OK, REALLY Redding's stats here:

Dick Redding's Real Stats

I’ve constructed some MLEs for Dick Redding, based on his stats as available at the Baseball Gauge’s Negro Leagues database and attempting to imitate the methods of Eric Chalek and Chris Cobb. I will describe these MLEs in three major steps:
1. Yearly innings estimates
2. Regressed yearly ERA+s
3. Potential problems/oversights in the MLEs

Here are the stats that I used:

Year Team Lg IP ERA+
1911/1912 Fe Cuba 96.7 69
1912/1913 Lincoln Giants Cuba 24 63
1912/1913 Fe Cuba 138.3 137
1914/1915 Lincoln Stars Cuba 48.7 108
1915/1916 ABCs Cuba 76 82
1916 Lincoln Giants NL 71.7 175
1917 Chicago American Giants NL 153.7 334
1918 Brooklyn Royal Giants NL 18.7 INF
1919 Brooklyn Royal Giants NL 8.3 294
1919 Atlantic City NL 88 258
1920 Atlantic City NL 112.7 90
1921 Atlantic City NL 224.3 120
1922 New York Bacharach Giants 141.7 131

For 1923-1927, I used the Hall of Fame numbers, which don’t give league averages. Eric Chalek presented league average runs per game estimates using league BA and SLG and the RC formula, to which I have added 0.46 runs per game to account for the impact of errors, as determined by comparison of Eric’s estimates to actual runs data for the few years we have it. Lastly, before computing RA+ I adjusted Redding’s RA numbers downward by 8%, because his team during this period, the Brooklyn Royal Giants, essentially functioned as a road team.

Year IP RA+
1923 53.3 94
1924 59 96
1925 43.3 115
1926 34 124
1927 26.7 74

I assigned each winter league season to the preceding summer (eg 1911/1912 becomes 1911) in order to maximize the number of MLE seasons, avoiding an overlap between 1915/1916 and 1916. While the most logical method would seem to be to assign half of each winter-league season to the surrounding summers, this method is both much more cumbersome and produces 1 more MLE season than Cuban Winter League season, so I opted for the simpler method described above. Of course, this means that since these MLEs are based on the Cuban Winter League only for 1911-1915, they aren’t directly comparable to Chris Cobb’s MLEs for this time period.

For Negro League seasons, I used the conversion rate of 0.85 used by Chris Cobb and Eric Chalek in their previous pitcher MLEs. For the Cuban winter leagues, I compared the ERA+s of major league pitchers touring in Cuba to their surrounding-season major league ERA+s, and weighted the ratio by innings pitched, arriving at a conversion factor of 0.82.

I estimated Redding’s MLE innings pitched by assessing whether he would have been a #1, 1/2, or 2 starter that year based on his team and league rank in innings pitched. For a #1 starter, I average the innings of all 16 team-leading pitchers in the majors; for a 1/2, I used the average IP of all 32 team-leading and second-on-team pitchers in the league; and the innings for a #2 are computed similarly. In no season do I see Redding as anything less than a #2 in terms of usage.

I used this method to estimate Redding’s workload for all seasons in which we have complete seasonal data from the Baseball Gauge: 1912, 1914-1917, and 1919-1922. 1911 is left out because Redding’s horrible MLE (57 ERA+) will result in a rating substantially below replacement level, to the point that he would not have pitched in the majors for very long at that level of performance. I won’t go into all of the details of how I determined which seasons corresponded to what workloads, because I’m only going to use this method to set total IP anyway, for reasons I will elaborate on after presenting the table. I conservatively estimated Redding’s 1913 innings as 250, compared to 282 in 1912 and 283 in 1914. For 1918, I considered Redding a #1 starter to determine war credit, as Redding’s best two seasons came in 1917 and 1919 and I consider him a #1 starter from 1917-1921. From 1923-1927, I used the data from the Hall of Fame study to guide my estimates; after pitching between 100-200 innings in 1920-1922, Redding slipped to around 50 IP from 1923-1927, indicating that he was only used as a spot starter/reliever. I thus estimated that he would have accrued 100 innings in the majors in each of these five seasons.

Year MLE IP Role
1912 282 #1/#2
1913 250
1914 283 #1/#2
1915 272 #1/#2
1916 280 #1/#2
1917 299 #1
1918 265 #1
1919 263 #1
1920 297 #1
1921 291 #1
1922 248 #2
1923 100
1924 100
1925 100
1926 100
1927 100
TOT 3530

My estimate of Redding’s total innings is quite similar to Chris Cobb’s/i9s’ earlier estimate of 3556 IP, placing Redding comfortably in the middle of his cohort and thus validating my estimation method. However, there are some notable incongruities: first, the dip in 1918 and 1919, even as Redding remained an ace, and second, the lack of seasonal variability in innings pitched, with Redding pitching as many innings in his mediocre first half of the decade as in his outstanding second half. The first issue results from the shortened schedule in 1918 and 1919; giving Redding short-schedule credit bumps his 1912-1922 total innings up to 3118 and his career total to 3618 IP. There is no reason to adjust this number downward to match with Chris’s estimates, as Chris did not give Redding any short-season/war credit in those innings estimates. The second issue is a real artifact of the method used, as using average innings for #1, #1/#2, or #2 starters heavily regresses Redding’s seasonal innings pitched to the mean. The career shape produced is highly unrealistic—Chris/i9s has Redding topping 300 innings 4 times, peaking at 354 IP in 1921, while my method has no 300+ innings seasons. Furthermore, this method produces little relationship between Redding’s innings pitched and his ERA+, even though we would intuitively expect more IP in a season with a better ERA+: the correlation between Redding’s innings pitched and his final regressed MLE ERA+s is just 0.02. I computed similar correlations for all members of Redding’s cohort as identified in Chris’s post: Johnson, Alexander, Grimes, Rixey, Faber, Coveleski, Cicotte, Shocker, Cooper, Mays and Luque. Considering only seasons with >200 IP, correlations for these pitchers ranged from 0.17 to 0.65, with the mean at 0.44, so Redding’s correlation is clearly unrealistic.

In order to correct for this lack of correlation, I regressed ERA+ against innings for all seasons among Redding’s 11-man cohort in which the pitcher threw more than 200 innings, since we only expect the relationship to exist for regular starters, spot starters and relievers having more variation in their performance. I then used this relationship to predict Redding’s yearly innings pitched given his seasonal regressed ERA+s. Since this method yielded 2886 IP between 1912 and 1922, compared to the earlier estimate of 3118, I scaled the estimates to force the total to equal 3118 (I’m leaving unchanged Redding’s innings estimates from 1923-1927, since he wasn’t a regular starter then).

Year MLE IP
1912 272
1913 273
1914 274
1915 265
1916 284
1917 321
1918 309
1919 299
1920 266
1921 275
1922 280

The variation in ERA+ only accounts for about a quarter of the variation in innings pitched, and as a result, Redding’s resulting innings estimates still appear artificially flat. However, the other three-quarters of the variation is difficult to ascribe to any one cause and can thus mostly be attributed to random variation. While such randomness could be accounted for in Redding’s MLEs, it would bias the estimates by asymmetrically moving innings from worse seasons into better seasons (or vice versa), without any good reason for doing so. As a result, the innings estimates above are probably superior to any attempt at “unflattening” them.

Yearly ERA+s were first calculated by multiplying the ERA+s from the Baseball Gauge database above by 0.85 for Negro League seasons and 0.82 for Cuban Winter League seasons. These estimates were then regressed by filling in the remaining innings (MLE innings minus actual innings) with career-average pitching. This process is based on the assumption that the standard deviation about Negro League ERA+ is the same as the standard deviation about major league ERA+, given equal innings pitched. While this assumption is probably not true, quantifying the actual standard deviation is difficult and I’d rather just use the simplest possible assumption, that the two are equal. Incorporating the difference in standard deviations probably won’t result in a major change anyways. In any case, this is most likely a more valid process than simply regressing all seasons the same amount, because the seasons in which Redding threw more innings actually do have less variation about their ERA+s and thus deserve to be regressed less. After regressing, I tweaked the seasonal ERA+ estimates such that Redding’s career ERA+ using the regressed estimates equaled his ERA+ calculated using the initial estimates (104); this tweak resulted in very minor changes, however.

Unfortunately, this method contains a contradiction: regressed ERA+ is calculated using seasonal innings estimates, but seasonal innings estimates are dependent on regressed ERA+. In order to resolve this issue, I repeated the process iteratively. First I calculated regressed ERA+ using the first innings estimates presented above, which do not depend on the seasonal ERA+. Next, I computed innings based on this regressed ERA+ (the second innings estimates presented above). Then, using the initial unregressed ERA+s, I calculated new regressed ERA+s based on the new innings estimates. I then repeated this process once; at the end of this iteration, the innings estimates had stabilized, with seasonal estimates changing by less than 1 inning from a potential third iteration, and by at most 3 innings from the first iteration.

For 1913 and 1918, I simply filled in the season with the average of the two surrounding ERA+s. As a result, Redding’s war credit year comes out as his second-best season. However, I don’t think this is an unreasonable assumption, as Redding’s two best years clearly came in 1917 and 1919, and Redding was lights-out in the 18 innings he did manage to pitch in 1918.

WAR was calculated in the standard way by computing Redding’s seasonal winning percentage, subtracting replacement level (.427, the average of Redding’s cohort, slightly higher than .420 to account for relief innings) and multiplying the result by innings pitched divided by 9. Win Shares were estimated from WAR based on the relationship between WAR and WSAB, then converted from WSAB to regular Win Shares by inverting the WSAB formula.

Year IP ERA+ WAR WS
1912 273 100 2.2 15
1913 274 101 2.3 16
1914 275 102 2.5 16
1915 266 91 0.8 11
1916 286 114 4.2 21
1917 318 151 9.1 36
1918 306 138 7.5 31
1919 297 127 6.0 27
1920 266 91 0.8 12
1921 276 103 2.7 17
1922 281 108 3.5 19
1923 100 90 0.2 4
1924 100 90 0.2 4
1925 100 102 0.9 6
1926 100 105 1.0 6
1927 100 89 0.2 4
TOT 3618 107 44.2 247

Redding also didn’t completely embarrass himself at the plate, with a 62 OPS+ that translates roughly to something around 40 OPS+ in the majors. Using his contemporaries as a guide, that means Redding probably added another win or so to his totals through his hitting prowess. Depending on popular demand (or how close Redding comes to my ballot) I may compute yearly batting estimates for Redding to get a better handle on his value added through offense.

The picture we get here is a high but very short peak player, reminiscent of maybe Bucky Walters or Dwight Gooden. Redding is also highly dependent on the war-credit year of 1918, but again, I think it’s completely defensible to project him as one of the league’s top pitchers that year. But overall, the picture painted above does not look like a HOM quality player, unless maybe you give a massive amount of weight to single-season peak.

However, there are reasons to believe that these numbers underrate Redding. First, this MLE’s career shape is quite different from Chris Cobb’s earlier estimates (I included Chris’s war credit estimates for consistency):

Year My WS Chris’ WS
1911 0 27
1912 15 11
1913 16 5
1914 16 14
1915 11 40
1916 21 33
1917 36 27
1918 31 15
1919 27 13
1920 12 19
1921 17 21
1922 19 19
1923 4 13
1924 4 5
1925 6 8
1926 6 2
1927 4 7
1928 0 6
TOT 247 267

My numbers are drastically lower in 1911 and 1915, much higher in 1918 and 1919, lower in 1923, and terminate a year earlier than Chris’s. Since Redding only pitched in one game in 1928, I think it’s reasonable not to give him MLE credit for that year. Similarly, Redding’s 1923 numbers are pretty mediocre by the HOF data, with just a 93 RA+ even with the road adjustment, so I feel that 4 WS is somewhat more justified than 13. In 1918 and 1919, the evidence that Redding was a top-flight pitcher simply appears to be a lot stronger now than it was then—all Chris had to go on was a 3-3 record in 1919, a far cry from Redding’s 261 ERA+ from the Baseball Gauge numbers. Chris also notes military service in 1919, though this appears inconsistent with Redding’s rank of fourth in the league in innings pitched.

This leaves 1911 and 1915—and those differences are gigantic. In 1911 I see Redding as a sub-replacement pitcher, based on his extremely poor CWL performance (69 ERA+), whereas Chris, using Redding’s excellent statistics as presented by Holway (5-1, 5.35 TRA) projects him as a star. In 1915, Redding again performed poorly in the CWL (82 ERA+) despite outstanding statistics according to Holway (6-2, 2.55 TRA, similar to 1917 stats) and MVP-level performance in Chris’s MLEs. Indeed, none of Redding’s CWL performances come close to Chris’s MLE estimates for 1911 and 1915.

Brent/134 suggests that Cuban umpiring may have been suspect and possibly biased against American players, partially explaining Redding’s poor Cuban performances. It’s also likely that Redding simply performed much better in the U.S. than in Cuba. While Redding’s poor Cuban pitching did happen and can’t just be dismissed, it seems defensible to bump up Redding’s 1911 and 1915 considerably in accordance with his excellent numbers from Holway. In particular, his 1915 performance earned him Holway’s George Stovey Award, a performance inconsistent with just 0.8 WAR. Right now, I feel that 5 WAR for 1915 and 2 WAR for 1911 is appropriate, though I’m open to suggestions and wouldn’t advocate using these numbers as anything more than wild guesses. With 1 WAR of hitting credit, this brings Redding’s career totals up to a more respectable 51 WAR.

Overall, I think that these MLEs can answer the question posed by Paul in 92 of whether Redding is a peak or career candidate with an emphatic answer in favor of peak. While the overall MLEs fall a bit short of the established HOM line, they by no means completely discredit Redding’s case (though he’ll definitely have a harder time making my ballot). If you like peak candidates, feel my numbers underrate Redding from 1911-1915 (and maybe also 1923-1927), and agree with the substantial war credit I give him (all reasonable positions to hold), then the above estimates aren’t inconsistent Redding as a good candidate for the HOM.

Alex (and anyone else), can you take a look at my last post (#176) on the number of electees by year thread to let me know if I'm on the right track there? Thanks!

BTW, thanks for the update on Redding. I'd been thinking of him more like Grimes, but seeing him as Walters or Gooden is definitely a change. I'll have to re-evaluate him.

Compare him side by side with contemporary Babe Adams (who could also hit a little bit).

Year WAR-DR WAR-BA
1909 XXX 3.6
1910 XXX 3.4
1911 2.0 6.8
1912 2.2 2.2
1913 2.3 8.5
1914 2.5 3.4
1915 5.0 2.0
1916 4.2 -1.9 (zero this out)
1917 9.1 XXX
1918 7.5 -0.7 (zero this out)
1919 6.0 5.9
1920 0.8 4.6
1921 2.7 3.7
1922 3.5 3.5
1923 0.2 1.6
1924 0.2 1.6
1925 0.9 0.1
1926 1.0 XXX
1927 0.2 XXX
TOT 51.0 50.9

It's a slightly higher peak for Redding and more concentrated but these two players look remarkably similar. Adams even had a 50 OPS+ bat.

Alex King's work in 140-141 is very impressive. I especially appreciated the detailed description of how he regressed the numbers.

With the update of 1914 & 1915 seasons at seamheads.com, we now have even more data for Redding!

And Alex King - Fantastic work!

Dick Redding Career Stats

Thanks for the new data, KJOK & Gary! I’ve recomputed Redding’s MLEs using the 1914 and 1915 data posted above. 1911/1912 and 1912/1913 still correspond to 1911 and 1912, respectively, but I’ve now split 1914/1915 equally between 1914 and 1915, and done the same for 1915/1916. After changing the initial actual IP and ERA+, I followed the same process as before.

Year IP ERA+ WAR
1912 271 102 2.5
1913 269 100 2.2
1914 268 98 1.9
1915 295 130 6.3
1916 277 109 3.5
1917 318 157 9.7
1918 307 144 8.1
1919 297 132 6.7
1920 264 93 1.2
1921 273 104 2.8
1922 279 112 3.9
1923 100 90 0.2
1924 100 90 0.2
1925 100 102 0.9
1926 100 105 1.0
1927 100 89 0.2
TOT 3618 112 51.9

As before, you can add another 1 WAR for batting and 2 WAR for 1911.

Again, Redding’s 1915/1916 doesn’t seem consistent with his surrounding Negro League stats. Redding goes from a 333 ERA+ in 1915, to an 82 ERA+ in 1915/1916, to a 175 ERA+ in 1916. His strikeout/walk ratio also swings from outstanding in 1915 and 1916 (2.82 and 2.00 respectively) to poor in 1915/1916 (0.86). Redding’s 1914, on the other hand, fits much better with the surrounding Cuban winter league seasons.

However, there doesn’t seem to be any glaring change of league strength between 1914 and 1915, as the league size stayed constant. The two other Negro League pitchers also performed worse in 1915/1916 than in surrounding years, however—Dicta Johnson had a 90 ERA+, compared to 97 and 127 in 1915 and 1916, and Jim Jeffries had an 83 ERA+, compared to 141 and 110 in 1915 and 1916.

It’s possible that, as alluded to earlier, the umpires may have been biased against American pitchers—KJOK/Gary A, do you have any specific information regarding the 1915/1916 season? Currently, I’m giving equal weight to Cuban Winter League innings as to Negro League innings, but this could change if someone can convince me that the Cuban Winter League seasons are a less valid assessment of Redding’s skills than the surrounding Negro League seasons.

Alex, first of all, great work! Probably the most important thing to know about Redding and Cuba is that a lot of his Cuban performance comes from American Series games--that is, visits by Negro league teams to Havana (abbreviated "ASN" in the DB--for Redding, his games for the Lincoln Giants in 1911/12, the Lincoln Stars in 1914/15, and the ABCs in 1915/16). The Negro league teams ought to be considered the road team for these games, and Redding's record in those series is in effect a road record, so you might want to adjust for that.

His games in the Cuban League, on the other hand, should be considered neutral site games--they were literally all played in the same park, Almendares Park--so no home/road or park adjustments are necessary.

The ERA+, OPS+, WS, & so on for the regular Negro leagues in the U.S. all contain adjustments for disparities in home/road games, strength of schedule, park effects, and so on, but we haven't done that for the American Series games in Cuba yet.

Hope that makes sense...

Thanks for the advice Gary, I didn't realize that ERA+s in the American Series weren't adjusted for home/road. If I increase the American Series' pitchers ERA+s by 8% (the same home/road adjustment used earlier), the Negro Leagues to Cuba discrepancies seen in 1915/1916 look somewhat more plausible, as Dicta Johnson's new 97 ERA+ is identical to his 1916 total (the other two pitchers are still significantly below their Negro League ERA+s but the difference has obviously become smaller).

Before I compute the revised MLEs, is 8% the correct adjustment to use for home/road? I'm assuming that home field advantage is the same in the Negro Leagues/American Series as in MLB, which probably isn't true--what adjustments are you guys using at Seamheads?

I'll have to defer to Kevin on that one--I'm not sure what exact adjustment he uses. From some work I did a few years ago on home/road splits it did seem to me that the Negro league home field advantage was bigger than the typical major league advantage (probably because the home teams tended to hire the umpires), though I couldn't give you an exact figure. This would hold true in Cuba, where the American Series (Negro League) games were worked by the regular Cuban League umpires.

When the major league teams visited Cuba, by the way, they brought their own umpires.

I've been doing the home field advantage calculation on a year by year, league by league basis, taking the actual measured advantage and regressing it. For MLB the historical home advantage is 10%, and I'd say on average the Negro League home advantage is slightly higher.

Updated Redding MLEs, with a conservative 10% adjustment for the American Series and 1923-1928 seasons.

Year IP ERA+ WAR WS
1911 2.0 WAR
1912 272 104 2.7 17
1913 269 100 2.1 15
1914 266 96 1.6 14
1915 298 134 6.8 29
1916 278 111 3.8 20
1917 318 158 9.7 38
1918 306 144 8.1 33
1919 297 133 6.7 29
1920 263 94 1.2 13
1921 273 104 2.8 17
1922 279 112 3.9 20
1923 100 93 0.4 5
1924 100 93 0.4 5
1925 100 105 1.1 6
1926 100 109 1.3 7
1927 100 92 0.4 5
TOT 3618 112 55.1 273 (not including 1911)

I'll also try to create some actual batting MLEs for Redding rather than relying on a wild guess of 1 WAR for his batting value, though I'm not sure when I'll have these done by (currently, I'm working on Ben Taylor's MLEs).

Doesn't really advance Redding's case a whole lot, but 1912 & 1913 statistics are now up at seamheads.com:
Dick Redding Career Stats

I'll wait until 1911 stats are up in a few weeks to re-run Redding's MLEs. In addition, I'll hopefully have batting MLEs ready for Redding by that time. Other players I'm hoping to run MLEs for include Ben Taylor (currently in progress, but again I'll wait for 1910-1911 stats before publishing here), Bill Monroe, Spot Poles, Julian Castillo and Carlos Moran (the latter two will serve more as a kind of calibration to make sure my process is correct, since Brent's already run MLEs for them using basically the same numbers).