hot topics

Well, if John won't do it, I will.

Richard! Richard! Richard!

(What'd you think I was going to say?)

LOL

1921 Dick Lundy
NNL associate Bacharach Giants

(these stats are still incomplete, but should give you a good idea of his season)
G-33 (team 53)
AB-132
H-45
D-7
T-4
HR-3
R-24
W-3
HP-2
SH-1
SB-7
AVE-.341 (NeL .263)
OBA-.365 (NeL .324)
SLG-.523 (NeL .361)

Lundy was injured in Detroit in early/mid-May, at the beginning of the Bacharachs' western tour, and was out for about a month. He missed 18 games against top teams during that stretch.

1928 Dick Lundy
Bacharach Giants (manager) + Eastern League Stars

Batting
G-56 (teams 60)
AB-206
H-72
D-12
T-3
HR-4
R-36
W-10
HP-2
SH-3
SB-8
AVE-.350 (NeL .279)
OBA-.385 (NeL .333)
SLG-.495 (NeL .384)

Fielding-ss
G-56
DI-481
PO-117
A-170
E-21
DP-10
RF-5.37 (eNeL ss 5.09)
FPCT-.932 (eNeL ss .917)

Dick Lundy data from Holway

In Negro Leagues
1918 .125 for Hilldale; ss
1919 .351 for Hilldale; ss; all-star
1920 .286 for Bacharach Giants; ss
1921 .361 for AC Bacharach Giants; ba 1st, 3 hr (4th), 14 hr/550 (4th) 4 3b (1st), 4 sb (4th); ss; all-star
2-15 in playoff vs. Hilldale
1922 no data (Riley lists him with AC Bacharachs, hitting .335. Lloyd was with Bacharachs this season and Holway lists him at ss and a non-entity at 2nd; I hazard a guess that Lundy moved to second, or that Lloyd actually played second. Riley says Lloyd moved to second in 1924 when he returned to the Bacharachs as player-manager)
1923 .290 for AC Bacharachs; 9 hr (2nd), 25 hr/550 (3rd), 13 2b (4th), 4 3b (5th); ss; all-star
1924 .339 for AC Bacharachs; 4? 3b (4th); ss
1925 .297 for AC Bacharachs; ss
1926 .320 for AC Bacharachs; 17 2b (1st); ss; all-star, MVP
11-36 in World Series vs. Chi Am Giants
1927 .303 for AC Bacharachs; ss; 11 hr (2nd); ss
9-36 in World Series vs. Chi Am Giants
1928 .414 for AC Bacharachs; ba 2nd, 4 3b (2nd); ss; all-star
1929 .336 for Bal Black Sox; ss
1930 .344 for Bal Black Sox; ss
1931 .276 for Bal Black Sox; 13 hr/550 (3rd); ss; all-star
1932 .328 for Bal Black Sox; 3rd in league; ss; all-star
1933 .167 for Phi Stars; ss
1934 .279 for Nwk Dodgers; ut (easing out as full-time player)
1935 no data (listed as manager for Nwk Dodgers)
1936 no data
1937 .167 for Nwk Eagles; ut and manager

Vs. Major-League competition
1919 1-8 vs. major-league competition
1920 1-3 vs. major-league competition
1930 2-10 vs. major-league pitching


In Cuban Play
1920 3-8 in Cuban Play
1926 52-127 in Cuban Play
1927 44-137 in Cuban Play
1929 65-195 in Cuban play

Career
938-2998, .314 according to Holway
Median avg. 1918-1934, .303

I'll have win shares estimates as soon as I've incorporated Gary's more complete data for 1921 and 1928.

In 1934 Lundy was out most of the season with an injury--I *think* some kind of eye injury, but I don't remember for sure. I'll check to make sure over the next couple of days.

A little more Jamesian fielding analysis: in 1928, eastern shortstops accounted for 25.6% of their teams' assists (excluding catchers and outfielders). Bacharach Giants' shortstops accounted for 29.3%.

I'll check out what happened with Lundy in 1922, since I'm going to be trying to figure park factors for that year anyway.

Here's a little more on his Cuban League play:

1920-21

The Bacharach Giants entered the Cuban League as the third team (to Habana and Almendares), but despite headline players including Oscar Charleston, Louis Santop, Dick Lundy, and Dick Redding, they were only 4-11 by January 13, when they decided to forfeit all remaining games and return to the U.S.

8 AB, 1 R, 3 H, 0 XBH, 0 SB, .375.

1923-24

Lundy played with Almendares, which placed 3rd of 4 teams, 18-29. This is the year when most of the biggest stars (Charleston, Dobie Moore, Oliver Marcelle, Frank Warfield, Bill Holland, Dave Brown, Jose Mendez) were assigned to Santa Clara, which was 36-11.

90 AB, 11 R, 31 H, 3 2B, 2 3B, 0 HR, .344

1924 (Second season)
With Santa Clara sweeping the regular season, the 1923-24 season was ended early (in mid January), the weakest team was folded, and the players were redistributed for a special second season. This was very competitive, with Lundy's Almendares in 3rd place at 12-13, only 1 game behind first place Santa Clara.

53 AB, 3 R, 13 H, 0 SB, .245

(Bienvenido Jimenez is also listed as a SS with 46 AB, suggesting that Lundy may not have finished the second season.)

1924-25

Lundy is again with Almendares, and John H Lloyd, Oscar Charleston, Newt Allen, Biz Mackey, and Bullet Rogan also join the team, which this time won the title with a record of 33-16. Lundy leads the league in stolen bases with 11.

139 AB, 24 R, 49 H, 11 2B, 2 3B, 3 HR, 11 SB, .353

Note - the season was again ended early because Almendares was running away with the championship. To finish out the season, the players split into "All Cubans" and "All Yankees" squads for a contest between the Cuban and North American players. The "All Yankees" won the series 5-3.

1925-26

Lundy again played with Almendares, but has only 71 AB. John H. Lloyd is listed as the regular SS with 134 AB. Almendares has a 34-13 record, but 12 of the victories represent forfeits by a weak San Jose team. Almendares edges Habana for the title by 2 games.

71 AB, 27 R, 23 H, 1 2B, 0 3B, 0 HR, 0 SB, .324
(These numbers, especially the runs scored, look suspicious.)

1926-27

Miguel Angel (Mike) Gonzalez and Dolf Luque formed a separate league, known as Triangular, taking with them most of the better players. They played at the University of Havana stadium, which appears to have been more hitter friendly than Almendares Park, where most of the Cuban League games were played until 1930. Lundy played for the Alacranes and was one of four .400 hitters (along with Pablo Mesa, George Carr, and Oscar Charleston). With Dolf Luque, George Carr, Judy Johnson, and with Jose Mendez throwing his last victory before his death of tuberculosis, the Alacranes won the pennant in a 3-team league with a record of 22-15.

127 AB, 25 R, 52 H, .410 (extra base hits not recorded)

1927-28

Dick Lundy (and the other players) returned to the Cuban League, where he again played for Almendares. Habana was running away with the pennant, and after Almendares lost a game 18-4 to Habana on January 21, they "withdrew from the league in shame" and forfeited their final 8 games. Final record (including 8 forfeit losses) was 17-23. Other players on Almendares that season included Lloyd (at 1B), Marcelle, George Scales, and Dolf Luque.

137 AB, 26 R, 44 H, 9 2B, 2 3B, 2 HR, 1 SB, .321

1929-30

Lundy was with Almendares again. The Almendares Blues finished in 3rd place (of 4 teams) with a record of 23-26. Martin Dihigo was his most prominent teammate.

195 AB, 28 R, 65 H, 7 2B, 3 3B, 1 HR, .333

1930

The regular season broke down after 5 games because of a contract dispute with the management of the new stadium (La Tropical), so a special season, known as Unico, of 25 total games was held in November in the old stadium (Almendares Park). Lundy played with Almendares, which placed first of the four teams with a 9-4 record. Other prominent players on his team were Martin Dihigo and Judy Johnson. Lundy tied with Charleston for most hits, with 19.

56 AB, 11 R, 19 H, 0 2B, 1 3B, 0 HR, 3 SB, .339

Don't have all the details yet, but what happened in 1922 was that the Bacharach Giants' owner John Connor moved the club to New York City for the season, along with most of the important players (including Redding and Marcelle). The Columbus Buckeyes were disbanded, and Lloyd was transferred to the NY Bacharachs to become manager (Redding voluntarily stepped down). The Bacharachs were still associate members of the NNL (along with Hilldale), and so their player transactions were governed by the league.

In the meantime, interests back in Atlantic City started a rival team, sometimes referred to as the "original" Bacharach Giants, and signed Lundy away from Connor to manage it. Generally, they seemed to be the inferior team, as most of the previous year's team went to NY; Sam Streeter and Nip Winters did spend at least part of the season in Atlantic City.

The NY Bacharachs made two extensive tours of NNL teams in the west, and got some press in NY; I've hardly found anything about Lundy's Bacharachs.

Bump! Keeping the new candidate on the Hot Topics list. Trying to figure out how he rates compared to Sewell/Beckwith (or even Cronin/Appling).

DICK LUNDY MLEs

Here are my estimated MLEs of Lundy’s at bats, hits, batting average, total bases, and slugging average. First, the methodology.

Playing Time. I am using i9 projections for playing time EXCEPT

1) no major-league credit for 1916 & 1917
2) reduce playing time for 1918 to .7 of listed
3) Playing time for 1921 reduced to 125 games in light of Gary A.’s injury report
4) no reduction in playing time for 1922

NeL-ML conversion factor. I’m using a .87 conversion factor for batting average, .82 conversion factor for slugging, based on the data posted on the Beckwith thread. Slugging average must be estimated for most seasons because of incomplete data. I’m doing this by finding the ratio of slugging average to batting average in the seasons for which complete data is available (1921 and 1923) and using that to estimate a career ratio.

Offensive Context Adjustment. I am adjusting offensive context as follows. NeL offensive level is 1918-1921 .91 of ml, 1922-23 .95 ml, 1924 .96 ml, 1925 .97m1, 926 .98, 1927 .99, 1928 100%

This adjustment based on 4 data points. 1917 NeL west Median avg. .86 of ML avg (my study of Holway data), 1921 NeL avg. .91 of National League (Gary A.’s data), 1924 NeL west median avg. 96 of National League (my study of Holway data), 1928 NeL avg. .99 of Major Leagues (Gary A’s data). These points suggest a fairly smooth trend in the NeL gradually catching up to the majors in offensive production over a ten year period. More data would make this aspect of the conversion more exact.

Park Factors: Gary A. gives us park factors for the Bacharachs in 1920 and 1921, which are 107 & 100 (single season). I will reduce Lundy’s 1920 and 1921 totals by 2%. I will hope that the 1921 park factor for the Bacharachs foreshadows their park as playing fairly neutral through the 1920s!

Regression to the mean: I will calculate career MLE batting average and estimate career MLE slugging average for Lundy, then use those values to regress the outlying high and low seasons. I’ll list the unregressed values (which are based on the league conversion, the offensive adjustment, the park factors, and the slugging estimations), and then list the regressed values to the right, with the percentage of regression towards the mean. I balanced regressions up and down to keep the career values from changing by more than a percentage point and I regressed more extreme seasons based on smaller samples more, but there isn’t a “system” at work here.



Here’s what Lundy’s MLE at bats, batting and slugging look like. The left-hand columns shows straight converted values, the right-hand columns show regressed values for outlying seasons. I’m not confident in my handling of the regression, so I’ve provided both values. The career totals should be about the same, but more value would shift into Lundy’s peak seasons without the regression.

      I9s Non-reg.   Regressed seasons
Year  ab*  ba, slg      ba, slg 
1918  354 .120, .170   .200, .291 (50% up)
1919  540 .339, .458   .324, .446 (25% down)
1920  529 .271, .339
1921  488 .315, .455   .315, .455 (25% down)
1922  589 .307, .435
1923  621 .266, .462   .269, .450 (25% up/down)
1924  584 .307, .427
1925  653 .263, .362
1926  562 .284, .468   .282, .440 (50% down)
1927  645 .266, .431
1928  581 .305, .406
1929  410 .292, .424
1930  568 .299, .419
1931  533 .240, .373   .250, .382 (25% up)
1932  487 .285, .385
1933  363 .145, .196   .212, .304 (50% up)
1934  205 .243, .328
career 8715 at bats, .280 ba, .403 slg, 2441 hits, 3515 total bases

*taken from i9s except as noted above.

I’ve usd these values, plus fielding estimates, to calculate win shares. That’ll go in the next post.

Dick Lundy estimated win shares

Batting win shares determined by finding closest major-league matches for ba and slg listed in previous post, using i9s as a guideline for plate discipline, then prorating to Lundy’s estimated plate appearances. NL matches used in even-numbered years, AL matches in odd.

Fielding. Based on his reputation as possibly the best-fielding ss in NeL history, I’ve estimated fielding win shares for Lundy based on him being an A shortshop for his career, earning 6.19 ws/1000 defensive innings. This is about equal to Dave Bancroft, better than Joe Sewell. Based on his plate appearances, I’ve estimated him at

18460 def innings, 114.2 def ws

I’ve distributed these through his career, with some random variation between A+, A, and A- seasons during his prime and a gradual decline to a B defender at the end of his career.

Win Shares
1918 2.8 + 4.4 = 7.2
1919 17.3 + 7.8 = 25.1
1920 12.8 + 7.8 = 20.6
1921 14.4 + 6.5 = 20.9
1922 14.8 + 7.4 = 22.2
1923 13.6 + 8.6 = 22.2
1924 14.3 + 7.9 = 22.2
1925 8.0 + 8.8 = 16.8
1926 15.9 + 7.6 = 23.5
1927 15.1 + 7.6 = 22.7
1928 14.9 + 8.0 = 22.9
1929 8.3 + 5.1 = 13.6
1930 10.5 + 7.4 = 17.9
1931 7.3 + 6.6 = 13.9
1932 11.5 + 5.9 = 17.4
1933 0.0 + 4.4 = 4.4
1934 4.3 + 2.4 = 6.7
Totals 185.8 + 114.2 = 300

Thanks, Chris! These estimates really help.

career 8715 at bats, .280 ba, .403 slg, 2441 hits, 3515 total bases

My less sophisticated MLE's came out to:
AB - 8674
H - 2423
2B-464
3B-122
HR-234
BB-855
So-1579
SB-583
CS-268
R-1289
RBI-1207
AVE - .279
OBP - .344
SLG - .442
Runs Created - 1330
OPS+ - 122

These numbers are MLE'd to a 1955-2003 National League context.

First glance at these....he's in contention for a ballot spot, but he's not going to be very high, even if he makes it.

Taking Chris's estimates of the 3 major NL shortstops on the ballot (typos are possible here):

Career

Beckwith 319
Lundy 301
Moore 256

Best 3

Moore 103
Beckwith 93
Lundy 72

Best 5

Moore 158
Beckwith 149
Lundy 117

Best 7

Moore 205
Beckwith 201
Lundy 161

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

Career voters will like Lundy better than I do, but I'm a career/peak mix, and I can't see him above Moore. Hell, I'm not sure I can see him very close to Moore in a peak system.

Right now I have Lundy 19th, one ahead of Maranville (who has a longer, but more peakless, career).

Without any discount at all, he's Sewell, at the bottom of my ballot. Applying even a modest discount, he's off the bottom, though not very far off.

Chris Cobb once again, thank you very much indeed. You're not at all responsible for the use the rest of us make of your estimates, but they are extremely helpful. Waiting with bated breath to see whether I shall be forced to grit my teeth and vote for Foster (my strong instinct says no -- we've already elected one NL Foster I don't think we should have.)

I doubt that KJOK's MLEs are less sophisticated than mine.

I note that we agree closely on playing time and batting average.

The main difference in our estimates is slugging percentage, which is the aspect of my estimates in which there is the most guesswork, both with respect to the conversion factor and with respect to the data for Lundy himself.

I'm going to have to seriusly consider a spot for Lundy now. I'll have to work on him this weekend to see if he belongs or doesn't.

Chris's WS estimateds paint a picture of a player who is very close to Rabbit Maranville, but not quite as meritorious.

Their 162-adjusted peak/prime/extended prime are all quite close, though Rabbit's is a little better overall. And he's got just a little more career to boot.

Another guy to compare Lundy to, in a more modern sense, might be Dave Concepcion. Better than average hitters in their good years. I see Concepcion's peak as a little higher, but his career and extended prime as a little bit lower.

Are Chris Cobb's Win Share estimates for 154 or 162 game seasons?

DavidFoss,

I just looked at what would happen to Lundy if I prorated CC's WS estimates to 162 games. The effect is to have him draw exactly even with Rabbit. Their careers would be virtually indistinguishable.

My estimates are for 154-game seasons. The goal is to show what the player's performance would have been if he played in the major leagues that year instead of in the NeL, imagining that the same chance events of injury, hot-and-cold streaks, etc., would have occurred in the majors as they did in the NeL.

Thanks... 154 makes sense and makes it easier for me to compare to Sewell, Maranville, etc. I was just double-checking.

Chris, great work! Btw, the raw park factor I have for the Bacharachs in 1928 is 95...so, based on the best evidence we have now, you're correct in your assumption their park was pretty much neutral in its effects.

If his Slugging is .442 (as per KJOK) instead of .403, how much of an effect would this have on him Win Shares wise? Right now he is slightly above Joe Sewell, but still not in my top 20.

The Negro Leagues Register in the 8th edition of the Macmillan gives Lundy a league-leading 14 home runs in 1924. The complete stat line:

G-60
AB-228
H-84
D-9
T-3
HR-14
SB-11
AVE-.368
SLG-.618

The home runs are something of an anomaly--his next highest totals in the Macmillan are 7, 5, and 4 (3 times).

My estimates are for 154-game seasons. The goal is to show what the player's performance would have been if he played in the major leagues that year instead of in the NeL, imagining that the same chance events of injury, hot-and-cold streaks, etc., would have occurred in the majors as they did in the NeL.

Just to make this clear - my MLE's are for 162 game seasons and show what the player would do in an "average" MLB season (NL 1955-2003).

Okay, I see.

One more question. Gary A just stated that Lundy let the league with 14 HRs and no one else was really close. Again, does this seem like someone who would only have a .403 SLG?

Is it really possible that he hit all of his HRs in games taht were recorded as opposed to the games that weren't while his contemproaries did nearly the opposite?

Actually, according to the Macmillan stats, Oscar Charleston tied for the lead with 14.

Let me be clear: I'm not the source for these figures. These are from the stats compiled by SABR's Negro Leagues Committee for the Macmillan encyclopedia (edited by Dick Clark and John Holway). The same statistics are used in The Negro Leagues Book.

However, Holway's Complete Book paints a totally different picture. These are his ECL home run leaders for 1924:

Charleston, Harrisburg 9
Thomas, Hilldale, 7
Hudspeth, Lincoln Gts, 6
Beckwith, Baltimore 5

Which would mean, presumably, that Lundy had 4 or fewer. That's quite a difference from 14.

I could try to check through box scores I have to see which is closer to the truth, but I don't have Atlantic City boxes for '24 (they're very hard to get), so I can't guarantee how much that will tell us.

In #28 above, I meant that Lundy's next highest totals for a season, according to the Macmillan stats, are 7, 5, and 4 (three times).

A quick'n'dirty spot check of 1924 box scores revealed four home runs for Lundy in 20 ECL games. I'm not sure how much more information I can get in the near future, but I'll post here if I find anything else.

The case of Lundy really highlights the statistical uncertainties we face.

According to the data Holway presents, 1921 and 1928 look like Lundy's two best hitting seasons. Gary A.'s more complete data shows them to be very good years for Lundy, but not as good as they looked in Holway. Substituting Gary's numbers for Holway's dropped Lundy's career slugging percentage from .410 to .403. 1924 in Holway looks like a typical season for Lundy (he does appear to have been rather consistent in the early 1920s), but the MacMillan data (prepared in part by Holway) shows 1924 as a big slugging year. If I were to accept that data and regress it to the mean in my usual way, it would raise Lundy's career slugging back towards .410. If I accepted the more favorable data in each case, Lundy's slugging would move towards .420.

These are not _massive_ discrepancies, but for a candidate on the border as Lundy is, they may move him back and forth across that border.

To answer Jschmeagol's question about whether a career .403 slugger could put up the 1924 numbers:

If we convert Lundy's .618 slugging to an MLE for 1924, it comes out to .529 .

If we regress towards the mean by 25% in converting 60 games to 154, that drops it to .499.

I think that a peak slugging year of 90 points above career average is a deviation from the mean at peak prowess that fits with typical career patterns. It's a little higher than Sewell's highest deviation, about in line with Bancroft's, if I recall correctly.

If I were to accept the 1924 MacMillan data, it would lead me to raise my calculation of Lundy's career slugging, but not to change them radically.

Thanks guys. I currently have Lundy in the low 20s (above Sewell adn the rest of the SS glut) but after all the talk I have heard of him coming into this 'year' I figured he was be a HOM selection. I just wanted to make sure that this impression that I have of him is valid after all of the good talk you hear of him. I guess he is sorta like the anti-Beckwith in this regard. Well, he or Johnson.

And sorry Gary A. if I misrepresented what you said at any point. My bad.

I think Lundy is deserving of all the talk about him. He was an excellent player and leader, often the manager of his teams. In the context of the Hall of Fame's erratic standards, he has a very reasonable case for election.

Our standards, especially for the 1920s and 1930s, are markedly higher than the HoF's have been.

Judy Johnson is more clearly the anti-Beckwith.

The pitchers are usually the worst, but this year I feel like I've got a handle on Bill Foster--he's Stan Coveleski in blackface.

OK, but now we've got a real glut of SSs going, and what to do with Dick Lundy is issue number one. Initially, he looked Joe Sewell. Now, he looks more like Rabbit Maranville.

But frankly the numbers I've seen just don't add up. Here are the top currently available SSs ranked on four measures. The set includes Jennings, Sewell*, Bancroft, Rabbit, Tinker, Long, Lundy, Beckwith*, Moore, 9 in all. (*Not a full-time SS. I see Jennings as a full-time SS because his years NOT at SS are NOT prime much less peak years. Sewell had 1 prime year at 3B, Beckwith several.)

Career adjWS
1. Beckwith 319
2. Rabbit 302
3. Lundy 300
4. Sewell 277
5. Bancroft 269
6. Long 265
7. Tinker 258
8. Moore 256
9. Jennings 214

Peak (3 + 5 year for shorthand) adjWS
1. Moore 261
2. Jennings 247
3. Beckwith 242
4. Long 213
5. Sewell 209
6. Bancroft 199
7. Tinker 196
8. Lundy 189
9. Maranville 184

OPS+
1. Beckwith--I've seen different numbers but even the lowest one (130-something) is #1
2. Moore--haven't seen one but surely in the 130s, could possibly beat Beckwith
3. Lundy 122
4. Jennings 118
5. Sewell 109
6. Bancroft 98
7. Tinker 95
8. Long 93
9. Maranville 82

Career adjDWS (defensive WS)
1. Maranville 142
2. Lundy and Tinker 114
4. Bancroft 105
5. Long 103
6. Sewell 89
7. Moore 76--not sure where I saw this or if reliable, as I think it represents only 6.5 NeL seasons. If credited for army play, of course, all of his numbers go up except possibly OPS+
8. Jennings 66
9. Beckwith 52

So what's the problem. Well, Lundy is tied for #2 in defensive WS and trails only Beckwith and Moore on OPS+. Nobody beats him in both categories. Nobody is within 10 on both. If I open up my consideration set through 1964, only Appling is within 10 for both, and only Cronin and Appling are within 20.

And sure, Lundy ends up #3 in career WS. But unfortunately that puts him in Maranville territory, and his peak is also down at the bottom of the list with the Rabbit-man.

Well how does a guy accum. as many WS as Rabbit in 10 fewer seasons with an OPS+ edge of 122-82 end up with the same peak? Frankly, if you make the peak a dependent variable and look at CWS, DWS and OPS+, well, like I said he's Cronin or Appling.

And for that matter, how does a guy with a 122 OPS+ versus Sewell's 109, who earns 114 DWS to Sewell's 89, and does so in 14 years of 10 or more WS versus Sewell's 12, end up well behind for peak WS? Well, sure he spread them out over 2 more years.

But seriously do those OPS and DWS (Lundy's) sound like a guy who wasn't as good as Sewell? Not to me. But Lundy's numbers are speculative, so we don't quite know.

But I would just say this--don't rush to the judgement that Lundy was a black Maranville, when in fact there is reason to think of him as at least a black Sewell, and maybe a black Appling.

In any event, as of today, the SSs on my 1943 ballot look like this.

1. Jennings as always
2. Moore, the black Jennings
3. Beckwith
4. Sewell
5. Lundy--was #7 on my prelim, moves down but may very very well be better than Sewell. Must think. In any event, Lundy should stay in the picture and not be forgotten until we can better explain his numbers. All of the first 5 could be on my ballot. I would like Lundy to be there, though on the other hand he may be off-ballot with the perennially off-ballot but not forgotten Bill Monroe.

6. Bancroft--very possibly PHoMer someday when we get to the 3-fer years.
7. Maranville--very unlikely no matter how many we elect.
8. Tinker--still the best of the poem guys.
9. Long--slip sliding away.
10. (just for fun) Art Fletcher

1922 14.8 + 7.4 = 22.2

Is this correct for '22, Chris? The I9's have Lundy with less than 300 PA, while the numbers don't look all that special.

John,

That's the estimate I've made for 1922. Whether it's truly correct is another question.

1922 is an odd year for Lundy. As Gary A. explained above in post 10, the Bacharach's owner move the team from Atlantic City to New York that year, and Lundy was hired away by other Atlantic City folks to manage a rival team remaining in Atlantic City. They were a lesser team and don't appear to have played much against top teams. I infer that the i9s folks worked from data that showed Lundy playing only a few games for the NY Bacharach Giants that year, and so gave him credit for only a partial season.

Given that, as far as we know, Lundy actually _played_ that season, just for a team whose games were not as well documented, and that the batting average Riley reports for him is a typical average for him, I decided that there was no reason not to give him a full season of MLE credit for 1922, so I did.

Given that, as far as we know, Lundy actually _played_ that season, just for a team whose games were not as well documented, and that the batting average Riley reports for him is a typical average for him, I decided that there was no reason not to give him a full season of MLE credit for 1922, so I did.

I think you made the right call, Chris.

I like Lundy better than Sewell, but he won't make my ballot this time. There's no denying that he was an outstanding shortstop, though.

This isn't much--but, looking through 1922 Philadelphia papers, I found 3 games for Lundy's Atlantic City Bacharach Giants played in Philadelphia during June, all against white semipro teams. Lundy played shortstop and batted cleanup in all three. I only checked June, so they could well have played more games in Philly that year.

I will also work on finding out more about Lundy's 1924 season and the divergent home run totals.

Maybe I should have weighed in originally, but Lundy's hitting WS strike me as too low.

Let me first explain that I started with the i9 numbers and I recognize their limitations. The i9 numbers are .297/.355/.434 with 2,644 hits. To correct (imperfectly) I discounted the hits, doubles, triples, homers, walks, stolen bases and caught stealing by 10%. That gives .269/.323/.394 and 2,392 hits. That's actually lower than the .280/---/.403 with 2,441 hits in the MLEs above, but not wildly different considering we're speculating.

The advantage of using the i9 numbers is there is a separate number for doubles, triples, homers and walks, which allows a bit more precision in calculating his estimated WS. Assume for the sake of argument that i9 has those in the right proportions (though they may be wrong, we don't have anything better).

So I used those discounted numbers and applied the handy TangoTiger formula to convert Linear Weights to Win Shares for batting: Batting WS = ((LWTS Runs/10) + (Outs/100))*3. To calculate the linear weights, I used the National League average runs created for the applicable year, increased by 10% (so the league average is 10% higher than the NL). I do not apply park factors. Park factors are known to be somewhat suspect even in MLB when we have a ton of data points (although I use them in MLB player analyses). Using park factors in the NeL offers the opportunity for wildly disparate results because, frankly, we have no idea whether they are right for a multitude of reasons.

In sum, I assume the player would hit 10% worse in the NL than he did in the NeL and I assume he would be doing it in an NL that scored 10% more runs than the NL actually scored (in a neutral park).

If you do that, you get 224 batting WS for Lundy, as compared to 186 above. That's 38 more WS using lower numbers than the MLEs above.

My point is not to criticize the WS or MLE calculations above, or to posit that my method is the right one. However, there is obviously considerable latitude in the way we come up with these numbers. If Lundy got those extra 38 WS, he'd be at nearly 340 WS, which I think would drastically change the electorate's view of him.

Jeff, good question/good answer. Way back when Lundy became eligible I asked how a GG SS with a 122 OPS+ in 8700 AB could end up with 300 WS when...

Frankie Frisch, slightly less than GG 2B with 9100 AB and 110 OPS+ has 366.

Jeff,

Have you checked the linear weights-to WS conversion method on any major-league contemporaries of Lundy to check its accuracy?

Knowing that it has been cross-checked on actual win shares would increase my confidence in the estimates greatly!

I expect that my WS estimate for Lundy is a bit low and will rise upon revision, but I expect the rise to be more like 10%, rather than the 20% your analysis is proposing.

Sunnyday2,

Keep in mind that the 122 OPS+ estimate was KJOK's, the WS estimate was mine. My MLEs do not suggest that Lundy was a 122 OPS+ hitter; it estimates him at around 104.

My estimate will rise with the update of the translations, but the discrepancy is in the OPS+ estimates, not in the conversion of OPS+ and playing time into win shares.

Also, Frisch was an A+ defensive second baseman, a great defensive player, not a slightly less than GG defender. From 1923-35, he won 3 WS gold gloves, was second 3 times, and third 5 times. Still less valuable than an A/A+ shortstop, but comparable in value to a B+/A- shortstop.

Have you checked the linear weights-to WS conversion method on any major-league contemporaries of Lundy to check its accuracy?

An excellent question. I checked this for Lundy's era, and I wish there was an easy way to summarize the results. I'll just say the formula seems to work for a very narrow range of teams and seasons. The formula doesn't work well for player seasons if their teams have unusually high batting WS (formula underpredicts) or unusually low WS (formula overpredicts). The formula also loses integrity if the actual batting WS are single digit (formula overpredicts).

That reminded me to take a look at Bill James' short-form WS estimate at the beginning of the Win Shares book. For the same sample group I used short form hitting WS instead of the linear weights formula. It produced different results on a season-by-season basis, but in the aggregate, the total error was the same.

Then I noted the following on page 15 of Win Shares: "In history, 28% of all Win Shares estimated by this manner are exactly right, 37% are off by one, 16% off by two and 8% off by three. The other 11% are off by four or more Win Shares."

So 65% are within 1. However, 1 might be a huge number if the total is small (i.e., 1 of 8 is a 12.5% error and 1 of 25 is a 4% error). In addition, a 1 WS error over a near-20 year career like Lundy's and you've got a 5-10% variation. A 10% variation among the players we are evaluating can make the difference between a ballot spot, or not.

Back to the drawing board for me. :)

For kicks I found all the player seasons during Lundy's career which were within +/- .005 of his discounted career BA (.264 to .274) and within +/- .015 of his discounted career SLG (.379 to .409).

That produced 103 seasons, but I eliminated any seasons with less than 100 ABs, which took me down to 84 seasons.

I was hoping to smooth the errors of the short form WS formula (and TangoTiger formula) by using the average of comparable players.

Using short form hitting WS, those players averaged .029005 WS per out and .019880 WS per PA. Applying those to Lundy's outs and PA, I got hitting WS of 192 and 192, respectively, using the i9 numbers.

Using the TangoTiger formula in the same way, those players averaged .02633 per out and .01805 per PA. Applying those to Lundy, I got hitting WS of 175 and 174, respectively.

If I was to average those estimates, I would get 183, which is nearly identical to Chris' estimates.

Jeff,

Thanks (belatedly) for cross-checking these formulas against a larger body of data. It's unfortunate that they are not accurate enough for our needs in individual cases, but it's very helpful to know that in their application to a larger body of data they tend to confirm the accuracy of the player-comparison estimate method within 2-3 percent.

Revised MLEs for Dick Lundy

These new MLEs differ from those previously posted in 3 ways.

First, they use updated league offense level estimates, which have produced some small changes in Lundy’s translated stats.

Second, Lundy’s late-career plate appearances have been revised to fit a normal playing-time decline curve for middle-infielders. This has lowered his career totals in counting stats a bit and raised his career rate stats a bit.

Third, Lundy’s walks and obp have been systematically estimated. Walks and obp were not included under the old estimates: I just eyeballed them as part of the player’s record when choosing comps for the win share estimates.

The estimates for Lundy’s walks call for some discussion. I estimated Lundy’s walks using his walk rates for 1921 and 1928, derived from Gary A.’s data, which I turned into an MLE rate for those seasons by multiplying by the major-league walk rate and dividing by the NeL walk rate. I then used Tangotiger’s aging patterns data to extrapolate rates througout his career from those two data points. I have favored the extrapolation from 1928 over the extrapolation from 1921 (which is incredibly low) in my projections.

Even favoring the higher rate and adjusting Lundy’s rate to reflect the lower walk rates in the NeL, his walk rate is historically low for a player of his career length and batting ability. Lundy’s hit totals place him among the top 100 major-league players in hits (with no other NeL players counted, he would be at #70 with 2600 hits, one more than Steve Garvey). Among the post-1900 players in that group, Lundy’s BB/AB+BB is the fourth lowest, ahead of only Willie Davis, Bill Buckner, and Stuffy McInnis. It’s possible that the small base of data is not representative of his plate discipline overall, but I doubt that a larger base of data would raise him out of the bottom quintile of walk rates among the all-time leaders in hits. It’s possible, though, that he should be projected for 50-100 more walks for his career. (Adding 75 walks would raise his career OBP from .333 to .340, for example). There’s no data that supports this higher projection, but when a small slice of data produces an extreme outlying value, that value should be viewed with some skepticism. If there are any additional statistics on his walks from other NeL seasons or from Cuba, or if there is any anecdotal information on his plate discipline, it would be very helpful to have that.


Overall, these projections in BA an SLG will not look very different from the earlier set. Changes, if any, will show up in OPS+ and WS, since we have more exact OBP estimates and Lundy’s final five seasons will be more precisely calibrated to the National League.

As is now standard practice, Lundy is projected to a major-league norm through 1929 and to a National League norm thereafter.

Without further ado, the data.


Dick Lundy MLEs

Year Team    EqG  PA   BB  Hits TB   BA   OBP  SA
(1918 AC      90  360   9  105  137 .299 .316 .391)
1919 AC**    140  588  15  194  259 .338 .354 .452
1920 AC      135  567  16  181  263 .329 .348 .477
1921         127  533  17  163  237 .316 .337 .459
1922 NY/AC#  151  634  29  191  286 .315 .346 .472
1923 AC      154  647  32  172  274 .280 .316 .446
1924         152  638  31  194  295 .320 .353 .487
1925         154  647  35  163  219 .267 .306 .357
1926         149  626  32  184  228 .310 .345 .384
1927         154  647  35  174  200 .284 .324 .326
1928         152  638  35  186  237 .309 .346 .393
1929 Bal     108  454  26  123  151 .287 .328 .352
1930         147  617  35  173  211 .297 .336 .362
1931         138  552  31  157  178 .301 .341 .342
1932         128  512  29  145  151 .301 .340 .313
1933 Phi      94  394  24   87   94 .236 .283 .253
1934 Bkn      52  198  12   48   50 .258 .303 .269
1935 NY/NWK## 77  267  17   65   76 .258 .304 .303
(1937 NWK     25   75   5   14   19 .193 .244 .271)
Totals*     2212 9160 451 2600 3408 .299 .333 .391

*1918 and 1937 not included in career MLE totals.
**Also played for Hilldale
#Season split between New York Bacharach Giants (AC team moved to NY that year) and the rival AC Bacharach Giants, which formed to replace the team that had gone to New York.
## New York Cubans and Newark Dodgers

I look forward to DavidFoss’s OPS+ numbers and hope that I or Dr. Chaleeko will be able to produce bws from them soon.

Dick Lundy OPS+

Pefatory Comments

Well, Lundy’s OPS+ numbers indicate to me like there is some factor unaccounted for in the translation system that is leading to Lundy receiving too little MLE credit after 1923 and possibly too much MLE credit early in his career.

The most likely possibility is that the competition was getting better, though it’s possible that the offense-level adjustments for the mid-1920s are not high enough or that there are undocumented park effects at work. The projection of a steady offensive decline from age 21 onwards is otherwise rather hard to explain.

The impact of this problem shouldn’t be overstated, for the following reasons. Lundy’s lack of plate discipline hurts his OPS+ numbers more than I had expected when I was not treating his walks systematically. With a historically low walk rate, he gained in batting win share estimates that focused on batting average and slugging.

We would expect to see a player for whom speed is an important offensive tool to peak fairly early in this period. Sewell’s offensive peak came at age 23, Maranville’s at age 25.

Even with those caveats, Lundy’s trajectory as a hitter presented in the numbers below is surely incorrect: it remains to find how much correction is needed. Whatever effect is present in Lundy would also be affecting his contemporary Biz Mackey to some extent, though Mackey's pattern of offensive decline is not out of keeping with what we might expect to see in a catcher's career.

Having warned you against the numbers, I now present them:

Dick Lundy MLEs, with OPS+ added:

</pre>

Year Team EqG PA BB Hits TB BA OBP SA OBP+ SLG+ OPS+
(1918 AC 90 360 9 105 137 .299 .316 .391)
1919 AC** 140 588 15 194 259 .338 .354 .452 106.9 125.6 132
1920 AC 135 567 16 181 263 .329 .348 .477 101.4 124.2 126
1921 127 533 17 163 237 .316 .337 .459 94.4 110.3 105
1922 NY/AC# 151 634 29 191 286 .315 .346 .472 96.4 113.7 110
1923 AC 154 647 32 172 274 .280 .316 .446 88.8 110.1 99
1924 152 638 31 194 295 .320 .353 .487 99.1 120.0 119
1925 154 647 35 163 219 .267 .306 .357 84.1 84.0 68
1926 149 626 32 184 228 .310 .345 .384 97.2 95.5 93
1927 154 647 35 174 200 .284 .324 .326 91.3 80.3 72
1928 152 638 35 186 237 .309 .346 .393 97.5 95.4 93
1929 Bal 108 454 26 123 151 .287 .328 .352 90.4 81.5 72
1930 147 617 35 173 211 .297 .336 .362 90.8 78.0 69
1931 138 552 31 157 178 .301 .341 .342 99.1 84.8 84
1932 128 512 29 145 151 .301 .340 .313 100.8 76.0 77
1933 Phi 94 394 24 87 94 .236 .283 .253 86.5 67.3 54
1934 Bkn 52 198 12 48 50 .258 .303 .269 88.6 65.9 55
1935 NY/NWK## 77 267 17 65 76 .258 .304 .303 89.1 74.4 64
(1937 NWK 25 75 5 14 19 .193 .244 .271)
Totals* 2212 9160 451 2600 3408 .299 .333 .391 92
1933 tot. 2110 8695 422 2487 3282 .301 .335 .397 94

</pre>

*1918 and 1937 not included in career MLE totals.
**Also played for Hilldale
#Season split between New York Bacharach Giants (AC team moved to NY that year) and the rival AC Bacharach Giants, which formed to replace the team that had gone to New York.
## New York Cubans and Newark Dodgers

Argh, I hit submit instead of preview and missed the key typo in the pre- tag. Here's the table again, properly formatted.

Dick Lundy MLEs, with OPS+ added:

Year Team    EqG  PA   BB  Hits TB   BA   OBP  SA  OBP+ SLG+  OPS+
(1918 AC      90  360   9  105  137 .299 .316 .391)
1919 AC**    140  588  15  194  259 .338 .354 .452 106.9 125.6 132
1920 AC      135  567  16  181  263 .329 .348 .477 101.4 124.2 126
1921         127  533  17  163  237 .316 .337 .459  94.4 110.3 105
1922 NY/AC#  151  634  29  191  286 .315 .346 .472  96.4 113.7 110
1923 AC      154  647  32  172  274 .280 .316 .446  88.8 110.1  99
1924         152  638  31  194  295 .320 .353 .487  99.1 120.0 119
1925         154  647  35  163  219 .267 .306 .357  84.1  84.0  68
1926         149  626  32  184  228 .310 .345 .384  97.2  95.5  93
1927         154  647  35  174  200 .284 .324 .326  91.3  80.3  72
1928         152  638  35  186  237 .309 .346 .393  97.5  95.4  93 
1929 Bal     108  454  26  123  151 .287 .328 .352  90.4  81.5  72
1930         147  617  35  173  211 .297 .336 .362  90.8  78.0  69
1931         138  552  31  157  178 .301 .341 .342  99.1  84.8  84
1932         128  512  29  145  151 .301 .340 .313 100.8  76.0  77
1933 Phi      94  394  24   87   94 .236 .283 .253  86.5  67.3  54
1934 Bkn      52  198  12   48   50 .258 .303 .269  88.6  65.9  55
1935 NY/NWK## 77  267  17   65   76 .258 .304 .303  89.1  74.4  64
(1937 NWK     25   75   5   14   19 .193 .244 .271)
Totals*     2212 9160 451 2600 3408 .299 .333 .391              92
1933 tot.   2110 8695 422 2487 3282 .301 .335 .397              94

I think there is something wrong with our statistics, our analytic methods, or our philosophical approach to Negro League players. Our numbers for the early 1920s seems to correllate with players' reputation and our numbers for the late 1930s and early 1940s don't seem that far off, but from 1925 to 1935 or so we are projecting guys like Lundy, Bell, and Mackey to be little better than replacement level major leaguers. I don't buy that for a second.

I have no real sense of what is causing this problem, but among the hypotheses I'd like to test are the following:

(1) The Negro Leagues were MUCH better in quality in the late 1920s and 1930s than we are giving them credit for, perhaps approaching parity with the major leagues.

(2) We are erring in assuming that the low Negro League walk rates woild translate into low major league walk rates b/c/ those walk rates reflect a group decision to play a game with as few walks as possible rather than anything about the players' abilities. The only Negro Leaguers who walk at all are the guys who teams pitch around. Everybody else swings at everything. If these guys played in the major leagues where walks were acceptable, some of them would have high walk rates, some of them average walk rates, and some of them low walk rates--it is wrong to translate them all with low walk rates.

(3) Our park factors are out of whack by huge margins.

Quick responses to these reasonable hypotheses:

1) The history of league formation (comparing NeL with the formative years of the white majors in the 1870s and 1880s) is consistent with the idea that league quality would rise significantly in the years immediately after league organization.

The history of the Negro Leagues' economic struggles and the competition from the Mexican League is consistent with the idea that the Negro Leagues declined in quality somewhat in the late 1930s and never rose to their former level.

Since the translation rates are based on 1940s data, this later drop in quality would lead to underrating the period of top competition in the translations.

Gadfly has argued for a shape somewhat like this for a long time: we are beginning to have enough data to see some evidence of it in the statistics.

2) The difference between NeL walk rates and ML walk rates is not so large as to suggest to me a consistent difference in style of play. Also, walk rates among players who _would_ be pitched around are highly variable. Mule Suttles and John Beckwith didn't walk much, Jud Wilson and Oscar Charleston did. George Scales appears to have walked a lot. He was an above-average hitter, better than Lundy and Mackey, but not a terror to the league. So I am doubtful that the walks in the NeL were any more concentrated into a handul of players being pitched around than they were in the majors.

Gary A.'s seasonal data for 1921 and 1928 might offer convincing support or refutation of this hypothesis.

3) Park factors out of whack are possible, but insofar as we are seeing consistent patterns emerge for players on different teams, I think a problem with the competition adjustment is the more likely explanation.

We do however have other NEL competitors of Lundy, Bell and Mackey, such as Stearnes, Suttles and Beckwith, who have projected nice and HOM'able ML careers. If Lundy, Bell and Mackey are too low, then the other guys must have been the second comings of Babe Ruth -- the differential between the two groups of players is surely about right, except for random fluctuations.

Is it not possible that we are OVERestimating in the early 1920s and early 1940s, rather than underestimating 1925-35?

In any case, Lundy, Bell and Mackey are hardly replacement level. Lundy was an SS and Mackey a catcher; Rabbit Maranville is in the HOF with an OPS+ of 82 -- back then there was little expectation that those two positions would hit much, and surprise when they did. Bell was lightening fast, and got huge points for that from his contemporaries, like Maury Wills, Lou Brock, and other speedsters who weren't great hitters. The sabermetric low average, high walks hero was less valued then than today -- Mark Bellhorn would have been out of an ML job in 1935!

Some observations on Lundy's walks.

1) Gary's numbers on 1921 and 1928 show Lundy with 3 walks in 132 ABs and 10 in 206 ABs respectively. This is surely not a completely representative sample, but Gary's research shows that with one walk for every 18 at bats, the base on balls was simply not part of King Richard's game in these two seasons.

2) Contrasting that, KJOK's MLEs (which he admits to being less sophisticated) show Lundy drawing a walk every 10.2 ABs or so for his career.

3) Chris's MLEs show Lundy drawing a walk every 20.6 ABs, or about twice as often as KJOK suggests, and about 13% less often than the stats we have from Gary say he walked.

4) I9s has Lundy walking once every 11.27 ABs.

5) The Cuban records posted above do not contain walk information.

IMO it's equally possible that:
-Chris is coming in too low
-KJOK is coming in too high
-I9s is coming in too high
-Lundy's 1921 and 1928 walk rates were out of character compared to his other seasons.

Without further raw data from game accounts, we have to decide which treatment of Lundy's walks is most apt
-a treatment that is consistent with the two seasons we have data for
-a treatment that gives him the benefit of the doubt.

I think the first of these is the most appropriate until the group (or more likely Gary and Gadfly) uncover further evidence.

On the other hand, maybe there's some kind of way around it? Is there any way for us to estimate his plate appearances based on a combination of his teams' runs scored and the number of known outs, as well as his known totals of Games and ABs? (Or does one of us do that already?)

Is it not possible that we are OVERestimating in the early 1920s and early 1940s, rather than underestimating 1925-35?

Well, I think it is unlikely that we are overestimating the period from 1944-48, since the conversions there are based on analysis of players changing leagues.

It is possible that we are overestimating some part of 1918-1923 and 1940-42. I think it probable that there is some overestimation for those seasons and some understimation 1925-1935.

In any case, Lundy, Bell and Mackey are hardly replacement level. Lundy was an SS and Mackey a catcher; Rabbit Maranville is in the HOF with an OPS+ of 82 -- back then there was little expectation that those two positions would hit much, and surprise when they did.

True, though there are seasons within their careers where they are approaching replacement level.

If Lundy, Bell and Mackey are too low, then the other guys must have been the second comings of Babe Ruth.

It depends on how large the errors are. Here's an example, chosen for illustrative purposes, not because I believe that this is the correct conversion factor, but because it's an alternative that has been suggested, and it would be a rather large change.

A quick look at Lundy's 1926 season -- 93 OPS+ in the current system -- indicates that changing the 90/82 conversion rate for batting and slugging to a 95/90 conversion rate would move Lundy up to a 106 OPS+: a significant but not earth-shattering change. The percentage change would stretch at the higher end, of course, so John Beckwith in the same season with the same change would go from a 149 OPS+ to about a 167 OPS+. This would move him from #9 to #2 in the majors in OPS+ that year. The top 10 are 227 (Ruth), 158 (Williams), 154 (Manush), 153 (Gehrig), 153 (Heilmann), 153 (Goslin), 152 (Hargrave), 151 (Wilson), 147 (Waner), 142 (Simmons). That is a significant increase, but not enough to set him apart from the cadre of top hitters with which he is already grouped, not enough to turn him into the second coming of Babe Ruth, or even Rogers Hornsby.

3) Chris's MLEs show Lundy drawing a walk every 20.6 ABs, or about twice as often as KJOK suggests, and about 13% less often than the stats we have from Gary say he walked.

This is an interesting outcome, which might be explained in two ways.

1) It accurately reflects that a projection of Lundy's full career, which began at age 21, would have a walk rate somewhat lower than a walk rate weighted towards his age 30.

2) The formula I am using to calculate projected walks is too low.

A matter calling for study!

Everyone,

I should have said that Chris's MLEs show Lundy walking half as often as KJOK suggests, not twice as often. Sorry for any confusion.

Well, one reason for the low career totals is that NeL players had a better opportunity and more incentive to have 25 year careers. The replacement level of the NeL was a good bit below the replacement level of MLB AND the salaries werent' good enough/ there were less opportunities after baseball for these players so they tended to stick around in baseball. I doubt that the careers or Mackey or Lundy would have been as long as they were if they had been playing in MLB. However, they would also have had fewer seasons approaching replacement level and better career rate stats.

I'm not talking about end-of-career seasons. Those I have already removed from Lundy's and Mackey's records.

Lundy's record shows the following age/season/OPS+ numbers

27/1925/68
29/1927/72
31/1929/72
32/1930/69

Numbers like these, which are approaching replacement level, are not the product of a 25-year career.

Maranville, from 31-35 in 1923-27, not quite his prime but a decade before his retirement, went 76-86-53-69-50 on OPS+. We're not talking A-Rod here! It was a very different game, particularly for hitters like Lundy and Maranville whose value was concentrated in BA, which was the stat people tracked.

Here's a list of the OPS+ for the 16 starting shortstops in 1925 as listed in baseball-reference.

119
109
102
92
91
87
85
75
71
(68-Lundy)
60
60 [replaced]
56 [replaced]
55 [one year away from being replaced]
53 [replaced]
50 [replaced]
43 [replaced]

So Lundy is just below the median, but this is clearly a very weak pool of shortstops in that 5 of the 7 below Lundy in OPS+ were replaced as starters, either during the season or after the season. One of these is Maranville, who would bounce back later, but all the rest were essentially done in the majors.

So at this point in time, Lundy as projected was just a little below average as an ML-shortstop, but he was also just a little bit above replacement.

A sidenote:

hitters like Lundy and Maranville whose value was concentrated in BA, which was the stat people tracked.

Maranville was actually pretty good at taking a walk. His batting averages were what kept his OBP down. He was surely being employed for his flat-out defensive brilliance, though perhaps his wacky lovable nature also had something to do with it . . .

I should have said that Chris's MLEs show Lundy walking half as often as KJOK suggests, not twice as often. Sorry for any confusion.

1. At the time, I had NO walk data for Lundy, so was relying 100% on I9 walk data.

2. The I9 walk rates seem to be typically "league average" types of rates, without too much variance between players (although I haven't systematically looked at this).

3. Without having walk data for every Lundy season, any estimates of walk rates using actual data should probably be heavily regressed towards the mean.

4. Conclusion - I think my rates are too high, but maybe Chris' rates are too low....

2) Contrasting that, KJOK's MLEs (which he admits to being less sophisticated) show Lundy drawing a walk every 10.2 ABs or so for his career.

3) Chris's MLEs show Lundy drawing a walk every 20.6 ABs, or about twice as often as KJOK suggests, and about 13% less often than the stats we have from Gary say he walked.

4) I9s has Lundy walking once every 11.27 ABs.

Also, just a reminder again that the difference between my 10.2 BB/AB and I9 11.27/AB is simply because my MLE's are projected into a higher walk environment (1955-2003 National League).

I have 4 1932 Boxscores for Lundy. Looks like early in the year he batted 3rd, and then later in the year he was batting 6th.

The limited data:
AB-16
H-5
2B-1
SB-1
BB-0

A very interesting thread. Although I was quite aware that Judy Johnson walked very little and was over-rated because of this (among other things), I was not aware that Dick Lundy possibly walked very little. I will look and see if I can find any additional evidence on the subject.

In general, when considering Negro League walk rates, I usually look to see if a player's ABs per game are very high or low and then consider whether team offense or batting position can explain any differences (Cool Papa Bell is a good example, he had tons of ABs but walked at a respectable clip anyways).

Some players (Gibson later in his career, Buck Leonard always, Spot Poles) quickly show up using this method as players who could really draw a walk. Other players (like Johnson) quickly show up as players who would not draw a walk [Every once in a while there is some real evidence also, such as the Mexican League statistics or the Pittsburgh Crawfords published stats in 1933].

One of the most interesting things that I've learned in the Hall of Merit was to what extent Jud Wilson was an 0BP monster. I knew that he walked and got hit by pitches a lot, but, Jesus Christ Almighty, not to that extent.

But most players just fall into the middle. Of course, the fact that the Negro Leagues were a low walk enviroment helps hide the walk rate in any event.

As far as how this impacts on Negro League seasonal evaluation, I'm not sure that it does. I think that Chris Cobb and company have a pretty good handle on that, outside of our disagreement on the actual conversion rates.

I think it only really impacts individual player evaluation. Modern sabermetrics had the same initial discoveries.

For instance, Pie Traynor (who is Judy Johnson's almost dead-on comp) was considered for years to be the best third baseman of all time. But now, his modern reputation is toast, mostly because he really didn't walk much.

We are basically doing a sabermetric analysis and should not be surprised that some players are now being upgraded because this analysis favors the modern OBP over BA.

The Negro Leagues, just like the Major Leagues of that time, had a Batting Average bias.

To end by beating my favorite drum, I notice that some people are picking up on the fact that the conversion rates being used pretty much uniformly do not match the reputations of the Negro League players themselves.

Reputations that come from being compared, by the era's own players and officials, with their contemporary Major League peers.

There are, of course, two possible answers to this cognitive dissonance.

One, proposed by Karl Magnus, is that all those sucky Negro Leaguers are over-rated anyways.

Two, which some people seem to be picking up on, is that the conversion rates are simply flat wrong.

As far as Dick Lundy goes, I have had him listed in the lower regions of my ballot for a while. I did not even realize that his Win Share data had been posted. I, off the cuff, had been comparing him to Frankie Frisch, who I believed to be a very good comp.

By my own personal adjustments to this WS data, I would estimate that he would have had about 362 career Win Shares from 1918 to 1934 with peak seasons in the low 30s. With the straggling years after 1935 included in, Lundy would have possibly gotten up to almost 400.

Weirdly enough, by repuation and my off the cuff analysis, this is exactly where I had evaluated him.

But, if Lundy really walked this little, a more appropriate comp may be Pie Traynor good enough to play shortstop rather then a slower stronger Frankie Frisch with a strong enough arm to play shortstop and Lundy is not as good as I thought.

Fascinating.

Hi, I wanted to bump this thread because I have a couple of items to discuss/announce here:

I have talked to GJM about becoming a voter. I was planning on voting in 1988 but frankly, I think it would be a disservice for me to vote without doing A LOT more work. Part of that work is doing some back-reading of the stuff you guys have already done.

I must say, I had always considered myself to be fairly knowledgeable about Negro Leaguers. Then I read these archives and I am simply amazed at the work done here.

However, it appears that a lot of these links have not been updtaed for more than a year. As the Hall of Fame has inducted a slew of Negro Leaguers this year, it has done a lot of research and seemingly has uncovered career statistics for several candidates.

The HOF did NOT induct Dick Lundy. One of the things I find most impressive about the work done by the HOM is that you guys have actually discerned differences between reputation and statistics and, most likely, have uncovered mistakes made by the HOF in electing Negro Leaguers.

Anyway, the hold up on Lundy seems to be the exact quality of his offense, in particular how much he walked. According to the BBHOF Web site, Lundy posted a career .306 average with a .363 OBP. This suggests a player who would walk 50-60 times a year, hardly the notorious nonwalker gleaned from fragmentary information earlier. It also says Lundy was 13th all-time in RBIs. If you were to look at No. 10-15 in RBIs among black/Latin players from 1950-1999, you have guys like Jim Rice, Dave Parker, Billy Williams. I'm not suggesting Lundy hit like those guys, but it is worth mentioning, if only for the fact that he must have bad enough EBH pop to bat in the middle of the order for most of his career.

To me this indicates that Lundy was, at worst, a Dave Bancroft-type clone, but very possibly something considerably better -- maybe a Billy Herman/Frankie Frisch type. Given this information from the HOF profiles, is it worth revisiting Lundy and some other players (John Donaldson, for instance, is credited with 4,000 strikeouts and a record of 262-94. I don't know what to make of that, but everyone from his era with 2,000 strikeouts is a HOFer)?

The most recent MLEs for Lundy are .299/.333/.391/92 with 300 WS.

I guess I'm as much of a moneyball guy as anybody, but to focus in on a record like that a bash a guy because he didn't walk does seem a little bit misguided. If we said he was a .275/.333 guy...would that make him better? I mean, .275 for a Gold Glove SS would be a pretty acceptable number and now we're talkin' 35 walks per 600 PAs. Still not a lot but more than 20. Does this make him a better player?

That is exactly half of Sewell's 70 BB per 600 but Lundy was a much better defender and earned more WS, granting that they are MLEs. Is 35 walks really that important?

Bancroft v. Lundy comes down to 98 OPS+ v 92 OPS+. Bancroft wins.

Er, the MLEs for Lundy have him walking at historically low levels. The HOF numbers suggest he walked more than that.

Er, the MLEs for Lundy have him walking at historically low levels. The HOF numbers suggest he walked more than that.

For whatever reason, it appears that the NeLers didn't have the same type of plate discipline like the MLers did. Whether or not the MLEs exaggerate this difference is certainly debatable.

OK, let me recap the Lundy conversation as I understand it:

1. Lundy's available stats are sketched together from the standard sources, Riley and Holway, as well as a couple of seasons of data from Gary A.

2. Preliminary MLEs are made, with similarities yet significant discrepancies. Debate ensues as to the validity of a .442 SLG or something closer to .403. Preliminary OBPs at this time are in the .340ish area. Projected OPS+ numbers range from 104 to 122.

3. Using available walk data from the Gary A. seasons, Lundy is projected to be a fairly extreme non-walker. In fact, a historical non-walker. Chris Cobb notes in post #50:

It’s possible that the small base of data is not representative of his plate discipline overall, but I doubt that a larger base of data would raise him out of the bottom quintile of walk rates among the all-time leaders in hits. It’s possible, though, that he should be projected for 50-100 more walks for his career. (Adding 75 walks would raise his career OBP from .333 to .340, for example). There’s no data that supports this higher projection, but when a small slice of data produces an extreme outlying value, that value should be viewed with some skepticism. If there are any additional statistics on his walks from other NeL seasons or from Cuba, or if there is any anecdotal information on his plate discipline, it would be very helpful to have that.

4. Chaleeko notes in post #56:

1) Gary's numbers on 1921 and 1928 show Lundy with 3 walks in 132 ABs and 10 in 206 ABs respectively. This is surely not a completely representative sample, but Gary's research shows that with one walk for every 18 at bats, the base on balls was simply not part of King Richard's game in these two seasons.

2) Contrasting that, KJOK's MLEs (which he admits to being less sophisticated) show Lundy drawing a walk every 10.2 ABs or so for his career.

3) Chris's MLEs show Lundy drawing a walk every 20.6 ABs, or about twice as often as KJOK suggests, and about 13% less often than the stats we have from Gary say he walked.

4) I9s has Lundy walking once every 11.27 ABs.

5) The Cuban records posted above do not contain walk information.

So it is acknowledged that Lundy's walk rates cannot be stated with extremely high confidence; Chris Cobb in fact requests any other statistical or anecdotal evidence related to Lundy's walks or plate discipline.

5. No further help comes, and thus the most recent work shows Lundy to be a below average MLB hitter, with a 93-94 OPS+.

6. Fast forward to 2006, Cooperstown does some housekeeping and clears out the backlog of Negro League candidates. In doing so, it apparently conducted more exhaustive statistical research than was previously available, and finds Lundy to have a career .306 average with a .363 OBP, and notes that Lundy was No. 13 all-time in RBIs.

The most recent MLEs for Lundy show a player who would walk around 25-35 times in a full season. A .306 hitter with a .363 OBP would presumably walk around 50-60 times a year. Someone who is among a league's all-time RBI leaders is, presumably, someone who amasses a fair number of total bases and/or slugging percentage.

I'll point out again that, at some point in the conversation, Lundy was seen to be a fairly good hitter. Decisions made on the best available information at the time have projected him to be significantly worse than originally thought, although there was always some debate.

So, what I want to know is if this information provided by the Hall of Fame more accurate and does it impact the MLEs we have for Lundy?

These are good questions, and I wish that I could answer them satisfactorily with respect to the MLEs that I have calculated. Unfortunately, I have a lot less time to work on matters HoM-related than I did two years and more ago wen I calculated most of the MLEs. I haven't even looked at the HoF's published batting data.

It's my sense that Gary A.'s data is the most reliable that we have available, since he has explained his methodology and has been able to discuss issues of completeness/incompleteness in the data he has provided. I believe that the HoF's data should be pretty reliable, however, and, as it covers many more seasons than Gary A.'s does, it should provide a better view of Lundy's hitting, as long as we can trust it. It must surely be better than Holway's data . . .

If I were to start assessig the HoF data, I would begin by comparing their findings for 1921 and 1928 with Gary A.'s and checking to see whether the season-by-season data presents a view of developing/declining plate discipline for Lundy that fits our usual expectations about how a player's walk rates change over a career. If the HoF's numbers are consistent for those seasons with Gary A.'s, and chart a comprhensible career path for Lundy as a hitter, then I would say that MLEs for Lundy ought to be recalculated based on the new data, and we should reevaluate his standing based on the results.

So, how does the HoF view of Lundy's 1921 and 1928 seasons compare to Gary A.'s? If someone could post the data here, or link to it, that would help us move forward with deciding whether and how to re-open Lundy's case.

Here's the link to the profiles at the HOF site:

http://www.baseballhalloffame.org/hofers_and_honorees/negro_leaguers.htm

There are no line-by-line stats,simply statements like "Oms had a lifetime OPS of 883," that are quite tantalizing.

The HOF has not, as far as I know, released the data that was used to evaluate the new NeL class.

-- MWE

For NeL HOF finalists, the HOF posted PDFs with line by line stats for hitters and pitchers (together with "league" batting averages and slugging percentages).

In 1920, 1921, and 1923, as a young player, the HOF study shows Lundy walking 16 times in 93 games. From 1924 through 1929, his walks vary from about 10 to about 15 percent of his at bats; in the 1930s they tail off, largely (I would guess) due to lack of information in box scores.

Thanks, Gary A., for the links to the pdf files! This is the first time I actually looked at the batting files, and I was quite surprised to see that they include _league_ batting average and slugging average. _That_ will make for more reliable MLEs, as the ones I have done hitherto had relied on estimates of league BA derived from Holway's data, and slugging estimates derived from those, except for the seasons for which Gary A. was able to provide league data. I am not going to be able to work on updating my MLEs until the semester is over, but I am going to have to find the time to dig into them again now that we have league data for every season!!! Now if only we had full leaugue batting data, pitching data, and park factors . . . Surely the HOF will publish that encyclopedia sometime soon ???

baudib,

Thanks for reopening the discussion on Lundy! I hadn't done any MLEing for Lundy since Chris and KJ worked him up previously, but I might try a base-line MLE later during the week. (Don't anyone hold up a vote or hold their breath in excited anticipation...). The spring catalogue season at my company begins in earnest this week so I don't know what time I'll have, but I'll at least try to run the basic mechanical stuff on the HOF data and report back, catch as cat can.

Last I heard, they did not have a publisher for the encyclopedia (astoundingly)...however, word has it that the data might be posted on the SABR website within the next few months. I haven't seen that confirmed anywhere, though.

FYI: the HOF data for 1928 is actually based on my research, though there are now differences that result from 1) further research on my part and 2) changes and edits they made to my work. Unfortunately, I have no idea what they did to my work--they did give Lundy significantly more walks (16 in 53 games, while I have him with 10 in 58).

This is the latest statistical line I have for Lundy in 1928, with HOF study for that season second:
G: 58 (team 62) / 53
AB: 217 / 195
H: 73 / 68
D: 13 / 12
T: 3 / 3
HR: 4 / 3
R: 36 / 34
W: 10 / 16
HP: 2 / --
SF: 2 / --
SH: 1 / 1
SB: 8 / 8
AVE: .336 / .349
OBA: .368
SLG: .479 / .487

Gary, so which is correct, 10 walks or 16? It seems to be a significant question.

PS. The alt OBA is .398, not the mark of a weak-ass no-walk bum. Of course a .349/.398/.487 hardly sounds like a 94 OPS+ either.

PS. The alt OBA is .398, not the mark of a weak-ass no-walk bum. Of course a .349/.398/.487 hardly sounds like a 94 OPS+ either.

If the HOF's data are correct, we might be looking at something like a 120 OPS+ for Lundy for 1928, which appears to be probably one of Lundy's top three seasons on offense (probably his third best, at a guess). That would be up from about a 100 OPS+ according to the most recent MLEs that I did.

Mind you, that 120 is a) a very rough estimate and b) unregressed, and given that Lundy's slugging goes from being 28% above league average in 1928 to 2% below league average in 1929, I am certain that there would be significant regression. So, although the new data show Lundy's offense more favorably, I am doubtful that he will end up looking as good offensively as Joe Sewell. I think it likely that he will look as good or better than Dave Bancroft, however, which could bring him back onto at least some ballots, given that Lundy's defensive reputation is stellar.

I will say, though, that the discrepancies between the HoM data and Gary A.'s are troubling: I hope we will eventually get a fuller accounting of their methodology.

I will say, though, that the discrepancies between the _HoF_ data and Gary A.'s are troubling.

I'm more accustomed to typing HoM than HoF for some reason . . .

OK, I just went ahead and cranked through the numbers. This is sumamry level data only, based strictly on three sources:
1) the line-by-lines and league averages presented by the HOF study
2) the team game totals for majority of seasons from the Clark/Lester book
3) the team game totals for a few seasons from the Holway book.

Method was down-n-dirty:
1) compare AVG and SLG to HOF study's lg avgerages.
2) discount average by 10% for QoP and SLG by 19% (aka: 90/81 per usual protocol)
3) in absence of NgL walk rates, for walks, assume his bb/ab remains constant from NgL to MLB
4) adjust into NL seasons in all instances
5) for playing time, take ratio of G to Team G, multiply by 154. For a few seasons I had to adjust the Team Gs upward since the new study sometimes had Lundy in more games than earlier studies had for his team's games. Those seasons end up as 154 game years for him. And in 1934, he was putrid, and rather than give him a near full season (as the record appears to warrant), I gave him about 40 games.

Worth noting: I have not made any attempt to add in a 1922 season, though it should be done. See above for details on why info is sketchy on 1922. And no park factors either.

OK, so what did I get? Well, pretty much what's been got before.


There's a couple things to mention here.
1) His SLG really gets killed in conversion. Given his reputation as a power hitter and his high career RBI totals, there must be a park effect in play. Or else the power is not as real as everyone thought. His league SLG was .375, his SLG was .416 per the big study. That's a 11% above the league. He ends up 10% below. Makes sense given the conversion factor.

2) His walks are OK. His OBP turns out 1% better than the league by this estimation. However, short of disavowing the accuracy of my walks translation, I'd like to talk for a second about the walks. There's several reasons to be concerned:
a) no NgL walk rate info
b) no good way to estimate reliably
c) i wasn't sure how walks translate from non-MLB to MLB; that is we use .90/.81 for AVG/SLG, so what do we use for walks? I didn't know, so I used 1.0.

All along, I've had hte hunch that NgL walk rates were probably around 2% lower than MLB walk rates. Why? Well, a couple things. We all know that the NgLs were well reputed for the slash-n-run style of play and for inside baseball. We've also often speculated that errors were probably more frequent in the NgLs. Some of this is probably born out by the fact that the NgLs consistently batted around a dozen points lower as a league than the NL. With more errors and more productive-outs and maybe also more attempts at bunt hits, you'd figure averages would be lower. Not to mention that NgL pitchers were allowed to load up the ball. Finally, I have to wonder if the quality of umpiring was all that good; given a rougher style of play (Jud Wilson or John Beckwith knocking out umps) and the leagues' shoestring budgets, I'd have to guess that umping wasn't superb or at least not consistent. When umping isn't great, batting averages and walks go down because batters don't have a consistent sense of the strike zone and have to swing at whatever's close.

So given all this, I've figured that it probably means the NgLs were at 6-7 walks per 100 ABs when the NL was at 8-9. Now, if I followed this hunch, and estimated the NgLs bb/ab at -.02 of the NL's what would happen to Lundy's walks and OBP and OPS+? Well, he'd pick up 169 walks (raising him to 678), his OBP would rise to .359, and his OPS+ would rise to 96. Looking seasonly, his top seven OPS+ by each method:

original 118 114 112 112 110 106 104
alt bb 124 117 117 116 116 111 110

Sadly for King Richard, that's where the good news ends. In both methods his 100 OPS+ seasons dry up from there.

Which is another thing to note. He really went off the cliff. From 1931-1935 he either totally lost it (OPS never reaching 90 by first method, reaching 90 only once by alt-bb method), or he encountered an extremely difficult hitters park after moving to Balto, while also beginning to decline. Anyone got a handle on this aspect of his career?

Anyway, so adding an estimated 1922 might help, though his surrounding years would be the source of the conjecture, and the OPS+s are roughly 100 by the first method and 105 by the second. Not bad, obviously, but essentially bulk filler, not a peak-level difference maker.

One thing I just noticed. If the method-one version of Lundy is correct in my previous post, he's not appreciably different than Silvio Garcia. I'd say the difference would be that 104 OPS+ season you see on the tail end of his good OPS+ years. In our current backlog bonanza, that's enough differentiation in all probability to be on a couple ballots versus being an afterthought, but I found the comparison interesting in many ways and thought it was worth sharing.

I should also note that I did not use any regression, nor did I include my usual assortment of winter league data to offset the small NgL sample sizes. I think that's all the disclaiming I can think of! ; )

A couple of quick clarifications:

1. I would say Gary A.'s data is defintely the most reliable, as it's balanced for all games included, between pitchers and batters, etc. (ie if batters hit 570 doubles in total, pitchers allowed doubles will total 570...)

2. My MLE's were based on I9 website data, which I understood to be based on data gathered for what became the HOF Negro League Data.

3. Any HOF Negro League data published for Lundy in an encyclopedia, or anywhere else, will likely be the exact data contained in the .pdf files, so in effect we already have as much data on Lundy as we're likely to have for awhile (until Gary A produces even more seasons...)

Well, having the stats is awesome, although it doesn't look good for Lundy.

There must be other players who must be reassessed -- frankly, Oms' numbers don't look particularly stunning at first glance.

It isn't much, but I just posted 1924 ECL park factors, if that helps at all.

You know, I was always under the impression that i9s was just based on published sources at that time (same old Riley, Holway, etc., maybe up to Figueredo's Cuban books), rather than on unpublished NLRAG (HOF) stuff.

I just remembered I have these stats for Lundy with Almendares in the 1927/28 Cuban League.

G: 29 (team 32)
AB: 124
H: 43
D: 9
T: 2
HR: 2
R: 26
RBI: 21
W: 5
HP: 1
SF: 3
SH: 0
SB: 1
AVE: .347 (lg .301)*
OBA: .368 (lg .361)*
SLG: .500 (lg .413)*
*-league averages with pitchers excluded
league ERA: 4.11
league TRA: 5.56
league FPCT: .950

FIELDING: SS (Lundy also played a single inning at first base.)
G: 29
DI: 260.7
PO: 62
A: 108
E: 14
DP: 16
RF: 5.87 (lg 5.39)
FPCT: .924 (lg .917)

Keep in mind this was a three-team league. The other shortstops were Francisco Correa for Cuba and José Olivares and Manuel Cueto (a veteran and former major leaguer who was usually a third baseman) for the champion Habana team.

Playing first base for the same team, Lloyd hit .350/.387/.476. Marcelle at third hit .339/.395/.426, and Scales at second hit .272/.336/.368. The best hitters in the league were Wilson (.432/.540/.739) and Dihigo (.381/.448/.571) of Habana, and Charleston (.351/.472/.570) of Cuba.

There must be other players who must be reassessed -- frankly, Oms' numbers don't look particularly stunning at first glance.

Certainly, we should look again at everyone's numbers. Without having looked at Oms' numbers from the HoF, let me note that his numbers in Cuba play an important role in his case. The CWL data lacks bases on balls, which is too bad, but they are generally more complete and reliable than the NeL numbers, and they will not have changed as a result of the HoF project.

If I recall correctly, the Cuban Stars were a road team in the 1920s, so they would lack the home-field advantage that usually boosts 1/2 of a player's stats. Oms really was at home playing in Cuba, so I think in his case the Cuban numbers are especially important. The MLEs I have done for Oms don't make any attempt to include an adjustment for home or away-from-home play, but these are factors to keep in mind in his case.

>AVE: .347 (lg .301)*
>OBA: .368 (lg .361)*

Let's just say for the sake of argument that Lundy hit .307 instead of .347 but that the OBA was still .368. Both still above league average which, for a Gold Glove SS, is exceptionally valuable.

Is .307/.368 more valuable than .347/.368? I didn't think so. So why do we penalize a guy who doesn't walk? Aren't those 40 points worth of base hits more valuable than the same number of BB? I think we way overreact to BB as opposed to OBA.

PS. this is why I just moved Lundy up from the 40s to the 20s. Still not on-ballot, but as we work our way into the backlog, I now want him to get into consideration for a ballot slot. His fellow gloves in the 15-30 range are Rizzuto, Ellie Howard, Ken Boyer and Bobby Doerr and Joe Sewell (the latter two HoM not PHoM). Lundy moved back up past Marvin Williams, Vern Stephens, Cupid Childs and Pie Traynor (in the 30s) in this latest move back up the scale. Bando, Clarkson, Trouppe, Bresnahan, Monroe, Bancroft and Rosen (40s and 50s) are in the next cluster of glovemen.

Could someone post the latest season-by-season MLE's for Lundy? Estimates of what his FRAA/Fielding WS might have been would also be greatly appreciated. You all know how much I like my shortstops, so he's a very good bet to make my ballot if I can get the data. Thanks.

So, I'm looking back at a few NgL guys to see if I've really got them right or not. And in so doing, I realized that the HOF data on Lundy is really different than the Holway data in post 6 that Chris used to create his initial projections.

here's a side-by-side look at how different:
In Negro Leagues


Chris did incorporate some of Gary A's data into the mix, presenting yet another variable. Anyway, as I've gone through the data, I'm finding Lundy's numbers to look a little different.



As you can see the difference is primarily in the OBP. I'm using the HOF data, which provides walks, and Lundy walks about 8% of the time, a little higher than Chris had him since Chris had limited walks data at the time.

The other big difference is in the BWS. I looked into this a little. I have Lundy with 5800+ outs. So I used the SBE to look up players from 1915-1940ish with 5000+ outs and a range of somewhat comparable OPS vs lgOPS and OPS+. I looked up their BWS, divided their BWS by their outs, multiplied by Lundy's outs to see how Dick stacked up and to see if my estimate, 40 BWS higher than Chris' in fewer PAs, was reasonable.

When I ranked them back out by their BWS in Lundy's outs I got this:

Manush 266 (similar outs, so i chose him as top-level outlier for reference)
Rice 237
Judge 233
Frisch 229
(Lundy 221)
Myer 217
Sewell 209
Traynor 206
Waner 188
McManus 164
Dykes 161
Bartell 157
Grimm 156

Given that Myer, Sewell, and Traynor all post better OPS+ than Lundy, his BWS appear to be about 20 too high. Or about 10%. The reason I think this occurs is that Lundy is posited by my system to be the only better-than-average or worse-than-average player on his team. And this status is calculated relative to position (see Clarkson thread for details). And since Lundy was usually above average for a SS, that means that he ends up being the best hitter above his position on these theoretically average teams and gets a little more credit for it. Eyeballing it, the effect appears to be worth a little more than 1 WS a year to him. I'm not sure theoretically how to make this go away, or whether I even should.

But another way to look at it is this: the new data and my means to translate it suggest that Lundy is a somewhat better hitter than Chris' most recent estimation, by about 13% on OPS+ (104 v. 92), and by about 19% on BWS (221 v. 185). Given this, we might suggest that applying the 13% increase on Chris' OPS+ to Chris' BWS might be a decent barometer of where I ought to be on BWS. That's about 209. So by that reckoning I'd be about 5% over but very much in the ballpark.

So after two or three years of MLEing, where I'm at today is that I'm prolly a few points over on the WS after several revisions trying to get them theoretically right. To be honest, WS is a cruel master for MLE purposes---I'm not sure there is any such thing as a correct answer to MLE WS. Buyer beware.

Well, of course there isn't, Doc. I think we know that. You're just coming as close as you can which is still great and valuable information.

Now then, here's the MLEs I'm getting for Lundy based on what you've read above. They include the new HOF data as well as Cuban League data, and they are "regressed" in the manner I've discussed before (filling in gaps with a mix of career and recent-prior peformance data), contextualized all to the NL of 1919-1937, with 1922 a whole-cloth year and nothing for 1936 at all.



OK, there you go!

So, Doc, from .299/.333/391/92 to .303/.360/.406/104...

And from 186 + 114 = 300 WS to 221 + 110 = 331. Or 347 adj to 162 games.

Wow, big difference.

Ozzie Smith played 162 games. His totals are 185 + 140 = 325 with an OPS+ of 87 and an A+ glove.

Adjust Lundy to 162 games and it's 232 + 115 = 347 with OPS+ of 104 and apparently a little less than an A+ glove, more like A- maybe. This is 67 percent bat and 33 percent glove. Tony Fernandez was 36 percent glove with a 101 and a B glove. He played 2100+ games over 17 years, 13 years of > 100 games, and 8600 PA. For Lundy it's 18 years, 16 > 100 games and 9300 PA (that's adj to 162 games). So he's Tony Fernandez with 3 extra years as a regular but a little less hangin' around time.

Those 347 WS compare to Vaughan 356, Cronin 333, Banks 332, Ozzie 325. I don't know exactly where Larkin ended up but I'm guessing about 350. All of these were < 30 percent glove. Oddly Ripken is a better parallel with 32 percent glove, but of course 427 WS.

Anyway, Doc. Which is the real Dick Lundy? And was Lundy better than Ozzie Smith? Was he better than Tony Fernandez? Was he Barry Larkin?

FYI on the defensive side for Lundy.

1919-1930: .056 fws/game, or 6.6 per 1000 innings.
1931-1937: .045 fws/game, or 5.3 per 1000 innings.

I think this is what Chris used in his original MLEs, and I just retained it.

6.6 fws/1000 would be synonymous wih an A+ fielder, a legit annual GGer.
5.3 fws/1000 would be synonymous with roughly a B+ fielder.

Any idea how many triples he would have had? I need that to estimate his non-SB baserunning runs.