MLEs, Take 2

Thanks again to all. Based on your feedback, I’ll take another stab at Buzz Arlett’s major league equivalencies, with the following changes from my first attempt (in # 21 and 22 above):

Run environment.

While it would be most appropriate to figure the run environment for the league, then calculate park factors for the parks BA played in, I will take a shortcut and simply take his team’s runs scored per game averaged with an estimate of runs allowed per game as an estimate of his run environment. This shortcut has been used in early sabermetric work, for example, in the 1986 BJHBA.

Since I haven’t yet been able to obtain actual runs allowed per game, I use the following “inverse Pythagorean” method to estimate them. Usually the Pythagorean method projects winning percentage based on runs scored and allowed:

W/(W+L) = (R/G)^2 / [(R/G)^2 + (RA/G)^2].

In this case, we know (thanks to KJOK and Gary A) the wins, losses, and runs scored for BA’s teams (for 1928-29 and 1932-36), so I will solve for estimates of runs allowed:

RA/G = (R/G) * [(W+L)/W – 1]^(1/2)

For example, for 1928, the Oaks were 91-100 and scored 4.495 runs per game, so we estimate that they must have allowed about 4.712 runs per game. Averaging Oakland’s runs scored and estimated runs allowed produces a estimated run scoring environment of 4.603 runs per game, which compares to the major league average of 4.730 runs per game. The environment adjustment factor is 1.027 = 4.730/4.603. I estimated the following environment adjustment factors for the years for which KJOK and Gary A provided data:

1928 1.027
1929 1.063
1932 0.792
1933 0.773
1934 0.727
1935 0.807
1936 0.813

Baltimore and Minneapolis were clearly very batter-friendly during this era, while Oakland appears to be somewhat pitcher-friendly. For his other seasons with the Oaks (1918-27 and 1930), I will use 1.045, the average of the two years for which I have an estimate.

The environment adjustment factor for each season is then multiplied by the quality-of-competition factor (.82) to obtain the “m” factor that is used in the calculations described in posts # 21 and 22.

Other changes in assumptions

Next I reduce his estimated fielding win shares to 1 per 64 games, which works out to about 1.74 win shares per 1000 innings. This puts his defensive estimates in the same general range as Harry Heilmann (1.79 WS/1000 innings), Jeff Burroughs (1.69), and Greg Luzinski (1.73), though I don’t think BA was really quite that bad. I also will assume that he pinch hit and didn’t accumulate fielding win shares in 5 percent of his games prior to 1931 and in 10 percent of his games after 1931.

Finally, I had been assuming that long seasons were only relevant to his PCL years, but Gary’s data show that the IL also held long seasons (167 team decisions in 1932 and 164 in 1933). I’ll make a season length adjustment of 8.3 percent (1 – 154/168) for those years. The AA appears to have held 154 game seasons.

Revised MLEs

Pitching:

Year InnP ERA
1918 153.3 3.29
1919 268.0 3.66
1920 329.0 3.52
1921 245.7 5.33
1922 288.0 3.38
1923 096.3 7.02

Batting:

Year Gms ABs RNs Hit 2B 3B HR RBI BB AVG OBA SLG RC
1918 026 070 008 014 04 00 01 007 09 .200 .291 .300 006
1919 043 105 009 028 05 01 01 013 12 .267 .342 .362 013
1920 049 133 016 031 01 02 03 017 16 .233 .315 .338 014
1921 049 097 008 020 04 01 02 009 12 .206 .294 .330 009
1922 057 131 015 029 06 02 03 014 16 .221 .306 .366 015
1923 115 332 050 103 22 03 13 067 36 .310 .378 .512 064
1924 149 521 081 160 41 11 22 096 57 .307 .375 .555 109
1925 146 529 080 170 35 07 17 096 56 .321 .386 .510 104
1926 149 495 092 178 37 09 23 092 50 .360 .418 .610 126
1927 144 490 081 161 38 04 20 081 52 .329 .393 .545 105
1928 146 416 072 142 33 02 16 073 43 .341 .403 .546 091
1929 154 538 098 190 50 05 26 127 55 .353 .413 .610 136
1930 136 460 087 156 41 04 20 094 48 .339 .402 .576 106
1931 121 418 065 131 26 07 18 072 45 .313 .387 .538 086
1932 135 439 084 127 24 02 32 086 67 .289 .383 .572 096
1933 146 450 078 130 29 01 23 085 66 .289 .380 .511 087
1934 116 397 063 104 25 01 24 079 55 .262 .352 .511 071
1935 122 394 060 122 21 01 17 067 40 .310 .373 .497 073
1936 050 181 037 049 08 02 10 035 27 .271 .365 .503 033
Tot 2053 6596 1084 2045 450 65 291 1210 762 .310 .381 .530 1344

Win Shares:

Year BWS PWS FWS WS
1918 02.5 09.5 --- 12
1919 05.0 14.9 --- 20
1920 04.8 25.7 --- 31
1921 02.8 07.4 --- 10
1922 04.6 28.0 --- 33
1923 13.6 -2.9 1.4 12
1924 24.3 ---- 2.2 27
1925 20.4 ---- 2.2 23
1926 32.0 ---- 2.2 34
1927 24.0 ---- 2.1 26
1928 21.2 ---- 2.2 23
1929 29.7 ---- 2.3 32
1930 20.2 ---- 2.0 22
1931 ---- ---- ---- 16
1932 20.6 ---- 1.9 23
1933 20.2 ---- 2.1 22
1934 13.6 ---- 1.6 15
1935 14.8 ---- 1.7 17
1936 05.9 ---- 0.7 07
Tot 280.3 82.6 24.6 405

The last half of BA’s career now looks a lot less impressive than in my first iteration, but the PCL years are still very strong. Even if one drops his 1918-19 and 1936 seasons, we’re still left with a 16-year career worth 361 win shares, with 4 seasons worth 30+ WS and 7 more worth 20-29.

Responding to some more of your comments –

KJOK: In post # 25, you provided BA’s fielding statistics for 1929, and I noticed that he only played 115 games in the outfield, while his batting statistics for that year show 200 games (and 722 at bats). I’m wondering about the difference. BJ’s short biography mentions “an ill-advised attempt to return regularly to the mound in 1929,” but doesn’t give any details. I also wonder about games played at other positions, such as first base. If you have any info, I’d be interested.

In # 29, Joe Dimino wrote:

I think a very conservative trimming would be to cut all of the pitching (based on Brent's post #2), and start his MLB career in 1924. If your estimates are correct, that leaves him as a 335 WS guy.

Because I think each voter ought to be free to use their own judgment about how much of his career should be trimmed, I’ve provided MLEs for all of the years, even though I don’t plan to use them all in my own evaluation. But FWIW, I think cutting off all of his pitching seasons is too harsh. By the end of 1920 (assuming there aren’t any surprises lurking in missing data on his Oakland PCL run environment) he had clearly established that he was capable of pitching very well at the major league level and was also unusually good hitter for a pitcher. IMO, he clearly would have advanced to the majors at that point if it weren’t for two unusual circumstances that were entirely out of his control: (1) the temporary refusal by the PCL to participate in the minor league draft (according to the link provided by Cliff Blau in # 5 above, the PCL opted out of the draft entirely during the years from 1919 to 1924; afterward, if I read the rules correctly, only one player a year could be drafted from any team); and (2) the 1920 spitball ban and fact that BA’s spitball was grandfathered in the PCL, but wouldn’t have been in the majors. We’ve already elected two HOMers, Coveleski and Faber, who were grandfathered by the major league spitball rule. During 1920-22 BA was using the same pitch very effectively at the highest level where he was permitted to do so. Personally, I don’t think it is fair to exclude BA’s value as a pitcher, at least for the years 1921-22 after he’d established that he was a top major league-caliber pitcher.

Joe also wrote:

I'd really like to see actual numbers of scheduled games, and something that tells what the run environments of his leagues were, before I'm willing to vote based on this information though.

I agree that more information would help us ensure the accuracy of these projections. Regarding schedule length, I’ve tracked down win-loss records for a few more PCL teams from the 1920s, and they all seem to be in a range between about 190 – 204 games, so my estimate 200 is probably ok. But it would be better to get hard information. While I now have run environment information for the last half of his career, including two seasons in Oakland, it would still be good to fill in a few years in the first half.

In # 30, karlmagnus wrote:

449 WS for Buzz Arlett (whose OPS+ of 138 in 1931 is perfectly OK for a poorly fielding outfielder, but not at all special) looks about as overstaed as the NL figures, and gives me further confidence that the translation process is fatally flawed, with at least a 20% bias to the upside.

I’m sure my latest revisions won’t satisfy you either, Karl. I’ve already applied an 18 percent penalty (in units of runs) for differences in quality. If I were to add another 20 percent, it would imply that the relative gap between the major leagues and the top minors then was wider than the distance between the majors and the double-AA leagues today. No way can that be true! In fact, I think the institutional arrangements of the 1920s argue that the gap may have been smaller – PCL teams were routinely able to hold onto outstanding players like Earl Averill for a couple more years after they had clearly established their ability to be stars at the major league level. That would rarely happen today.

During those years lots of players passed through the PCL and other top minor leagues on their way to or from the majors and can be directly compared with Buzz. BA played side by side with Ernie Lombardi from 1928-30 and out-hit him every year. He pitched side by side with Ray Kremer from 1919-22 and out-pitched him (BA was 95-71 over 4 years, Kremer was 64-77). In 1924, at age 31, Kremer moved to Pittsburgh where he began a very effective ML career (143-85, 113 ERA+).

If I had the time and the data, I’d love to do the study that provides definitive MLEs for the minor leagues of the teens, twenties, and thirties. But even without doing the study, I’m sure that 18 percent is much closer to the true quality discount than 38 percent.

In # 31 Tom H wrote:

The one major league season Arlett had was clearly the Worst by these projections. That makes me real queasy that he really would have racked up 30 WS per year had he played MLB the other seasons.

The short biography by Bill James in the The Baseball Book 1991 discusses BA’s 1931 season and BJ surmises that Arlett probably suffered a minor injury in mid August that hurt his performance for the rest of the season. BJ notes that BA played well in the first half of the season, playing right field regularly and hitting .324 through August 9. That was followed by an interval on the bench, followed by usage patterns that suggested he “was nursing an injury of some sort.” (I assume BJ is referring to his appearances as a pinch hitter or at first base.) As the Phillies regular outfielders, Chuck Klein moved from left field back to right field and Doug Taitt entered the lineup in left field.

Over most of his career, BA appears to have largely avoided injuries (other than the one that ended his pitching career) and stayed in the lineup pretty well. So I think his lower performance in 1931 is explainable. Also, the 1931 season looks less out of line now than it did with my first set of estimates; with new environment adjustment factors for the IL and AA, his career now follows a more typical pattern of decline after age 30 (1929).

Another comment on his 1931 season is that playing right field in the Baker Bowl must have required some significant re-learning. The Baker Bowl’s right-field wall was sort of the mirror of Fenway’s left-field Wall, except that it was located about 40 feet closer to home plate. I agree with Chris Cobb’s comment that moves (especially jumping from one level to another) often requires an adjustment period of at least one season. More convincing evidence that the MLEs work would have to be based on a group of players, each of whom spent several years in the PCL and in the majors. There were plenty of such players during the 20s and 30s, but I don’t have the database to do such a study.

In # 34, Chris Cobb wrote:

I'd also be interested to see how the short-form projections that Brent is using compare to the projections one would come up with by matching Arlett's seasons to their closest major-league equivalents by raw stats and finding the ws from there. That's how I do the NeL projections for batting.

Chris, I’d welcome alternative projections by you or others. While I think I understand the basic approach you are using, I didn’t feel confident enough with the specifics to attempt to do the projections that way, so I opted instead for the simple formulas that BJ provided.

OK, I've got Oakland's W/L records:

1918: 40-63 6th
1919: 86-96 5th
1920: 95-103 6th
1921: 101-85 5th
1922: 88-112 6th
1923: 91-111 7th
1924: 103-99 4th
1925: 88-112 6th
1926: 111-92 2nd
1927: 120-75 1st
1928: 91-100 5th
1929: 111-91 4th
1930: 97-103 5th

Chris, I’d welcome alternative projections by you or others. While I think I understand the basic approach you are using, I didn’t feel confident enough with the specifics to attempt to do the projections that way, so I opted instead for the simple formulas that BJ provided.

I've done a short set of projections using my approach, which is to find the closest match for the player's rate stats in either the AL or the NL (alternating seasons) and giving him that player's win shares, prorated for differences in playing time and differences in production. Here are the batting win shares my method produced for Arlett, 1923-27, using AL in even numbered years and NL in odd-numbered. For ease of comparison, I've put Brent's numbers in parentheses following.

1923 14.1 (13.6)
1924 22.5 (24.3)
1925 20.6 (20.4)
1926 28.1 (32.0)
1927 24.3 (24.1)

As you can see, the matches are pretty close. Three factors may be involved in the discrepancies, which could be investigated by further testing. First, they could be random. Second, they may be attributable to differences in competition levels between the leagues. Third, they may be the effect of the full WS system tamping down on extreme values in a way that the short form does not. I'll try to run more seasons and cross-check across leagues to make clearer the source of the discrepancies.

Obviously, they're not huge, so if Brent's MLE projections are on target, the win share estimates are also pretty much on target.

If further comparison of short-form win shares to comparison win shares confirms that the short form overestimates top seasons, it would be appropriate, I think, to pro-rate Arlett's peak batting seasons as they appear in Brett's projections by a small amount -- say 5-10%.

Further investigation of offensive levels in the PCL and of the competition differential between the majors and the top minors should be done, if possible, but given what we know so far, I'm convinced that Arlett needs to be considered as a serious candidate.

Thanks, Brent, for all the great work you've done on Arlett's case. A good assessment of the PCL will be valuable for our consideration of candidates like Averill, too!

KJOK: In post # 25, you provided BA’s fielding statistics for 1929, and I noticed that he only played 115 games in the outfield, while his batting statistics for that year show 200 games (and 722 at bats). I’m wondering about the difference. BJ’s short biography mentions “an ill-advised attempt to return regularly to the mound in 1929,” but doesn’t give any details. I also wonder about games played at other positions, such as first base. If you have any info, I’d be interested.

I noticed this too, but didn't have time at the time to do additional searching.

Turns out he played 74 games at 1st base with:
Po-601
A-39
E-14
Pct - .979

He was 16th out of all 20 players listed as playing 1st base in Fielding%, and 10th out of the 10 with more than 50 games, so doesn't look like he was much of a 1st baseman....

There's a very nice biographical article by Bill James in "The Baseball Book 1991."

Does James (or any other source) explain why Buzz went to Baltimore after his one season with the Phillies?

From the Washington Post, March 5, 1931:

Winterhaven, Fla - The spotlight blazed on Buzz Arlett, might oak from Oakland, Cal., as he went through his first workout today with the Phillies. The new outfielder is 6 feet 3 and scales 225. "What a break to get in the majors at last," he says.

"There were reports that I didn't care about coming up, and that I was satisfied to stay in Oakland, but that was wrong. I gave my best for thirteen years and they held me back. At the end of last season I said I was through, so they sold me."

New York Times, March 8, 1931:

"The big leaguers are interested in the debut of Buzz Arlett with the Phillies. Arlett is a Coast veteran, 32 years old, and he has never been up before. He is supposed to be a poor fielder, but the players like him because he is a husky fellow with a heavy wallop. The boys in the trade have a great respect for base hits."

BJ’s short biography mentions “an ill-advised attempt to return regularly to the mound in 1929,” but doesn’t give any details.

SABR's Minor League Baseball Stars shows Buzz pitching very little 1924-27 (3-0, 8 G, 27 IP). In 1928 he was 1-0, 7 G, 27 IP and in 1929 he was 1-4, 5.76, 17 G, 61 IP. The next year, his last with Oakland, is the last shown in his pitching record, 3 G, 3 IP.

Regarding possible injuries, he seems to have missed a chunk of games in 1928. Gary's list has Oakland with 191 decisions while Arlett played 160 games. Also in 1930, Buzz played 176 games and the team had 200 decisions.

Washington Post, December 19, 1931:

BUZZ ARLETT RETURNS TO MINORS:

Aged Rookie Traded to Baltimore for Russ Scarritt

New York, December 18 (U.P.) - Opportunity to make good as a major league ball player came too late for big Buzz Arlett and the slugging 220-pound outfielder today was returned to the minors after a single season of prominence with the Philadelphia Philles.

Long regarded as the finest prospect in the minor leagues (he remained with the Oakland Pacific Coast for 13 years only because major league clubs were unwilling to pay the $100,000 purchase price demanded by Oakland owners). Arlett was allowed to slip back to the minors with no big league team willing to pay the $7,500 waiver price for his services.

With every big league team granting waivers, Philadelphia traded Arlett to the Baltimore International League club in exchange for outfielder Russell Scarritt.

Arlett played with the Phillies for a single season, but his heavy hitting and colorful play made him a prime favorite with the Philadelphia fans.........he ranked as the greatest slugger in the Coast League. Despite his bulk, his agility and fine throwing arm made him a valuable defensive gardener, and the Oakland club set a prohibitively high price on his services.

When the Phillies finally purchased him after thirteen years of minor league service, Arlett had started to slow down. He opened the season with a flash and gave promise of earning ranking as the most valuable "rookie" of the year, but injuries and advancing years took their toll. By the middle of August, Arlett had lost his regular place in the Phillies lineup and served only as a pinch hitter for the balance of the year.....

Arlett's chief weakness was in the field, and he finished far down among the outfielders with a percentage of .955.

From NY Times Feb 24, 1932:

"The truth is that outfielders are really hired for their hitting. The ability to catch flies is taken for granted. This is sometimes taking too much for granted. Ham Hyatt, Billy Bagwell and BUZZ ARLETT proved that...."

Arlett went to spring training with Brooklyn in 1933, and it was noted that he had dieted down to 210 lbs.

From March 21, 1933 Washington Post (apparently Arlett had been cut by Brooklyn) :

"...the reason big league clubs don't buy Buzz Arlett, the Oriole slugger, for pinch hitting duties is because Arlett demands a major league salary and he isn't worth it as a pinch hitter..."

Arlett's chief weakness was in the field, and he finished far down among the outfielders with a percentage of .955.

Huh... and RF in the Baker Bowl was not exactly spacious, either. Caroms off the screen might have been tricky, though. Anyhow, the Phils actually moved Klein to left to make room for Arlett and as soon as Arlett was gone, they hurriedly placed Klein back in right. Its not like the Phils were the pitching-and-defense minded team of their era, either.

Thanks. Lots of useful information is coming in.

Gary A: Thanks for the W-L records and rankings. Would you be able to tell us how many teams were playing in the PCL each of those seasons?

Anyone: It still would be very helpful if we could get Oakland team runs scored for at least a couple more years during 1918-27 or 1930.

Chris: Thanks for the alternative projections. You're right that the latest-and-greatest runs created formulas that are used in the Win Shares book tend to tamp down extreme values compared to the basic runs created formula that I used. But I'd have guessed the difference from that source would be smaller - maybe 5 percent rather than 10 percent.

KJOK and David: Great quotes. You've convinced me that Arlett was a poor outfielder (though the statistics never seem to imply that he was the very worst outfielder around). It's still hard for me to understand that teams weren't willing to live with the glove to have the bat that had ranked eighth 8 in OPS+. My guess is that Philadelphia's decision to trade him (and the other teams' willingness to let him clear waivers) may have reflected several points of confusion that sabermetricians now have a better handle on:

- the sudden drop in NL offense (from 5.68 R/G in 1930 to 4.48 in 1931) probably made his offensive contributions seem less impressive.

- while the fact that the Baker Bowl boosted offense was known, estimates of the magnitude of the effect weren't available, so there was added uncertainty about the quality of any Philly player.

- I've observed that good players on bad teams are often unappreciated (and their fielding deficiencies are sometimes exaggerated).

It's clear from my latest MLE projections that Arlett's HOM case will have to rest primarily on what he did in Oakland rather than on what he did in Philadelphia, Baltimore, or Minneapolis. But the fact that after leaving Philadelphia he continued to play professional baseball successfully at the next-to-highest level for 4 full seasons and contributed to two AA pennant-winning teams ought to count for something.

Yes, I'd forgotten that the PCL was still a six-club league through 1918. In 1919 it expanded to eight, and stayed there till the 60s.

The PCL source I have--*The Pacific Coast League: A Statistical History*, by Dennis Snelling--doesn't give team totals, but it does (for some reason) have a list of teams that led the league in a few categories. It looks like, until 1948, Oakland only led the league in batting ave. once (in 1921, .305). Oakland didn't lead the league in runs scored per game until 1947. They NEVER led the league in home runs, through 1957. There are, unfortunately, no pitching categories.

Also--it seems that the two major themes of KJOK's quotes are fielding AND money. Arlett was roundly thought to be a poor fielder, AND he (and Oakland) seemed to be tough negotiators, perhaps a little too tough. It would be interesting to have salary info for him (I have no idea if anybody's ever done research on minor league salaries in this era). Were minor league teams willing to pay him more than the majors were in the thirties?

Snelling also provides some league leaders in fielding:

In 1920, Arlett led the league in errors by a pitcher with 10 (tied with 3 others)

In 1924, Arlett led the league in errors by an OF with 19.

He never led the league in OF assists, or fielding pct. at any position.

One theory I have about Arlett's fall from MLB:

Thru June 6th, Arlett was hitting .384 and apparently had been leading the NL batting race.

He hit his 11th HR on June 13th.

On June 19th, he suffered a broken thumb, and apparently had to sit out for at least a few weeks.

Apparently, when he came back from that injury, he was either rusty, or in a slump, or just suffering "regression to the mean", etc. but obviously from that point on he was a less effective player, and his shortcomings seem to have started to be magnified, with his hitting skills overlooked...

Well, luckily I found what I was looking for, sort of:

As of June 5th, Arlett was 61 out of 159, for a .384 average (and 10 HR's), which means that for the remainder of the year he was:

70 out of 259 (8 HR's), for a .270 average. That is quite a dropoff, and by September the more recent performance is probably what was focused on.

And FYI, here is the link to Arlett's PCL batting and pitching stats:

Buzz Arlett PCL Stats

Possibly a minor quibble, but I still have to make it...

Since most of the players he’s being compared to had two or three good years in the minors before establishing themselves in the majors, in evaluating Arlett I dropped his 1918 and 1919 seasons and gave only half credit for 1920. He wouldn’t have made it to my ballot purely as an outfielder – I see him as similar to, but not quite as good as Cravath and Tiernan. But when you add one and a half seasons as a 30-win share pitcher, I believe he legitimately qualifies.

My understanding is that the spitball ban began in 1920. So if we're assuming a normal career path, and that he would have debuted in 1920, doesn't that mean the pitching has to be out, as he relied on the spitter? Certainly seems wrong to me to give him credit for pitching that there's no serious way he could have done in the major leagues.

----------------------------

As for Arlett overall, without the pitching (which I'm not giving), I can't see any way he's ahead of Cravath, who's not on my ballot. I'm probably being closed minded on the topic, but so be it.

Speaking of Cravath, if someone were to do a similar Win Shares calculation to the above for his "missing" years, it would be much appreciated. :)

Michael:

I guess I see the fact that the spitball ban went into effect at the same time as Arlett established himself as a star-quality pitcher as an historical accident that was entirely out of his control and shouldn't be held against him. The fact that he wasn't allowed to pitch in the majors during those years didn't make him any less of a pitcher. He was a great pitcher during those years, just in the PCL, rather than in the NL or AL. And in the days of independent minor leagues, the PCL was just as important to the fans in Los Angeles and Oakland and Salt Lake City as the NL was to the fans in Pittsburgh and Chicago.

If we aren't penalizing Coveleski and Faber for the spitball grandfather clause, I don't think it's fair to penalize Arlett either.

By the way, I agree that MLE's for Cravath would be useful. I am giving him credit for his minor league seasons, but I may be too conservative.

KJOK: Thanks for the information on the injury. Characteristically, Bill James seems to have gotten the essence of the story right but the details wrong. A broken thumb? Ouch!

Also, thanks for the link to his statistics. In his picture he doesn't look all that fat - why I'm sure there must be at least a couple dozen fatter major leaguers today! ;-)

I did a little browsing today to see what I could find. I got my hopes up when I found that the Library of Congress had on-line versions of a number of old issues of the Spalding Official Base Ball Guide, including the one for 1922. Unfortunately, it looks like for several years in the teens and early 20s Spalding made its readers by a second book, the Spalding Base Ball Record to get the minor league statistics. (Kind of like Bill James making people buy Win Shares to get the background info for the NBJHBA.) I suspect that copies of the Record will be even harder to locate than the Guide.

I also found an Oakland Oaks Web site that has info on the Oaks Ball Park.
http://oaklandoaks.tripod.com/oakspark.html

Oaks Ball Park was used from 1913 to 1955. The page also says that from 1913 to 1922 some of the Oaks' home games were played in San Francisco. I don't know how many games or how it might have affected the run environment. However, the info in Gary A's # 63 strongly suggests that the Oaks played in a neutral to pitcher friendly environment through most of their history.

Michael,

Let me try answering your question again another way. When I think about major league equivalents for minor league or Negro League players, I'm not literally thinking of what their career would have been if they had played the same years with a major league team. Instead, I'm thinking of taking their actual accomplishments in whichever league they played in, and converting them to units of major league value, in the same way that dollars are converted into euros or pounds. I'm saying that a 2.77 ERA in Oakland in 1922 was comparable in difficulty, after adjusting for the differences in quality of competition and run environment, to a 3.38 ERA in the majors the same year. I do not believe that the fact that the pitching was done for Oakland rather than a major league team made it any less valuable - just that the quality of competition was different, so a minor league pitcher who was as good as a major league pitcher (and there were many minor league players during those years who were as good as major league players) would be expected to have a lower ERA and better statistical record.

If we aren't penalizing Coveleski and Faber for the spitball grandfather clause, I don't think it's fair to penalize Arlett either.


See, I don't think that is a valid comparison. They *were* grandfathered. Arlett would not have been. The valid comparison is to the, I'm sure, many pitchers who debuted or would have debuted in the majors from 1920-1922 who had a key part of their pitch assortment banned. Is anyone arguing that those guys should receive extra credit? That's essentially what you're arguing with Arlett.


Or another way...we elected Ross Barnes, who took advantage of the fair-foul rule. Does that mean we should have been looking for post-1876 players who would have been great with that rule in place?

In my opinion, no; you get credit for what you actually did in the majors, or you actually could have done in the majors. You reference the Negro Leagues...Mendez had an entire section of his career as a SS. It was not particularly great, but it was above replacement, at a minimum. He gets no credit for that from me because I don't believe he would have done that in the majors. Wilbur Rogan was a tougher call, but I believe he'd have at least been given a shot as a two-way player. I understand those who might have not voted for him on the grounds that they don't think he could have. Like I said, he was a tough call.

Michael,

I don’t think of it as giving Arlett extra credit. Extra credit would be giving Arlett credit for pitching when he didn’t, but he did pitch during 1920-22, actually more than 1100 innings during those 3 years, and that’s what I think he ought to get credit for. When you say, no, he wouldn’t have been able to pitch in the majors, it seems to me you also need to ask whether in that case he would have moved to the outfield three years earlier, and then maybe he would have hit another 75 home runs (or whatever). Playing that kind of “what if” scenario is what I don’t want to get into—Arlett was a pitcher during those years, he was covered by a grandfather clause in the PCL the same as Faber was in the AL, and he was an excellent, ML-quality pitcher during those years.

You’re right that other young pitchers were caught by the ban, and some of them lost their careers as a result. That was clearly unfortunate and unfair to them, but we don’t know who they are, and even if we did I wouldn't give them credit for pitching that they didn’t do, just as we don’t give Barnes credit for baseball he didn't play because of the rules change. The difference in Arlett’s case is that he was allowed to continue pitching.

It seems that your premise is that only baseball played at the ML level really counts. For the period since about 1960, I’d agree with you; everyone understands that the minor leagues are primarily training programs for the majors and the seasons don’t really count in any meaningful sense. If someone tells you that someone had a great season in Triple-AAA, your reaction isn’t, “did it help his team win its minor league pennant?” Instead, it’s, “he’s probably going to make the ML roster next year,” or “why didn’t they call him up earlier when we could have used him.” If a minor league team is in a tight pennant race and its major league parent decides it needs the team’s best player to be called up for 3 weeks to sit on the bench and pinch-hit, it has not qualms about doing so because everyone understands that that the only pennant race that really matters is the major league one.

We need to understand is that 1920s baseball wasn’t like that at all. Minor leagues were largely independent and operated pretty much the same way the major leagues did. It was more like college sports, where there are stronger leagues and weaker leagues, but every league is more or less independent and equally important to its own teams and fans. (Actually, college basketball isn’t so much that way anymore, but I hope you're understanding my point.) Just as in college sports, in 1920s baseball most of the best players were in the best leagues (the majors), but not all of them were. If Baltimore thought its players were worth more competing in the IL than Philadelphia was willing to pay for them, then they would stay in Baltimore. And since the Baltimore baseball fans cared more about the IL than they did about the AL or NL (in those days before television or radio), there’s no reason a player couldn’t have been worth more to a top minor league team than to a major league one. If you were to pick the 50 best performing baseball players for each season prior to 1930, my guess is that about 30 would have been major leaguers, another 10 to 15 would have been in the Negro Leagues, and 5 to 10 ought to be minor leaguers. Don’t get me wrong; I think the very best players (the Ruth’s, Cobb’s and Johnson’s) were all major leaguers (or Negro Leaguers), but the next tier included some minor league stars who were every bit as good as some of the major league all stars.

When I think about giving Arlett or other players credit for minor league play, I’m not imagining a major league career that they didn’t have. I’m taking the career that they did have and adjusting it to its equivalent in difficulty at a major league level. It’s more like adjusting in the 1880s/90s for differences in quality of competition between the NL and AA or PL; playing in either league counts, but the quality of the leagues wasn’t all the same, so it needs to be adjusted for.

Adendum to Arlett Fielding:

With the Phillies, Arlett gets a "94" from Baseball Prospectus, and gets 2.0 Fielding Win Shares (in about 840 innings).

You’re right that other young pitchers were caught by the ban, and some of them lost their careers as a result. That was clearly unfortunate and unfair to them, but we don’t know who they are, and even if we did I wouldn't give them credit for pitching that they didn’t do, just as we don’t give Barnes credit for baseball he didn't play because of the rules change. The difference in Arlett’s case is that he was allowed to continue pitching.

When I think about giving Arlett or other players credit for minor league play, I’m not imagining a major league career that they didn’t have. I’m taking the career that they did have and adjusting it to its equivalent in difficulty at a major league level.

One pitcher who comes to mind is Frank Shellenback. He was a spitball pitcher who pitched for Chicago White Sox in 1918 and some of 1919. He was in the PCL when the spitball was outlawed and was thus not grandfathered in. He went on to compile a 295-178 record in the PCL. Should we adjust his PCL statistics to MLE? As Brent says, he did pitch those years, just as Arlett did.

Yes, Shellenback was an excellent pitcher - another one we should really be looking at. If someone can post his minor league data, I'll gladly do the MLE calculations - or you are welcome to do them yourselves.

Major league equivalents for Buzz Arlett, version 3

I’m making several modifications to my MLEs for Arlett. I located a library with copies of most of the old Reach Guides on microfilm. (I must say, after spending several hours staring at microfilm, my admiration for those of you who do serious historical research has gone up.)

- I have left the quality of competition parameters the same as shown in earlier posts # 21, 22, and 49, based on the factors given by Bill James: 0.82 for runs, about 0.92 for AVG, and about 0.89 for SLG (note – the Bill James method separately adjusts each element of the batting line, so the adjustments for AVG and SLG are only approximate). Gadfly has recently argued that the quality of competition may have been higher during the 1920s. I am sympathetic to his argument, but am holding off on making any change to these factors until I can complete a small study of players who moved between the PCL and the majors during the 1920s.

- For run environment, the Reach Guides provide information on the team’s runs scored per game, which I use together with an estimate of runs allowed per game based on the “inverse Pythagorean” calculation shown in post # 49. My microfilm did not include the 1925 and 1926 Guides, so I do not have data for 1924-25; for those two years I used an average of factors for 1922-23 and 1926-27. Also, the 1919 Guide did not contain runs scored for 1918, so I used an average of factors for 1919-21.

Oakland was pitcher friendly during the late 1920s, but seems to have been more hitter friendly during the early 1920s. (Partly, this reflects a shift in the league, which included hitter-friendly Salt Lake until 1925; however, even relative to the league Oakland appears to have been more hitter friendly during the earlier period.) The following table gives Oakland’s runs scored, games played, W-L record, estimated run environment = (runs + est. runs allowed) / 2, and the run environment relative to the averages for the PCL and for the major leagues. For example, for 1919 I estimate the Oaks’ run environment to have been 4.69 runs per game, which is 8.0 percent above the PCL average and 21.0 percent above the MLB average.

Year Runs Gms W-Loss REnv PCL MLB
1919 0834 183 86-096 4.69 108.0 121.0
1920 0850 199 95-103 4.36 106.4 100.0
1921 0996 187 101-85 5.11 103.5 105.2
1922 0822 202 88-112 4.33 092.8 088.9
1923 0969 203 91-111 5.02 096.9 104.3
. . .
1926 0906 203 111-92 4.26 095.2 091.9
1927 0966 195 120-75 4.44 089.9 093.3
1928 0865 192 91-100 4.61 095.1 097.5
1929 1034 202 111-91 4.88 090.5 094.0
1930 1014 200 97-103 5.15 092.6 092.8

- I again used the formulas for converting minor league statistics to MLB equivalents from the Bill James 1985 Baseball Abstract, which are listed in post # 21.

- Given the lack of data on batters walks for the PCL, I decided to use a more sophisticated method. I started with a weighted average of Arlett’s walk rates for his 1931 major league season and of the MLEs for his 1932 and 1933 IL seasons, which are based on actual walks data. (The average was weighted 5/12 for 1931, 1/3 for 1932, and 1/4 for 1933.) This average is extrapolated for the PCL years with two adjustments: (a) for age, using the age profile shown on Tango Tiger’s Web site (TT shows that batter walk rates tend to increase throughout a player’s career), and (b) for changes in the ML average walk rate.

- I also did a better job of adjusting for PCL season length. Because the PCL did not follow a balanced schedule, I am not sure how many games were scheduled in each year. Instead, I adjusted for schedule length from the number of decisions by the team with the maximum number of decisions each season. For the missing years (1924 and 1925), I used the number of decisions by the Oaks. The numbers used are: 1919-182, 1920-199, 1921-188, 1922-201, 1923-202, 1924-202, 1925-200, 1926-203, 1927-196, 1928-192, 1929-202, 1930-201.

- For runs created, I have switched to a formula appearing in the Bill James Handbook 2005. I believe that this formula, which places “the individual hitter in a ‘neutral solution’ of eight ordinary hitters,” effectively eliminates the upward bias of my WS estimates for high-performing seasons that Chris Cobb noted in my earlier Arlett MLEs (see post # 52). The version I use is:

(A+2.4*C)*(B+3*C)/(9*C)-0.9*C
A = H + BB
B = 1.125*1B + 1.69*2B + 3.02*3B +3.73*HR + 0.29*BB
C = AB + BB

- I realized that I had been assigning too many batting WS to the years that Arlett was pitching. I fit a linear regression to seasons of several pitchers for which their OPS+ was 75 or above and obtained the following formula, applicable to seasons with OPS+ above 56:

bWS = (AB+BB)*(-.024024 + 0.00043449*OPS+)

- Influenced by Chris Cobb’s discussion of Negro League work loads, I decided to reduce Arlett’s innings pitched by another 7.5 percent beyond the proportional reduction implied by the difference in schedule length.

- Comparing the pitching WS to ML pitchers with similar innings and ERA, I decided that the short form formula was overestimating his pWS, so I multiplied the result by a factor of 0.90.

Here are Arlett’s MLE batting statistics.

Year Age Gm ABs Rns Hit 2B 3B HR RBI BB Avg OBA SLG OPS+ RC
1918 19 026 069 007 013 03 00 01 006 05 .188 .243 .275 063 005
1919 20 047 112 008 027 05 01 01 011 09 .241 .298 .330 089 012
1920 21 050 133 015 031 01 02 03 017 11 .233 .292 .338 079 015
1921 22 052 102 008 020 04 01 02 009 09 .196 .261 .314 054 009
1922 23 057 131 016 030 07 02 03 015 13 .229 .299 .382 082 016
1923 24 114 325 046 097 21 03 11 061 37 .298 .370 .483 131 060
1924 25 147 517 081 160 41 11 22 096 59 .309 .380 .559 153 108
1925 26 146 531 081 172 35 07 17 098 66 .324 .399 .512 139 110
1926 27 147 492 095 179 37 09 24 095 63 .364 .436 .622 188 126
1927 28 147 502 084 167 40 04 21 085 63 .333 .407 .554 160 111
1928 29 152 433 075 148 35 02 17 076 58 .342 .420 .550 161 098
1929 30 152 533 097 188 50 05 26 126 75 .353 .433 .612 170 133
1930 31 135 460 089 157 41 04 21 097 62 .341 .420 .585 153 109
1931 32 121 418 065 131 26 07 18 072 45 .313 .387 .538 138 085
1932 33 135 439 084 127 24 02 32 086 67 .289 .383 .572 158 095
1933 34 146 450 078 130 29 01 23 085 66 .289 .380 .511 152 089
1934 35 116 397 063 104 25 01 24 079 55 .262 .352 .511 133 072
1935 36 122 394 060 122 21 01 17 067 40 .310 .373 .497 136 074
1936 37 050 181 037 049 08 02 10 035 27 .271 .365 .503 130 034
Total 2062 6619 1089 2052 453 65 293 1216 830 .310 .387 .531 146 1361

His pitching MLEs:

Year InnP ERA
1918 153.3 3.29
1919 272.3 3.66
1920 306.0 3.52
1921 241.7 5.33
1922 265.0 3.38
1923 088.0 7.02
Total 1326.3 3.85

And here are his batting, pitching, fielding, and total win shares:

Year bWS pWS fWS WS
1918 00.2 08.5 --- 9
1919 01.8 13.6 --- 15
1920 01.5 21.6 --- 23
1921 00.0 06.5 --- 7
1922 01.7 23.2 --- 25
1923 12.4 -2.7 1.4 11
1924 24.1 ---- 2.2 26
1925 22.2 ---- 2.2 24
1926 32.1 ---- 2.2 34
1927 25.7 ---- 2.2 28
1928 23.2 ---- 2.3 25
1929 28.9 ---- 2.3 31
1930 21.0 ---- 2.0 23
1931 ----- ---- --- 16
1932 20.3 ---- 1.9 22
1933 20.9 ---- 2.1 23
1934 13.9 ---- 1.6 16
1935 15.1 ---- 1.7 17
1936 06.2 ---- 0.7 7
Total 271.1 70.8 24.6 382

This evening I will also post MLEs for Earl Averill. I hope to do the calculations for Berger in the next couple of days.

Brent, this is great stuff!

I look forward to seeing what Berger and Averill look like by comparison.

Brent-

I agree with Chris Cobb's assessment (great stuff) and have already made my comments on the conversion rates in the Averill thread, but the first thing that comes to mind when I look at your Arlett MLE is that the ABs seem too low.

In 1931, Arlett had 418 at bats for the Phillies in 121 games. He actually played 107 games in the field and pinch hit 14 times (with 4 hits for a .286 BA). So he had 404 at bats in 107 starts.

154/107x404=581 at bats in a full season.

For a check, I'll run 1929 since (in that season) you have Arlett having a career high 535 ABs.

Arlett with Oakland 1929: 200 games, 722 ABs
Oakland: 202 games
PCL 1929: 1628 games, 55774 at bats
NL 1929: 1232 Games, 43030 at bats

(I used the National League since that is the Major League Arlett actually played in.)

Arlett Games 154/202x200=152
Arlett AB 152/200x722=549
League adjusted (43030/1232)/(55774/1628)=1.019
Adjusted at bats=560

That still seems too low. I think what is happening here is that Arlett's AB are being artificially lowered by the BB conversion.

In other words, you need to adjust for Plate Appearances, not just at bats. When you adjust for PA, some BB will become AB if the BB rate is lowered (and that would also work in reverse). If you just adjust for AB, these PA disappear.

And, of course, this phenomenon will affect a player that walks a lot much more than one who does not walk much. Of course, there are no BB figures for Arlett in the PCL but it is known that he would take a walk.

I think his actual ABs in 1929 should be more like 575-580.

Interestingly, my seat of the pants analysis of Arlett showed him as a player capable of hitting over 300 HRs in the Majors with a BA of around .320 to .330. Your analysis, if you accept my BA conversion rate and adjust for the missing ABs, hits that right on the nose.

I look forward to your work on conversion rates for the PCL so that I am no longer a chorus of one.

gadfly is entirely correct. Thanks for looking closely at Brent's work!

For what it's worth:

In this era, the average number of PA per player per game tended to be about 4.2 (Arlett is very close to that for 1931), with variation from place in the batting order for individual players. When I need to make an off-the-cuff estimate of PAs in relation to games played, that's what I generally use.

The statistical projections regarding Arlett (though not the win shares estimates, which are higher) confirm my suspiscion that the very best career minor leaguers were among the top 500 players of all-time, maybe even the top 300, but that none of them quite crack the top 210. That's really a pity as this project would have been much more fun if we had to sprinkle a few career minor leaguers onto our ballots.

How come the WS projections are so far off the stats projections? Eyeballing the career statistics set out in post 76, I would have pegged Arlett at between 250 and 275 WS.

Daryn,

What leads you to believe that the WS are off? Are you remembering the pitching win shares?

Brent has Arlett with 7449 PA, a 146 OPS+, and 271 bws.

Harry Heilmann, a near contemporary, has 8960 PA, a 148 OPS+, and 325.1 bws.

If you prorate Heilmann's bws by Arlett's at bats, you get 270 bws.

That leads me to think that Brent's bws are pretty reasonably scaled.

They'll go up a bit, actually, once the Arlett's PAs are adjusted.

Thanks, everyone, for your comments.

Gadfly wrote:

…the first thing that comes to mind when I look at your Arlett MLE is that the ABs seem too low.

First, to explain how they were calculated – AB were determined residually by holding outs (= AB – H) constant, except for scaling for the number of games played. For 1929 his actual number of outs was 452 (= 722 – 270) (see post # 8 by Kelly from SD). Scaling this by the difference in schedule implies 345 outs = (154/202) * 452. After calculating 188 hits, I added the 345 outs to get 533 AB. A few comments:

- Holding number of outs constant is pretty standard in similar exercises done by other baseball analysts. In addition to Bill James, I believe the same assumption was used by Steve Treder in his recent analyses on hardballtimes.com, and my recollection is that Davenport makes a similar assumption.

- The assumption is generally justified because 27 outs is the one thing that stays (approximately) constant as the run scoring context changes. If you imagine moving a player and his team from one scoring environment to another, then I am sure that this is the appropriate assumption.

- For going from the minors to the majors, however, it may be more relevant to consider whether the player’s share of total team outs is likely to change. It seems to me that there are two cases to consider.

- For the very good player, moving to a better team is likely to increase his share of the team’s outs. An extreme example is Barry Bonds, who in recent seasons has used an unusually small share of his team’s outs. If he could be placed on a team of his “peers” (I guess I have in mind Ted Williams, Mickey Mantle, Lou Gehrig, Josh Gibson, etc.) his share of team outs would surely go up.

- For a marginal player, on the other hand, moving to a better team is likely to lead to a decline in plate appearances and his share of team outs, because a good team would be more likely to platoon him or use him irregularly. This factor suggests that if Arlett had spent his final seasons in the majors rather than in the AA, he might have been used as a pinch hitter or backup player. Consequently, the distribution of AB across seasons probably needs to change, but it is not clear that the total is that far off.

- Perhaps a more important factor in Arlett’s case is that his actual AB data for the PCL (see post # 8), it appears to show an unusually low average AB/G, at least for some seasons. For example, in 1928 he plays 190 games, but has “only” 561 AB. Two possible explanations come to mind.

- My estimates of BB could be much too low. Unfortunately, I only have actual walks data available for 1931 and later, so the estimates of BB for the PCL years are entirely based on my assumptions.

- Another possibility is that the Oaks may have used him as a pinch hitter fairly often. We know that the PCL seasons were unusually long, there were long distances to travel, and there must have been quite a few double headers. It seems possible to me that they may have given Arlett some rest by using him as a pinch hitter a number of times each season. It would be interesting to compare his fielding games to his total games. (Unfortunately, I didn’t think to photocopy the fielding statistics when I was going through the microfilmed Guides.)

I'll try to come up with a more appropriate distribution of AB in the next iteration of these MLEs. (I view each iteration as a learning opportunity).

Does anyone know what Arlett's principle position in the outfield was for each individual season?

Chris,

I wasn't intentionally ignoring the pitching stats, but I have to say I have trouble giving them credit for 100 WS. On the hitting, the 600 hits and 500 r+rbi difference between heilmann and Projected Arlett seem to be worth about a 75 WS drop, I guess. So maybe I'd expect those stats to come in a little over 300 WS (hitting plus pitching), not close to 400. Clearly, what's colouring my perception is that I don't think a minor leaguer could actually translate to first ballot NB status. For better or for worse, some pre-conceived notions are difficult to change. But like Samuel Jackson in Pulp Fiction, I'm trying to be a better person...

That should say 70 PWS, not 100.

Yesterday I wrote:

For example, in 1928 he plays 190 games, but has “only” 561 AB...

Another possibility is that the Oaks may have used him as a pinch hitter fairly often... It would be interesting to compare his fielding games to his total games.

It occurred to me that 1928 is the season covered by the 1929 Spalding Guide, which is available on-line at the Library of Congress site. So I looked up his playing and fielding games and learned... oops (he says sheepishly!) I've been using the wrong number of games played. Arlett's record actually shows him playing 160 games in 1928, rather than 190. He fielded 154 games in the outfield and pitched 7 games, not leaving much opportunity for pinch hitting! With that correction made, Arlett's AB for 1924-30 all vary between 3.44-3.74 per game--reasonable values for a player who drew a lot of walks. In my next iteration years I will have to slightly boost his MLE plate appearances for those years, as Gadfly sugggested, but will lower them slightly for 1932-36. (I will also double check all the data for typing errors.)

John Murphy asked:

Does anyone know what Arlett's principle position in the outfield was for each individual season?

I don't know, and I figure it will be difficult to find out (unless you're willing to look up old West Coast newspapers on microfilm). The Guides don't list outfielders games by outfield position, and I doubt any other reference does either. (See KJOK's post # 53 regarding games played at first base in 1929; I think he pretty much stayed in the outfield for his other seasons.)

I have always thought of Arlett as a right fielder (probably because that's what he played for Philadelphia), but when I went back to look at sources like the 1991 Bill James short biography, they don't mention his minor league positions other than "outfield." The PCL "All Century Team" (www.coastleague.com) lists him as their left fielder. I'm fairly sure he didn't spend significant time in center field, but don't know the split between the two corners.

Some of our younger voters may not realize that until quite recently, information on major leaguers' specific outfield positions played was generally unavailable. I don't know who assembled the information, but I imagine it must have involved collecting data from tens of thousands of microfilmed box scores. I salute whoever accomplished the project.

I'm no minor league expert or anything, so maybe this model is way off base. My perception of the PCL of Arlett's younger days is that it's sort of like the Japanese leagues of today, completely independent, lesser quality than the majors but better than the eastern minors, and with little or no standardized mechanism in place for funneling talent into the majors. They also played in some large markets and were far away. (LA and SF were Pittsburgh sized cities, larger than DC and Cincy and the split markets of Boston and St.Louis.)

To those who know more, in what respects is this model wrong?

Jimd,

I think you're mostly right - the 1920s PCL teams were independent and not vassals to the major league teams. Because they played in large markets, some PCL teams drew more attendance than some of the lesser major league teams. A player could be more valuable to a PCL team than to a major league team.

A lot of talent was always being funneled to the majors, but because the teams were independent, if they wanted to hang on to a player they could, at least until 1925 when they agreed to participate in the minor league draft. When they finally did agree to the draft, though, there were still a lot of restrictions and limits -- for example, I think they were allowed to hang onto a player for three years before he became eligible for the draft, and I think the draft limit initially was one player per team per year. I believe that most players of that era who moved from the PCL to the majors did so through trades or cash purchases rather than through the draft. Some of the cash purchases could be quite large -- in excess of $100,000.

The PCL, IL and AA were recognized as lower quality (or "minor") leagues relative to the majors, but they were not considered to be less "important"; they just played in smaller cities and had slightly lower quality players.

I thought I'd better explain that I'm not any kind of expert on the minor leagues or anything. In fact I feel kind of silly leading the charge on Arlett and other minor leaguers, since I don't even own the minor league reference books. I guess I first got interested with the first Bill James Historical Baseball Abstract (the 1986 edition), which contained quite a bit of material on minor league stars, much of which was dropped for the NBJHBA. So I'll gladly take advice from some of the posters here who are much more knowledgeable.

My general impression is that the PCL surpassed the IL and AA in overall quality beginning around 1925. It's just an impression, but I'm pretty sure that by the 1930s the PCL was considered the best. In the early 1920s, though, I think the IL's Baltimore Orioles were the great minor league team of all time.

the IL's Baltimore Orioles were the great minor league team of all time

IIRC, the IL was another minor league that went without a draft agreement for some time during the 1920's.

Also Baltimore, Buffalo, Newark, Milwaukee, Minneapolis/St.Paul, and New Orleans were all cities that were arguably of major league size (Cincinnati being the smallest single market, and half of St. Louis being smaller than that). Baltimore was the largest of them all, larger than Pittsburgh or LA, nearly the size of St. Louis, and only one team, not two. (I don't have Canadian historical census numbers, so I can't say anything about Toronto and Montreal in 1920.)

Brent:

Thanks for your response.

It drives me crazy when I see all of the outfield positions lumped together. Oh, well...

Based on a small sample of boxscores from the LA Times:
1924- RF
1925- RF
1926- RF
1927- RF
1928- RF
1929- RF/1B?

I'm gonna go out on a limb, and say he was a right fielder.

Cliff, you rule!

I discovered I made a mistake in doing the adjustment for run environment in calculating Arlett's pitching record--divided when I should have multiplied. Here's a corrected version. I have to say that I think this version looks more reasonable, particularly in comparing his W-L record to that of his team. This correction changes his career total WS (batting, fielding, and pitching) to 388.

Year InnP_ ERA_ pWS
1918 153.3 3.05 10.1
1919 272.3 3.02 20.3
1920 306.0 3.53 21.5
1921 241.7 5.07 08.5
1922 265.0 3.80 19.8
1923 088.0 6.73 -1.8
Total 1326.3 3.71 78.4

Regarding attendance at PCL games, I noticed that the 1931 Reach Guide, in describing the 1930 split-season pennant race in which second-half victor Hollywood beat first-half winner Los Angeles in the playoff, included some discussion of attendance:

"Therefore the strength of the circuit was concentrated at Wrigley Field, where these two teams play, and this park drew around 995,000 admissions from the two tenants, which was bigger than the 1929 mark."

The guide doesn't break it down by team, but if we divide by two that comes to 497,500 per team, which was larger than the 1930 attendance for 7 of the 16 major league teams.

In response to helpful comments and criticism, I’ve decided to make a couple tweaks to my minor league MLEs. Because the two changes pretty much offset each other, I will hold off on re-doing those for Buzz Arlett pending the completion of another study of players who moved between the PCL and the majors, but I will go ahead post a revised version of MLEs for new candidate Earl Averill, as well as my first set of MLEs for Wally Berger.

Gadfly (# 78) noted that my AB estimates were low. I agree – the AB should be reduced if they are in a higher run scoring environment, but they should not necessarily be reduced for the differences in quality of play (though they could be if the player is playing at a “marginal” level that wouldn’t ensure full-time play in the majors). I will restore the missing AB; for Averill this change raises his assumed plate appearances by about 5.5 percent.

Daryn (# 80) questioned whether my WS estimates were consistent with the MLE batting statistics. Chris Cobb (# 82) defended my estimates by comparing Arlett to Heilmann, but I realized that there is a problem with that defense, and that Daryn has a valid point. The OPS+ statistics I’ve presented are in comparison to the major league average OBP and SLG, but bbref calculates OPS+ from league averages excluding pitchers’ batting. That means the OPS+ that I’ve calculated are too high (I’m guessing by 4 to 5 points, but I will try to run some more precise calculations). After looking at Arlett’s MLEs and bWS relative to other similar players—e.g., Heilmann, Hack Wilson, Babe Herman, Chuck Klein—I’ve decided that the short-form batting Win Share formulas that I’ve been using are a little too high, so I’m reducing them by about 5 percent.

These two adjustments are essentially offsetting, so if you look at Averill’s revised estimates you will see little change in the estimated WS, though the batting lines will be larger because of the increased plate appearances.

By the way, I’ve decided that Hack Wilson is probably the most similar contemporary player to Arlett as a batter, though Arlett played longer, more consistently, and was generally a bit better hitter.

I've been giving Arlett's candidacy a lot of thought this week, and I have a question for you guys: Why has his support dried up so much amongst the electorate? I think he was only on one ballot in 1950...the MLEs suggest a clear HOMer. So what's the story? Is the issue one of not trusting the MLEs, or just not being able to pull the trigger on a career minor leaguer, or has something else been brought to light which detracts from his candidacy?

I'm interested in hearing input from as many people as possible, so please chime in.

Ran across this, which doesn't solve anything but strikes me as relevant to this thread:



Moneyball Redux
Slate talks to the man who revolutionized baseball.
By James Surowiecki
Posted Tuesday, June 10, 2003, at 10:15 AM PT

...

One of your most important insights is the idea that minor league batting statistics predict major league batting performance as reliably as major league statistics do. There have been certain players—think of 1980s players like Mike Stenhouse, Doug Frobel, Brad Komminsk—who seemed as though they would be terrific hitters but never really made it in the big leagues. Did they not get enough of a shot? Are they outliers? Or is there such a thing as a Four A (too good for Triple A, not good enough for the majors) hitter?

Well, no, there is no such thing as a Four A hitter. That idea, as I understand it, envisions a "gap" between the majors and Triple A, with some players who fall into the gap. There is no such gap. In fact, there is a very significant overlap between the major leagues and Triple A. Many of the players in Triple A are better than many of the players in the majors.

The three examples you cite are three very different cases. Stenhouse never had 180 at bats in a major league season, so one would be hard pressed to argue that he got a full trial.

Frobel is a different [instance], in which I think there probably wasn't a real strong case that he was a good hitter to begin with. Frobel hit .251 at Buffalo in '81, hit .261 at Portland—Pacific Coast League—in 1982. We would expect, based on those seasons, that he would hit .200, .210 in the major leagues, with a pretty ghastly strikeout/walk ratio—which is what he did. Then he had the one good year at Hawaii in 1983, looked like a better hitter, and fooled some of us into thinking that he was better than he was. But ... it was one year, 378 at bats, of performance that isn't that impressive. It wasn't enough, in retrospect, to conclude that he was actually a good hitter.

[Then] there are some players whose level of skill changes—drops—between two adjacent seasons or between two seasons separated by two or three years, usually because of an injury but sometimes because of some other factor. Frank Thomas is not the same hitter now that he was a few years ago; Tino Martinez isn't; Mo Vaughn isn't.

When those "disconnects" happen between major league seasons, we ascribe them to sensible causes—aging, injury, conditioning, motivation, luck, etc. Comparing major league seasons to minor league seasons, occasionally you get the same disconnect. Sometimes a guy simply loses it before he establishes himself in the major leagues. That's what happened to Komminsk, I think—he shot his cannons in the minor leagues.

I'm trying to make two general points here. Point 1: When there is a disconnect between a player's major league and minor league records, some people want to ascribe this to some mystical difference between major league baseball and minor league baseball. Unless you can say specifically what that difference is, this is akin to magical thinking—asserting that there is some magical "major league ability," which is distinct from the ability to play baseball. The same sorts of disconnects happen routinely in the middle of major league careers—not often as a percentage, but they happen. Everybody who plays rotisserie baseball knows that some guys you paid big money for because they were good last year will stink this year. It is not necessary or helpful to create some magical "major league ability" to explain those occasional disconnects between major league and minor league seasons.

Second point ... the creation of new knowledge or new understanding does not make the people who possess that new knowledge invulnerable to old failings. I can't predict reliably who is going to be successful in the major leagues in 2004, even if we stick with the field of players who have been in the major leagues since 2000. I can't do that, because there are limits to my knowledge, and there are flaws in my implementation of what I know. The principle that minor league hitting stats predict major league hitting stats as well as major league hitting stats predict major league hitting stats can be perfectly true—and yet still not enable me or you to reliably predict who will be successful in the major leagues in 2004, because I still have limits to my knowledge and flaws in the way I try to implement that knowledge.

David,

I am taking another look at Arlett, but I will say that he isnt' really ballot material for me. For one I am not going to give him pitching credit. He had a few decent years, but he had quitpitching by the time (age 24 or 25) he would have been moving to MLB anyways and he obviously didn't attract their attention as a pitcher.

Second, I give him no credit for 1923 and 1936 get no credit and I am debatign giving him credit for 1924. A player in the PCL did need to have a godo season to get him noticed. 1923 was his first season as a hitter so that is thrown out. He also wasn't terribly good at it (I give him 14 WS) so he may or may not have been able to attract attention. This brings in the question of 1924. In other words had he had a clear path to MLB would 1923 have been enough to get him there? Obviously 1924 was good enough but should it count?

Taking 1923 and 1936 off gives him 275 career WS with a decent peak. If I take 1924 off he is nto a contender really, 249 WS with a decent peak but not one that is eye popping in any way.