One of the last legal spitballers, that's probably going to hurt his chances with some of the voters right off the bat.

Seems like Byrd's case is, in a mild way, like Dihigo's. Those who strictly follow the value-to-team approach will find him very valuable to his teams, while those that want to understand him in an MLB context may see his reliance on the spitter as making him less likely to perform well in an MLB setting. Some type of hybrid thought process may be required here to understand how he fits into an integrated ballot.

Some spitball questions for our Negro League experts:
1) After 1920, was the spitter usually used in the Negro Leagues as a primary or complementary pitch?
2) Did Byrd, himself, have other pitches that were nearly or equally as effective?

A question for the whole group:
3) Assuming that baseballs were replaced more frequently after 1920 in all leagues, including the NgLs, would a spitballer gain nearly the same advantage as previous generations of spitballers did?

Not to be confused (please!) with Harry Byrd, or Charlie Parker for that matter.

As to Byrd's spitball, I will anticipate David: ML rules are irrelevant. I mean, it's one thing to ask what Buzz Arlett would have done in the MLs--he could have played in the MLs. Byrd could not have, so how he would have done in the real MLs is not relevant. How he compares to Redding and Mendez and Ray Brown and Satchel, et al, that is the question. And if, by chance, he was the best of them all (doubtful), then he is a HoMer regardless of whether Judge Landis or Happy Chandler liked his out pitch or not.

Career leaderboard information for Bill Byrd

WINS 6th at 133

LOSSES 4th at 84

DECISIONS 3rd at 217

WINNING PCT .613
(50+ Decisions) 31st
(25+ Decisions) 50th
(10+ Decisions) 84th

ADJ PCT OF TEAM DECISIONS 28.2%
(50+ Decisions) 9th
(25+ Decisions) 13th
(10+ Decisions) 25th

WAT 4th at 19.0

WAT PER DECISION .087
(50+ Decisions) 24th
(25+ Decisions) 40th
(10+ Decisions) 68th

APPEARANCES ON YEARLY WINS LEADERBOARDS

1936: 2nd in NNL with 10 wins.

1937: Tied for 3rd most wins in NNL with 9, tied for 3rd most wins in NgLs.

1938: Tied for 4th most wins in NNL with 9, 5th in NgLs.

1939: 2nd in NNL with 15, 2nd in NgLs.

1941: Led NNL and NgLs in wins with 15.

1942: Led NNL and NgLs in wins with 14.

1943: 3rd in NNL in wins with 9, 4th in NgLs.

1944: t-4th in NNL in wins with 8, t-5th in NgLs.

1945: t-3rd most wins in NNL with 7, t-5th most wins in NgLs.

1947: t-4th in NNL in wins with 9, t-4th most wins in NgLs.

1948*: Led NNL in wins with 11, t-3rd in NgLs.

*NOTE: Holway's data for 1948 is extremely limited, in fact he only shows wins for nine pitchers (all of whom had 6 or more wins, so it's not like it's him and a bunch of one-win, short-timer chumps.)

OPP+ and hOPP+ for Bill Byrd

OPP+ of 105

770.375 adj team decisions out of 736.686 average team decisions.

So he had five percent more opportunity for documented decisiosn than his peers in the Negro Leagues.

hOPP+ of 89

770.375 adj team decisions out of 867 historically average team decisions.

Byrd is at an 11% opportunity disadvantage versus an historically average NgL pitcher.

I'd like to take back what I said in post #2. Given Byrd's long-time record of effectiveness, his high WAT scores, and the fact that in comparison to the pitchers in the backlog he did so with less opportunity, I'm convinced he's potentially got enough to get on the ballot.

However, there is one sticky point: whether his 1940-1945 seasons should be discounted in some way and by how much. The reason for asking this is twofold:
1) 1940-1941 (and to some degree 1942-1945): the effect of the exodous of top talent to the Mexican League on the quality of stateside play

2) 1942-1945: the effect of WW2 on the quality of play

Byrd's most effective prime period by WAT is 1936-1947 where he racked up 18.2 of his 19+ career WAT. However 1940-1945 represent 12.3 WAT by themselves.

Determining what kind of discount structure is appropriate for him, if any, may determine whether he's a Ruffing/Rixey kind of pitcher (all bulk) or a true peak + bulk candidate.

Yeah, discounts for the early 1940s will have a significant impact on Byrd's candidacy (and on Ray Brown's, too).

I've started working on MLEs for Byrd; I hope to have them ready by the time the discussion of 1953 begins. (Win Shares for Willlie Wells are my other Memorial Day weekend project.)

I don't intend to change the conversion factors for any of the 1940-45 seasons, so that voters can apply the discounts, if any, as they see fit.

My own plan is not to discount 1940, to disount 41-42 by 3%, 43 by 5%, and 44-45 by 8%.

44-45 discounts are slightly less than for ML players in those seasons, because the NeL lost fewer of its top stars, I believe, to military service.

Chris, I'm confused about your "base" for these discounts. The 3-5-8% NeL discount of '41-'45 is vs. a base of what?

And what is the more or less "global" NeL conversion to MLE?

And it would change then for these years, right?, since both are being discounted but by different amounts? So even with an 8% discount in '44, the NeL conversion to MLE would be closer to 1.00 than normally?

But back to the "discount," do we know of NeLers who were out of baseball due to the war? Or is it more that they were in Mexico?

Specific to Byrd, aside from Erik's leaderboard info, have we actually seen Byrd's yearly results a la Holway or Riley? I'm not sure he doesn't look like #3 after Mendez and Redding, though probably closer to Winters and Cooper and Brewer (or whomever else is in the running for #3) than to the top two guys.

Chris, I'm confused about your "base" for these discounts. The 3-5-8% NeL discount of '41-'45 is vs. a base of what?

The National League norm for that season, which is what the MLEs aim to produce.

For these years, I will convert the NeL data from a given season to the corresponding major-league season, as usual, so 1945 MLEs will be intended to show what the player would have done in the National League that year. If you apply a competition discount to major-league baseball for 1945, you should apply that discount to the MLE as well. However, the Negro Leagues were somewhat less affected by the draft than the majors, so I propose myself to use an 8% discount for the Negro-Leaguers for this season rather than the 10% I will be using for the majors. This discount will _not_ be included in any MLEs I post (it was not included in the Josh Gibson MLEs, for example.)

Does that clarify things?

And what is the more or less "global" NeL conversion to MLE?

The conversion factors are .90 for batting average, .82 for slugging average. I've been using .85 for ERA.

But back to the "discount," do we know of NeLers who were out of baseball due to the war? Or is it more that they were in Mexico?

They were in Mexico in 1940-41, and there were a lot of them there, especially in 1941. A few were in Mexico 1942-45. A few were in the military in 1942, more in 1943, the most in 1944-45.

Lists of players in Mexico and players in the military have been posted somewhere on the site.

Specific to Byrd, aside from Erik's leaderboard info, have we actually seen Byrd's yearly results a la Holway or Riley? I'm not sure he doesn't look like #3 after Mendez and Redding, though probably closer to Winters and Cooper and Brewer (or whomever else is in the running for #3) than to the top two guys.

Yearly data from Holway hasn't been posted yet -- I'm working on it. He could well be #3 after Mendez and Redding. I don't know enough yet to have an opinion, except that 1) I'm definitely more impressed with him than with Chet Brewer and 2) he definitely threw a good deal more innings than Nip Winters.

Early projections place him with MLE innings between 3100 and 3600, with a support-neutral ERA+ of somewhere between 102 and 112. Unless the fine-tuned projections place him at the upper end of both ranges, he's not a HoMer, comping instead to Mel Harder, Paul Derringer, Larry French-type careers. And even at this upper end he would be a borderline candidate. But the upper-end projection is still possible, and until I examine his individual seasons and the makeup of his teams more carefully, I certainly won't rule it out.

One factor that he does have going for him is that he generally pitched on teams with very good pitching staffs, like Joe Rogan did, and he was still the best pitcher on these teams, again like Rogan. This means that his straight wins-above-team numbers, which are very good but not great, probably underrate him, and they project him to an MLE support-neutral ERA+ of 102. So, he was at least that good. How much better is what the fine-tuning analysis will seek to figure out.

Bill Byrd Data

Played most of his career for the Elite Giants: the franchise moved a lot, but it had the same owner and continuity in the core of players.

Born July 15, 1907

Since his leaderboard appearances are posted elsewhere, I’m not including them here.

Seasonal Data from Holway

1933 5-6 for Cleveland Blue Birds; team 21-26; #2 on team in decisions (tie), 0.1 WAT
1934 3-11 for Cle Red Sox; team 8-28; #1 in dec.; -0.2 WAT
1935 2-3 for Nashville Elite Giants; team 16-17; #3 in dec (3-way tie); -0.5 WAT
1936 10-5 for Was Elite Giants; team 27-18; #1 in dec; 1.5 WAT
1937 9-6 for Was Elite Giants; team 27-17; #1 in dec; -0.3 WAT
1938 9-2 for Bal Elite Giants; team 28-9; #1 in dec., 1.7 WAT
1939 15-4 for Bal Elite Giants; team 24-9; #1 in dec., 2.8 WAT
1940 In Venezuela – No Data
1941 15-6 for Bal Elite Giants; team 38-21; #1 in dec., 2.3 WAT
1942 14-4 for Bal Elite Giants; team 31-17; #1 in dec.; 3.8 WAT
1943 9-6 for Bal Elite Giants; team 15-26; #1 in dec., 5.5 WAT
1944 8-7 for Bal Elite Giants; team 21-18; #1 in dec., -0.1 WAT
1945 7-4 for Bal Elite Giants; team 19-14; #1 in dec., 1.0 WAT
1946 4-7 for Bal Elite Giants; team 21-26; #3 in dec., -1.2 WAT
1947 9-6 for Bal Elite Giants; team 44-41; #3 in dec., 1.5 WAT
1948 11-6 for Bal Elite Giants; team 38-20; #1 in dec., -0.2 WAT
1949 12-3 for Bal Elite Giants (this data from Riley, so not included in career totals)
1950 no data (started season with Elite Giants, but retired shortly thereafter)

Here’s a first stab at Bill Byrd MLEs. I’m going out of town through Sunday, so I won’t be adding any analysis to this until then, though I’ll try to post an explanation of the methods tomorrow from the road, if I can find a networked computer. It’s basically the same method I used for Bill Foster, so his totals can be compared to Foster’s. This is not a strict translation. Usage patterns were too different from the majors for innings pitched to be directly projectable, and the lack of team batting or fielding data (not to mention good park factors) for most seasons makes any formulaic calculation of pitching value for Negro-Leaguers impossible. This is the way I reconstruct Byrd’s career.

IP are estimated based on whether Byrd was used as a #1-#5 starter on his teams, how he ranked compared to other pitchers with that usage pattern in the NeL that year, and what the NL usage patterns were that year.

DERA is fielding neutral ERA, normalized to 4.5 r/g environment. It is meant to be equivalent to WARP’s DERA.

SnW-snL is support-neutral w-l

Snwp is support-neutral winning percentage

Year IP  DERA snW-snL    snwp
1933 189 5.39 8.9-12.8  .411
1934 107 6.55 3.9-8.4   .321
1935 171 5.35 6.6-13.1  .338
1936 282 4.09 17.8-14.6 .548
1937 225 5.01 11.5-14.4 .446
1938 240 2.99 19.1-8.5  .694
1939 305 3.32 22.7-12.4 .648
1940 259 3.63 18.0-11.8 .606**
1941 271 3.94 17.7-13.5 .566
1942 292 3.29 21.8-11.8 .651
1943 247 3.49 17.7-10.7 .624
1944 207 4.80 11.1-12.7 .468
1945 186 4.31 11.2-10.2 .523
1946 109 6.57 4.0-8.5   .319
1947 148 4.32 8.8-8.2   .520
1948 138 4.80 7.4-8.5   .468
16  3376 4.17 208.1-180.1  .536

**Byrd pitched in Venezuela this year. No data is available, so I averaged his performances in 1939 and 1941 and deducted 10% from his IP to make an estimate.

Final notes:

1) Byrd was a very good hitter for a pitcher and often played in the outfield on off-days. His offensive contribution is not accounted for separately, but as these numbers are derived, by a convoluted route, from his actual winning percentages, any contribution his hitting made to his victories is figured in. His actual DERA thus is probably somewhat higher, with batting value making up the difference.

2) The conversion method I am using here tends, I believe, to exaggerate extremes in ways that more sophisticated systems like WARP and WS do not. So I think the career totals are probably more reliable than the seasonal ones, which pull towards extremes. Byrd’s peak is very good, but probably not quite as good as suggested here, and he was also probably much more like an average pitcher than a replacement level pitcher outside of his peak. Regression would smooth things out a bit: these MLEs are not regressed.

3) Byrd’s weak performance in his early years is of a piece with the low estimates of value for all players in the 1932-36 period. Here I think the conversion factor underrates Byrd by underrating the conversion factor for the single-league years of the NeL.

Chris,

I agree that it looks as though the 1932-1936 period may be dragging him down a bit. 1934-1935 in particular caught my eye as being potentially low. If one were to see him in those two years as an improving young pitcher climbing to a combined .500 record, then he would be a .550 pitcher.

All that said, he feels less HOMeriffic than I thought he would, and he'll likely be down in the Rixey area of my ballot instead of up by the top guys.

Description of the projection method used above.

This is derived, ultimately, from wins above team. First, I assume that unless there is evidence to the contrary, the team record is due equally to hitting and defense. Setting the run environment to 4.5 r/g, I use the pythagorean method to estimate team rs/g and ra/g necessary to achieve their winnng percentage. I then examine the team roster to decide if hitting, pitching, or fielding is likely to have been a greater or lesser contributing factor to the team's success and adjust the rs/g and ra/g accordingly, attributing 2/3 of responsibility for ra/g to pitchers and 1/3 to fielders.

Using the team rs/g, I then use the Pythagorean method to find the ra/g needed for the pitcher to achieve his own winning percentage. I adjust that by the r/g the fielders are above or below average, as previously guestimated, and that gives me the equivalent of WARP's DERA for the pitcher -- fielding-neutral run average. I divide 4.5 by this number to get a DERA+, multiply that by .85 as the standard MLE conversion factor, and use the pythagorean method from there to generate a MLE winning percentage, which I apply to the estimated innings, divided by 8.7, to get the estimated support-neutral MLE won-lost record.

These estimates, overall, have four significant elements of uncertainty:

1) are my IP estimates reasonable?
2) are my adjustments of responsibility for hitting, pitching, and fielding for the team record reasonable?
3) Is the conversion factor accurate?
4) are the estimates at very high and very low ra situations distorted by my failure to adjust the exponent?

If I had more calculating power and a better understanding of pythaganpat/port/etc., I could probably eliminate element of uncertainty #4.

For the rest, I do the best I can, but I could easily be wrong. Use the estimates as you think best, and if your own examinations lead you to significantly different conclusions, I'd be happy to see what an alternative analysis looks like!

Hey, Chris, great work, a couple things:

1) IIRC, you use a different conversion rate for offense and pitching. Are these rates related in any way or derived from one another somehow?

2) I was doing the math along with you, and I'm getting a different result for 1935. I've come up with an .414 winning pct for that year. I think that's more square with the DERAs listed for 1933 and 1935, which are close in value. This would take him to 210-178, .541.

3) I just double-checked the 8.7 inn/dec rate to see if applying a blanket rate would under or overrate the number of MLE decisions Byrd would receive.

At 3376 MLE innings, with an 8.7 inn/dec rate, he's getting 388 decisions.

If we look year-by-year, going by probable rotation slot?

YEAR  BB'S SLOT  NL DEC RATE IN SLOT
1933  2          9.1
1934  1          8.5
1935  3          8.5
1936  1          8.3
1937  1          8.8
1938  1          8.5
1939  1          8.5
1940  1          8.6
1941  1          9.0
1942  1          8.9
1943  1          8.6
1944  1          8.3
1945  1          8.4
1946  3          8.8
1947  3          8.3
1948  1          8.3

Using Byrd's MLE innings, I got 392.8 decisions. So very close on the career side.  Comparing individual seasons, 

<pre>
YEAR    SLOT DEC  MLE DECDIFF
1933    20.8      21.7    -0.9
1934    12.6      12.3    0.3
1935    20.1      19.7    0.4
1936    34.0      32.4    1.6
1937    25.6      25.9    -0.3
1938    28.2      27.6    0.6
1939    35.9      35.1    0.8
1940    30.1      29.8    0.3
1941    30.1      31.2    -1.1
1942    32.8      33.6    -0.8
1943    28.7      28.4    0.3
1944    24.9      23.8    1.1
1945    22.1      21.4    0.7
1946    12.4      12.5    -0.1
1947    17.8      17      0.8
1948    16.6      15.9    0.7
        392.8     388.2    

Using the typical NL pitcher in his slot as the model would yield 4+ more decisions. It yields a record of 212-180, a .541 pct that matches the MLE pct within .0002 points.

Conclusion? The blanket 8.7 inn/dec works perfectly well.

Chris, you asked for an analysis of the innings pitched and called for alternatives, so here's an idea that I've been thinking about trying out.

The idea is to let Byrd's standing among his NgL peers determine his innings alotment, by using decisions as a proxy for innings.

In decisions, season by season, Byrd placed in the top group this way:
1936: 2
1937: 2
1938: t-4
1939: 2
1941: 1
1942: 2
1943: t-2 with three others
1944: 2
1945: 3
1946: t-7
1947: t-5
1948: 1

Obviously, Byrd's usage screams ace. If he was truly ace-worthy, and if innings tend to correlated strongly with decisions, then we can get a strong sense of his innings.

Here's the year-by-year leading totals in the NL in innings pitched alongside the average #1 starter's load year-by-year (the latter represented by a three-year rolling average)

YEAR LDR AVG#1
1933 308 275
1934 315 273
1935 325 277
1936 315 265
1937 292 265
1938 307 255
1939 319 255
1940 305 250
1941 302 251
1942 287 249
1943 303 258
1944 312 249
1945 284 243
1946 266 237
1947 289 242
1948 314 249

So if we do assume that Byrd's going to follow the pattern of pitching-staff slotting he established in the NgLs (which is a different discussion altogether!), and we also assume that the league leaders in innings are all #1 starters (a reasonable though not foolproof assumption), then we can start gingerly plotting Byrd's innings in an 8-team league based on his standings in decisions, figuring the #1 innings guy gets the most decisions and the average #1 starte comes in around fourth (in an eight team league).

Here's what I came up with for the years in which we know how he ranked in his own league. When he finished second or third in decisions, I averaged the league leader and the typical #1 guy as listed above and rounded down a little. When he finished first I came in close to the NL leader in innings. When fourth, the average #1. Then I tried to use a little sense about things after that, though I may have not exercised enough caution.

YEAR  BB'S RANK IN DEC  EST INN1936: 2                 290 
1937: 2                 275
1938: t-4               255
1939: 2                 285
1940: in VZ             275
1941: 1                 300             
1942: 2                 265
1943: t-2               275
1944: 2                 280
1945: 3                 255
1946: t-7               200
1947: t-5               230
1948: 1                 275

Total: 3460 innings

3460 by 8.7 inn/dec = 398 decisions, at the .541 pct established previously, that's 215-183.

Of course, we'd still have to deal with 1933-1935 and possibly 1949....

This allocation of innings would represent a different view of Byrd, one that would place him closer to the long-career Ruffing/Lyons/Rixey/Grimes/Faber 3750-4000+ innings pitchers than the Coveleski/Vance/Newhouser 2750-3250 pitchers.

Does this seem like it's reasonable?

One question I have is this and I raise it only because this has become very complex. You guys are contributing great work, but it is complex and processing is tough. So:

Doc's emphasis on WAT: Being based on Wins, is it not subject to all the problems with Wins? Well, not all of them. But the team Wins vary from pythag, and then pitcher Wins vary from the team by virtue of all the variables that occur day to day. You've equalized for a lot of that, but I wonder if one or the other of you could expound on the pros and cons of WAT in this particular context?

Sunnyday,

I think there's one essential reason to look at WAT: it's the best thing we've got to look at.

For individual pitchers, all Holway reports for (virtually) every guy is his decisions, that's it.

Not G, not GS, not ERA/TRA, not Ks or BBs. Just W and L. WAT is about the most complex stat you can come up with from just Wins and Losses.

And because WAT takes a larger sample than merely the individual's record, it gives one a little more confidence than purely looking at wins and losses.

By making a crude adjustment for opportunity, we can remove some of the biases that come from haphazard and unbalanced scheduling (across one and several years) and try to situate everyone's WAT in a single context to compare these guys against one another. Though with some teams having (count 'em!) two or five or ten decisions, it gets a little weird, and sometimes you just can't make a reasonable adjustment without introducing even more chance for error.

Anyway, between WAT, WAT/decision, win pct, # decisions, rotation slot, and leaderboards, I think the we can circle pretty closely around a guy, just not with the level of certainty we'd like, and I'm surely not the first one to call this process Heisenbergian.

Just for the curious, I just finished reading Frazier "Slow" Robinson's autobiography. He was the #1 catcher for the Elite Giants from 1946-49. His comments on Byrd:

"Bill Byrd was an old-timer who had been pitching since the early thirties. He was past his prime and got by on a variety of knucklers, spitters and big looping curves."

When he names his All-Star team of people he played against and with, he apparently preferred the pitchers he saw when he was with K.C. in 39/40. His rotation is Satchel Paige, Dick Bradley, Hilton Smith, Connie Johnson and Leon Day. On the other hand, he didn't like catching Byrd that much, because of having to handle the spitball, and because he really couldn't call the pitches.

Some of us like to use win shares in our evaluations. The following uses the Bill James "short-form" method to convert the Byrd's MLE innings pitched and DERA statistics calculated by Chris in # 12 to win shares. (I gave the details of the calculation on the Nip Winters thread, # 9).

Year WS
1933 3
1934 0
1935 3
1936 17
1937 6
1938 25
1939 27
1940 20
1941 18
1942 27
1943 21
1944 7
1945 10
1946 0
1947 8
1948 5
Total 197

If you prefer the innings estimates given by Dr Chaleeko in # 16, you can scale these upward.

An interesting comparison is with Vida Blue, career DERA of 4.17 in 3343 innings.

Brent,

I'm guessing you were taking the DERA straight from Chris's chart to plug into the WS formula.

IIRC, DERA is a total-run forumulation, while James uses ERA in short-form WS.

So would DERA need to be adjusted for the presence of UER before it went into the s-fWS formula?

You're right that James uses ERA for the short form method (for the long form, he includes both earned and unearned runs). But since everything here is being calculated relative to league average (which I understand DERA sets at 4.50), I don't think that any adjustment is needed for unearned runs unless there is some evidence that Byrd differed from average in that respect. I don't believe that information on earned and unearned runs is generally available for the Negro Leagues.

I suppose that the fact that Byrd relied on "junk" pitches--"a variety of knucklers, spitters and big looping curves" to quote Frazier Robinson from # 19 -- those types of pitches may be associated with more unearned runs than normal. I'm thinking of Chris J's new kind of the unearned run, knuckleballer Dutch Leonard. But to make the inference that this characteristic would have carried over to Byrd strikes me as speculative.

That should have been "king of the unearned run."

Since it hasn't been mentioned here yet, Bill James lists him as the best Negro Leagues pitcher for 1944, 1948, and 1949. 1949 isn't even included in the MLE/WS estimates - I understand why, since we don't have good numbers to use, but it shouldn't be left out of the record.

Hey, gang,

Please find my latest MLE for Bill Byrd here. As always, when new data arrives, these will be updated. Thanks!