized runs created by dividing 4.5 by his league’s r/g times his pf and multiplying all that times his RC

4.5 / (1.00 * 4.66) * 141 = 136 runs created

5) Convert his RC to a neutral MLB environment by multiplying step 4 by .825, the runs conversion factor I’ve used for double-A, as the Texas League was classified.

.825 * 136 = 112 runs created

6) Use the Willie Davis Comment Method (aka just Willie Davis) to figure out how the player's original stats translate into the new environment.

Without going into gory detail .324/.433/.602 becomes .285/.387/.530.

His PAs drop from 652 to 603.

7) Sum all converted seasons to get neutral career averages.

These turn out to be .284/.375/.467.

8) Figure playing time as g/tm g * 154 * ~4.2 PA/G (variable based on the era, league, player's likely slot in the order, etc....)

I actually used 4.1 for Clarkson because I had questions about his durability and didn’t want to over shoot. The NL of his era was at 4.3 PA per batter per game.

157 / 162 * 154 * 4.1 = 612 PA

9) Subtract the PAs figured in step 6 from step 8: this is how many PAs you will apply his career average to in building his full season.

612 – 603 = 9 PA

10) Figure how many runs he needs to create in the PAs from step 9 based on his career neutralized RC/PA rate (and his nearest three, if it's possible).

Since there’s no interruptions in Clarkson’s stats in this era, I use the average of his nearest three neutralized seasons, plus his career neutralized rates. 45% for previous seasons, 55% for career. This turns out to be about 1.6 RC for these 9 PAs. I know it’s pretty picayune.

11) Use Willie Davis to figure his batting line based on the RC in step 10 and the PAs in step 9.

Because it’s a small number of PAs, there’s more potential for wacky results: .388/.489/.640 in those 9 PAs. All of his other remainder projections are very, very close to the neutralized career totals with the exception of 1955 which is well below his norms and is only 14 PAs.

12) Add the batting lines from steps 6 and 11 (the neutral and "remaining" batting lines) to get the total neutral season line.

.287/.388/.531, 115 RC

13) Divide 4.50 into NL season's R/G.

4.56 / 4.50 = 1.01

14) Multiply 12 by the player's neutral rc/g to get his projected NL RC/G.

1.01 * 114.58 = 116.1

(rounding and partials end up a little above the 112 + 2 expectation for our final neutral total)

15) Use Willie Davis to figure his projected NL component line.

His projected line is .288/.390/.534.

16) Figure his WS as though he were on a .500 team scoring at the league average and figure OPS+ the usual way.

Starting with OPS+, he’s a 139, NL is at .345/.424.

For WS, I based it on pages 17-20 of the WS book. The idea is to create an average team and figure what the player’s contribution to such a theoretical team would be.
a) Figure the league’s runs/outs. Retain.

For the NL of 1954, it’s about .166. Might depend on whether you figure it from the pitching side or batting side.

b) Figure an average team’s expected RS as lg R/G times scheduled games.

4.56 * 154 = 702

c) Figure the average team’s marginal runs as .48 times step b.

.48 * 702 = 337

d) Figure available WS for batters as wins times three times .48.

77 wins * 3 * .48 = 111 WS

e) Figure the player’s expected background runs as step a times player’s outs times .52.

.166 * 377 * .52 = 32

f) Figure player’s marginal runs as RC minus step e.

116 – 32 = 83 marginal runs

g) Divide step f into step c.

83 / 337 = .246

h) Multiply step g times step d for the batting WS.

.246 * 111 = 27.3 BWS

i) For his fielding WS, I gave Clarkson average SS FWS through 1948, then average 3B FWS thereafter using the norms from SFWS, and figuring his games as PA / 4.1.

612 / 4.1 / 38 = 3.9

Total of 27.3 + 3.9 = 31.2 WS

I did do a double-check with SFWS (which I used in the previous system), and it reported 32 even WS with 28.1 through batting. I abandoned SFWS because I don't think it holds together well for comparing widely varying environments. Clarkson is within ten WS of his SFWS projection, but when we get to Marv Williams, I think we'll see bigger differences.

Here then are the final lines which the old and current system return:



The problem with the old system was threefold:
-Didn’t have enough info on the NgLs (since ameliorated by the HOF study)
-Didn’t treat walks correctly, underselling them badly.
-Discounts on AVG and SLG didn’t match up with reality since they made the comparison between higher-level players look like a comparison with lower-level players (see my previous long post about Joe Minors for an example of how this happens). This is mostly the result of my slugging discount cutting too deeply, but is also somewhat true of the average discount.

That said, there’s a couple possibilities about this particular season.
1) Maybe the TxL was not a strong league despite its AA status? If so I need to ratchet down its discount. OK, but I’ll need to ask whether or not that means each AA league of this period should be marked down the same way. And also, I think Marv Williams will show a lot of TxL time, and he’s the next guy I’m going to post up, so perhaps a comparison with him would be helpful before jumping the gun.
2) Maybe my system is completely wrong. That’s always possible, and now you have enough information to tell me so!
3) Maybe this is the Bonds in AA problem.
4) Maybe Bus Clarkson is a severe outlier.

Either way, I do believe this system is closer to baseball reality as we understand it than the previous one and I’m looking for ways to make it better. And i suspect that oddities like Bus's 1954 are matters of tweaking rather than breakdowns.... But I've been wrong in these threads many times before. ; )

Eric,

What's your source for the .825 discount factor for runs for Class AA? It seems much too high. James used .82 for runs for Class AAA (see Buzz Arlett thread). If you're looking at the discounts used by Davenport, keep in mind that they are being applied to EQA, not to runs; equivalent runs in the BP system is proportional to EQA raised to the 2.5 power.

Something doesn't make sense when trying to make OPS+ and BWS correspond. Let me take 3 of Bus Clarkson's better years (1948,1953,1954) and compare him to actual NL player performances.

1948

Bus Clarkson MLE .288/.367/.511 135 OPS+ 552 PA
Bob Elliott actual .283/.423/.474 145 OPS+ 675 PA

Bob Elliott wins BWS but just 22.7 to 21.7. That's with a 10 point higher OPS+ and 123 more plate appearances.

1953

Bus Clarkson MLE .309/.426/.495 140 OPS+ 570 PA
Jackie Robinson actual .329/.425/.502 139 OPS+ 574 PA

Bus Clarkson wins BWS easily, 25.3 to 21.1. The years look nearly identical.

1954

Bus Clarkson MLE .288/.390/.534 139 OPS+ 613 PA
Stan Musial actual .330/.428/.607 166 OPS+ 705 PA

Stan Musial wins BWS by about 1, 28.5 to 27.4. Stan the Man outpaces Bus by 27 on OPS+ and has 93 more plate appearances.

I know OPS+ and BWS don't have to line up perfectly. Yet it shows how you can get from a career translation 126 OPS+ player in 8800 PAs yet have 402 Win Shares. Eric seems to be working the numbers properly but why are Win Shares coming out high? Or is it OPS+ is coming out low?

Brent,

I'm glad we're talking about the discount structure because it's an area that I think needs the most feedback. You are right that James pegged AAA at .82. In the Estimating League Quality thread (around posts 145-155), I noted that Davenport sees .86 as the number. In the same thread, KJOK suggested that Clay's and Bill's AAA ratings were too low. He said that he had found .88 to be more likely based on an informal survey of research and his own observations. In the thread, I canvassed a few of Clay's articles to see what he said about various leagues. And then I created a discount table based on this that goes like this:


I decided to follow KJOK's .88 because the high minors were more independent from 1920-1955 than they are in the last fifteen years (Clay's study period) or the late 1970s and early 1980s (Bill's study period). Now, what's complicated is that the minors weren't stratified quite the same. Then it was

Open
AAA
AA
A
B
C
D
(And in a couple years there was 1A as well, which i think went between AA and A.)

I don't think that the lower leagues were specifically rookie or instructional leagues as they are now, and they were full-season leagues (or nearly so). And after integration, every stratification had some NgLers. Even so, IIUC, ex-major leaguers were moving up and down the MiL ranks and so were career minor leaguers.

Anyway, so my thought was to raise AA up a bit to .825 to match the .88 for AAA. I haven't made a similar adjustment for the lower leagues, feeling like the Barry Bonds in double-A question is a strong enough argument to keep them at roughly the same as Davenport's findings. So my current discount schedule:

Open/NgL: .90
AAA/1947-1948 NgL: .88
AA/1949-1950 NgL/MxL: .825
A: .75
B/CPL: .7
C: .65
D: .60

If the group thinks that I should go back down to .86 for AAA and .80 for AA, I'd be happy to do that. I'm not wedded to .88 and .825. And in general, I'd love to know if anyone's got any other thoughts on the subject.

Al,

I think the answer to your question could be something to do with:
a) number of team wins
b) RC vs. OPS+
c) a player's teammates.

I'd assume the variation is located somewhere within that area. And there may be partials here and there to tip things toward better hitters in non-theoretical MLE seasons.

For old Buster,
a) the team wins in all his seasons is locked at 77
c) all of his teammates are theoretical robots who march along at the league average and who among the other seven-plus lineup slots will accrue marginal 254 runs, or about 35 per lineup slot. That means the other lineup slots are worth about 12 batting WS, en route to a likely 15-18 total WS per lineup slot. As your examples above point out, this is sometimes an advantage, sometimes not. An advantage, probably, in comparison to a player playing in a loaded lineup. A disadvantage to a player in Wally Berger-like circumstances.

I could be wrong, though. I'm just guessing.

A few notes on Clarkson:

1) Riley, FWIW, seems to think considerably more of Clarkson than Holway does, devoting more than a full page to him. He gives him the 1918 birthdate, but says Clarkson played football and baseball at Wilberforce University (which is near Dayton, Ohio), and signed with the Pittsburgh Crawfords after he graduated—which actually matches better with the 1915 birthdate.

2) How certain are we that there’s no league record for Clarkson in 1937-38? Holway doesn’t list him as a regular for those years, but that book not always absolutely reliable. Clarkson isn’t included in the Macmillan encyclopedia (at least the edition I have), and he wasn’t nominated for the Hall of Fame, so I’m not sure where else stats for those years would have been published.

3) On the Texas League in 1953-54: I haven’t done a complete study, but just looking at the stats for those seasons, it seems rather obvious that Clarkson was the most valuable player in the league for both years.

Here are the best hitters (by OPS) I could find in the 1953 Texas League (asterisk marks someone who never appeared in the majors):

Joe Frazier, of, .332/.440/.560/1.000 – 30 years old
Les Fleming, 1b, 37 .317/.460/.538/.998 -- 37
Bus Clarkson, 3b, 38 .330/.453/.528/.981 -- 38
*Harry Heslet, of, .269/.350/.551/.901 -- ??
Frank Kellert, 1b, .289/.388/.501/.889 -- 28
Willard Brown, of, .310/.354/.519 -- 38

And in the 1954 Texas League:

Les Fleming, 1b, .358/.473/.562/1.035 -- 38
Bus Clarkson, ss/3b, .324/.433/.602/1.035 – 39
Frank Kellert, 1b, .316/.394/.585/.979 – 29
*Howard Boles, of, .306/.368/.607/.976 -- ??
Ed Mickelson, 1b, .335/.396/.538/.934 – 27
Willard Brown, of, .314/.353/.573/.926 – 39

There was also:
Ken Boyer, 3b, .319/.380/.506/.887 – 23

Clarkson and Fleming were easily the best two hitters in the Texas League in 1953-54, with Fleming just a shade ahead (they played for the same two teams in 1954, btw, and were evidently dealt together from Beaumont to Dallas).

Going through the other hitters:

Joe Frazier had a cup of coffee with the Indians in ’47 (1 for 14), but got another chance with the Cardinals in ’54, at the age of 31, and showed he could still hit (.295/.388/.500 in 103 plate appearances), but tailed off after that, finishing his major league career with a 97 OPS+ in 324 plate appearances.

Frank Kellert, a right-handed first baseman, got a chance with Brooklyn in ’55 (.325/.385/.575 in 91 plate appearances), even appearing in the World Series (1 for 3). He was shipped to the Cubs in ’56 (at the age of 31), and stopped hitting. He finished with an 85 OPS+ in 277 plate appearances.

Ed Mickelson had three cups of coffee, in 1950, 1953, and 1957, going an aggregate 3 for 37.

Ken Boyer, who in ’54 had probably the best offensive season by an infielder other than Clarkson during the two years covered (though he was still nowhere near Clarkson as a hitter at this point), went up to the Cardinals in ’55 and posted OPS+’s of 95, 123, 94, and 121 in his first four seasons.

Les Fleming is the most intriguing player, and makes an interesting comparison with Clarkson. They’re almost the same age—Fleming was born on August 7, 1915, just short of five months after Clarkson. He didn’t get established in the major leagues until 1942, when he hit .292/.412/.432 (OPS+ 144) with 106 walks as a 26-year-old rookie first baseman with Cleveland. He even garnered a few votes for MVP (finishing 25th). Then WWII intervened. When Fleming returned from the war, he was pushing thirty and had to fight to regain his job. He put up pretty good numbers (.329/.382/.493 in ’45), but got squeezed out anyway, and ended up in the minors. His overall record in the majors: .277/.386/.417, OPS+ 130 in 1572 plate appearances. Compare Eric’s MLEs for Clarkson above: .280/.371/.461, OPS+ 126.

In the 1953/54 Texas League, Clarkson and Fleming, almost the same age, seem like mirror images as offensive players (Clarkson batting righthanded, Fleming lefthanded): both high average power hitters with considerable control of the strike zone. Then consider that Fleming was (and had always been) a first baseman, while Clarkson was still spending a considerable amount of time at shortstop at the age of 39.

Thanks, Gary!

I've been making out some lists of these two seasons.

NgLers who appeared in the TxL in 1953-1954 (no minimum)

Bob Boyd (28)
Willard Brown (38-39)
Clarkson
Bill Greason (28)
Chuck Harmon (29)
Benny Lott (26)
Terrance McDuffie (44)
Jose Santiago (24)
Pat Scantlebury (36)
Jim Tugerson (31)
Charlie White (24)
Jesse Williams (40)
Sam Williams (31)
Lonnie Summers

Also, I made a list of guys who I recognized as possibly being MLB players. In other words, I haven't gone name-by-name yet, but I'm hoping to get to that this weekend (we lost power during a snowstorm last night, so these two lists are by candlelight...but that's the kind of HOM dedication I've got, I'm the Eckstein of HOM).

up = ascending to MLB
down = descending from MLB
between = between MLB stops

1953
--------------
Ed Bailey (up)
Larry Jackson (up)
Ryne Duren (up)
Joe Koppe (up)
Chico Garcia (up)

Les Fleming (down)
Max Lanier (down)
Frank Mancuso (down)
Fred Martin (down)
Cliff Ross (down)
Willard Brown (down)
Clarkson (down)

Cloyd Boyer (between)
Cal Hogue (between)


1954
------------------
Ken Boyer (up)
Don Blasingame (up)
Hal Smith (up) [don't know which one, both were Cs]
Jim Brosnan (up)
Luis Arroyo (up)
Billy Muffet (up)
Karl Spooner (up)
Norm Sherry (up)
Frank Robinson (up---token app)
Paul Schramka (up)
Jackson
Koppe

Clyde King (down)
Roy Hartsfield (down)
Fleming
Clarkson
Brown
Lanier
Martin
Ross

Bob Boyd (between)

There are probably more guys whose names I simply didn't recognize because I'm less familiar with the second-tier players of the 1950s and 1960s than I am of the second-tier players from many other decades. I'll find more info this weekeend.

Doctor South Central PA
In Clarkson's case, he is listed by Lester and Clark on the Philadelphia NNL roster for 1937-1938 but appears to have no record of league play. This may have been his woodshedding period or something like that, maybe seeing a lot of time in non-league contests after the regulars had played a few innings. I don't know. Someone like Gary A could probably tell us what that means.

Gary A
1) Riley, FWIW, seems to think considerably more of Clarkson than Holway does, devoting more than a full page to him. He gives him the 1918 birthdate, but says Clarkson played football and baseball at Wilberforce University (which is near Dayton, Ohio), and signed with the Pittsburgh Crawfords after he graduated—which actually matches better with the 1915 birthdate.

2) How certain are we that there’s no league record for Clarkson in 1937-38? Holway doesn’t list him as a regular for those years, but that book not always absolutely reliable. Clarkson isn’t included in the Macmillan encyclopedia (at least the edition I have), and he wasn’t nominated for the Hall of Fame, so I’m not sure where else stats for those years would have been published.

IIUC the emerging picture, Clarkson was age 24.1-24.6 during the 1939 baseb