Last week, we shared information showing that the standard deviation of offensive performance in the Negro Leagues and Latin leagues was considerably higher than that of the major leagues of the same era. But players in the Jackie Robinson era also played in the white minor leagues, and someone translating their performance to an MLB level would want to know whether the minors were more like the majors or the Negro and Latin leagues. Did those leagues have wider variation than the majors as well?
Before we turn to the record itself, let’s ask this question. Why might the white minors be dissimilar or similar to the majors in terms of STDEV? We explored some of the structural and game-play drivers of a wider standard deviation in the Negro Leagues last time out. Do they hold for the minors?
So we already see that the conditions that made the Negro Leagues’ performances spread out further from the average than MLB’s in most cases either don’t apply to the minors or are muted. But we need to look at the stats to know whether or not STDEV was nonetheless driven by other factors, or whether it mirrored the majors. As we look at this question, we’ll examine the results level by level using the same technique as in our previous article.
AAA and Open Classifications
Although right around Jackie’s time, these leagues got new classifications or jumped a class, they remained the highest rungs on the sub-major ladder. (For a graphic that helps to visualize how the classifications of minor leagues has changed over the last hundred-odd years, check out this article.)
YEAR MLB | PCL | IL | AA STDEV | STDEV ADJ | STDEV ADJ | STDEV ADJ =================================================== 1946 1.55 | 1.24 1.12 | 1.32 1.09 | 1.30 1.10 1947 1.46 | 1.61 0.96 | 1.58 0.96 | 1.52 0.98 1948 1.59 | 1.85 0.93 | 1.13 1.20 | 1.38 1.08 1949 1.50 | 1.44 1.02 | 1.58 0.98 | 1.56 0.98 1950 1.43 | 1.48 0.98 | 1.52 0.97 | 1.13 1.13 1951 1.50 | 1.66 0.95 | 1.22 1.12 | 1.71 0.94 1952 1.17 | 1.09 1.04 | 1.28 0.96 | 1.62 0.86 1953 1.49 | 1.14 1.16 | 1.53 0.99 | 1.52 0.99 1954 1.65 | 1.14 1.22 | 1.54 1.04 | 1.31 1.13 1955 1.43 | 1.18 1.11 | 1.51 0.97 | 1.20 1.10 1956 1.59 | 1.47 1.04 | 1.33 1.10 | 1.37 1.08 1957 1.85 | 1.23 1.25 | 1.24 1.25 | 1.41 1.16 1958 1.50 | 1.74 0.93 | 1.33 1.07 | 1.04 1.22 1959 1.39 | 1.12 1.12 | 1.42 0.99 | 1.16 1.10 1960 1.15 | 1.14 1.00 | 1.16 0.99 | 1.82 0.82 1961 1.64 | 1.54 1.03 | 1.27 1.15 | 1.17 1.20 1962 1.31 | 1.63 0.90 | 1.36 0.98 | 1.16 1.07 1963 1.22 | 1.01 1.10 | 1.08 1.06 | 1964 1.37 | 1.25 1.05 | 1.47 0.97 | 1965 1.29 | 1.36 0.97 | 0.87 1.24 | --------------------------------------------------- AVG 1.45 | 1.37 1.04 | 1.34 1.05 | 1.38 1.07
How about them apples?! The high minors actually had less standard deviation than the majors. Before we draw conclusions, let’s see if that’s how things play out down the ladder.
In each instance as we tour the minors, I’ve only included seasons where I’m aware that a top Negro Leagues candidate played in a given league. Therefore, there may be gaps in the information I’m presenting, especially compared to the AAA/Open leagues. As it turns out, we only have solid information for seasons in question from one league that would currently be considered AA, and that’s the Texas League. Other leagues included Negro League candidates, but their stats aren’t yet on BBREF, so I couldn’t include them.
YEAR MLB | TXL STDEV | STDEV ADJJ ========================= 1953 1.49 | 1.50 1.00 1954 1.65 | 1.59 1.02 1955 1.43 | 1.28 1.06 1956 1.59 | 1.51 1.03 1957 1.85 | 1.19 1.28 1958 1.50 | 1.26 1.10 1959 1.39 | 1.50 0.96 1960 1.15 | 1.44 0.90 1961 1.64 | 1.56 1.03 ------------------------- AVG 1.52 | 1.42 1.04
Yet again, a minor league is actually a little tighter than the majors….
Today we have Hi-A and Lo-A levels, but that split only occurred in 1990. Before that every A-level team was in the same category. Once again, we only have stats for one league (and one season in it) at this level.
YEAR MLB | WES STDEV | STDEV ADJ ========================= 1958 1.50 | 1.83 0.91
If we’re starting to get into the exurbs with B leagues, we’re going to be out past the boonies with C and D leagues. The lower in the classification system we go, the more localized the leagues and teams are.
YEAR MLB | IIIL | NORW | WINT STDEV | STDEV ADJ | STDEV ADJ | STDEV ADJ =================================================== 1951 1.50 | | | 1.39 1.04 1954 1.65 | 1.62 1.01 | | 1962 1.31 | | 1.27 1.02 |
YEAR MLB | AZMX | CAL STDEV | STDEV ADJ | STDEV ADJ ====================================== 1954 1.65 | | 2.12 0.89 1958 1.50 | 2.13 0.85 | 1.24 1.10 1959 1.39 | | 1.00 1.19 -------------------------------------- AVG 1.51 | | 1.45 1.06
YEAR MLB | FLOR STDEV | STDEV ADJ ========================= 1958 1.50 | 1.43 1.02
My boss’ boss has more than a few pearls of wisdom inside her. She likes to say that a good rule of thumb for making decisions says that if something happens once, it’s an occurrence. If it happens twice, it’s notable. If it happens three times it’s a pattern, and you need to take action. So looking at the minor leagues, we see that nearly every league we’ve looked at, and most seasons in each league we examined, show up as having a lower standard deviation than the majors of the same season. This was at first a surprising result. But maybe it shouldn’t have been?
From an anecdotal and qualitative perspective, it does makes sense that minor league standard deviations are closer to the big leagues than the Negro Leagues were. Rarities such as .400 hitters or 60-homer hitters don’t litter the annals of minor league history, but such batting averages and equivalent feats of batsmanship do occur more often in the Negro Leagues. But that’s also a clue to the minors tighter variance.
I initially thought that because the minors tended to employ less experienced players in the farm-system model, play would be somewhat uneven. Similarly, the minors in this time had more independence than today and were at liberty to sign MLB vets who could no longer keep at job in the show. In the case of the PCL, which signed many such players, guys originally from out west may also have opted to forgo the worst of their decline phase in MLB to play out west nearer their homes. Today a so-called AAAA player might be in his late 20s or very early 30s, but back then, fringe types might be older. Indeed, from 1947 to 1954, the average age in AAA/Open leagues was around 27.5 for the AA, 28 for the IL, and 30 for the PCL. The Coast League was the most active in terms of signing ex big leaguers, for example LA-born and Portland native Joe Gordon for his age-36 season. The majors at that time were around 28.5 years old. From 2011 to 2015, the average age in the PCL was this close to 27 and roughly the same in the IL. The PCL lost three years in average age by becoming a development league, whereas the IL had lost only one year because its teams had been mostly affiliated all along.
In fact, the average age of a league tells us something simple and significant about why standard deviations were so tight: Everyone in the league is basically at the same developmental level. In the minors, if you’re too good, you get promoted quickly. If you’re too awful, you get demoted. If you stay the whole year, you’re getting appropriately challenged for your level of experience. That’s the whole point of the minors! Today this is much more apparent because the average the different levels is more highly stratified than ever. Rookie ball is filled with 18–20 year olds. Short season ball is all 20 or 21 year olds, etc. So no matter what other factors may contribute to the variance in a league, age/experience may be the most important. To be sure, this is isn’t ironclad reasoning, but it does pass the smell test for me.
OK, so we’ve now had a look at the Negro Leagues themselves and some of the leagues that Negro Leagues expats played in. Next time out, we’ll take a look at how all of this may change our perspective on the offensive value of some famous blackball heroes.
How good were the Negro Leagues? If you’re considering translating Negro League statistics to a Major League setting, you have to have an answer to this question. If you want reasonable translations, you have to have a really good answer to the question. If you want all your translations to line up systematically, you have to answer that same question for many, many leagues. So today, that’s what we’re considering.
In some sense, organized baseball has answered parts of the question for us. Under the National Agreement, the minor leagues have been classified since the early 1900s. We’ll soon see how those classifications have changed repeatedly over the years, but fans today recognize this structure:
Today, the classification of these leagues represents a ladder that young players climb on the way to The Show. In past eras, however, the National Agreement based classification on the size of the population the league served. But when you think about, this was a strong proxy for quality. Of course, the larger the area you drew from the more talent you could scout locally, the more ticket sales you could do. But remember, unlike the organized minors of today, until the 1960s or so, most minor league teams were trying to win their league’s pennant. So fans of the time also exerted more pressure on minor league squads to win. The point is this: The ladder existed then and exists now.
So what’s this got to do with Black Ball? Simply put, if we can figure out the quality of play at each minor league level, we may be able to place the Negro Leagues and other independent leagues that signed dark-skinned players into the framework. It’s a method that can produce a reasonable and familiar estimate of play.
Here’s a timeline of minor league classifications presented for puzzlement/enjoyment. The hashed arrows indicate that a league shifted to a new level. It’s a pretty wacky timeline, so…this is a blues riff in B, watch me for the changes, and try to keep up. Okay?
What if we knew the discount (if you will) off of major league performance for each of these leagues? That is, if we knew that a person created 100 runs in AA, what percentage of those 100 runs created would he give back by moving up to the majors?
Luckily for you and me, others have plumbed these very depths and done the math. I pieced together information from an excellent article by Ben Lindbergh at the former Grantland as well as some of Clay Davenport’s work to reach some SWAGs for conversion rates for leagues in the current minor-league classification system. To the best of my ability to use Google, I haven’t been able to find an updated table that includes all levels and indy and international leagues and their conversion rates to MLB.
To provide some context, let’s see how the discount structure works using two players’ 2016 seasons. Mike Trout led the AL with 148 runs created, 64 more than average in his 681 plate appearances. Lorenzo Cain created 52 runs in 434 PAs, exactly average for a player in his playing time in the AL of 2016. What would we expect these guys’ major league performance to be if they had created the same number of runs in AAA? Or in AA?
CONV LEVEL LEAGUES RATE TROUT CAIN ====================================== MLB AL NL 1.00 148 52 AAA IL PCL 0.80 118 42 AA EL SL TXL 0.72 107 37 Hi-A CAL CAR FSL 0.62 92 32 Lo-A MWL SAL 0.58 86 30 SSA NYP NWL 0.50 74 26 NCRK APP PIO 0.50 74 26 CXRK AZS GCL 0.50 74 26 FNRK DMS VZS 0.50 74 26 AAA MXL 0.49 73 25
When you are as good as Trout was in 2016, you can be playing as far down on the farm as LoA and still produce an approximately average MLB season. Even down in Rookie ball, you’re not yet at replacement. On the other hand, Cain plummets to roughly replacement level in Hi-A. Now, obviously, this bootstrapping-like method has limitations. Guys in the Arizona Summer League have probably never seen a great breaking ball and won’t until they hit A ball. But on the whole, it appears defensible because it’s telling us that a Trout-like season by a veteran player in Lo-A would only appear as about average in the bigs. As we’ll see below, this may make good sense.
Let’s bust thing out a little further to include some foreign and independent leagues.
CONV LEVEL LEAGUES RATE TROUT CAIN ============================================ MLB AL NL 1.00 148 52 INT NPB 0.90 133 47 AAA IL PCL 0.80 118 42 WINT DMW 0.80 118 42 AA EL SL TXL 0.72 107 37 IND ATLANTIC 0.72 107 37 WINT AZF PRWL VZWL MXWL 0.72 107 37 INT CUBA 0.63 93 33 Hi-A CAL CAR FSL 0.62 92 32 IND AA CANAM 0.62 92 32 Lo-A MWL SAL 0.58 86 30 IND FRONTIER 0.58 86 30 SSA NYP NWL 0.50 74 26 NCRK APP PIO 0.50 74 26 CXRK AZS GCL 0.50 74 26 FNRK DMS VZS 0.50 74 26 AAA MXL 0.49 73 25 INT KBL 0.49 73 25
That’s a pretty reasonable spread to work from, right? So how would the Negro Leagues fit into this? The Negro Leagues are variously described as anywhere between Nippon Pro Baseball and AA quality. That would put them in the range of 0.75–0.90 of MLB. I suspect the truth is they come in at both ends of this spectrum at different times in history. Did I mention that the Negro Leagues are complicated?
The Negro National League and Eastern Colored League of the 1920s were probably close to NPB level leagues. The talent was well concentrated in those leagues, and while the cream of the crop were Hall-level players, the very bottom end were probably Hi-A or Lo-A players. The spread of talent was larger than in MLB, but the cream got more playing time, and the really bad teams with mostly nobodies tended to play fewer games and/or fold quickly. Compare that to the early 1940s. At that time, the league’s biggest names jumped to Mexico and/or went to war. Pending further research, the combination of the two seems likely to me to have lowered the quality of play to AA quality. Once the color line was broken and the exodus of talent hastened, the quality of play sank rapidly.
On the flip side there’s the Mexican League. With so many black stars jumping to it, the league’s quality rose steeply in 1940 and 1941 and ebbed and flowed in the 1940s. It imported several quality MLB players in 1946–1947 before Happy Chandler started handing out suspensions for signing a contract with la Liga. Although this requires more investigation, we can make some initial guesses. Today’s Mexican League draws primarily from Mexico and surrounding countries. Despite its AAA classification, Clay Davenport’s studies show it’s at about a Rookie ball level. Mexico’s best have rarely proven to be superior quality major leaguers. Fernando Valenzuela being an exception that proves the rule. So, add to that Rookie-level league a couple dozen high-profile MLB stars and some veteran AA and AAA players, and what would happen? The ceiling would absolutely rise, but so would the floor because the lesser native players would garner fewer appearances. So the guess here is that the league of the 1940s rose to about an overall AA level. Maybe a tad more, maybe a tad less depending on how much talent it imported for any given season.
Now here’s a kicker. In some seasons, the Cuban Winter League might have been NPB level or better. The Cuban leagues only included three or four teams each season. Numerous Negro Leagues stars made the trip south (or returned to their homeland in Martin Dihigo’s case), their numbers were augmented by the very best Cuban and Latin American players, as well as occasional white minor league or major league players (especially native sons Dolf Luque, Armando Marsans, Rafael Almeida, and Mike Gonzalez). The number of players who appeared for a team fluctuated from small (15) to large (25), and the championships were hotly contested. Talent burst at the seams of the league, though, like the Negro Leagues, this may have been more true some years than others.
Now, finally, we arrive at the organized minor leagues themselves. They are of concern for players who transitioned into organized baseball during integration. If Negro Leagues expat Marv Williams hit .401 at the age of 32 with 45 homers in the 1952 Arizona-Texas League, what does that mean? His stats (with an extrapolation for his walks and other peripherals at known career rates) probably compute to a runs-created total around 150, very close to Trout’s 2016 total. The Arizona-Texas league was a C-level league. Consulting my chart above on the history of league classifications, Williams probably played in a context around Hi-A or Lo-A level by our current nomenclature. It might well mean that Williams’ performance translates to very near the major league average despite the gaudy numbers (especially because the AZTX league was a very high-octane loop with 7.1 runs/game. Yeah, you read that right, 7.1). And that makes sense, doesn’t it? If Lorenzo Cain played all of 2016 in the Midwest League, wouldn’t we expect him to destroy it just like Williams did the AZTX?
So at the very worst, bootstrapping from today’s minor league setup gives us a strong foundation to build conversion factors from. There are issues with it, though. Leagues, especially lower level leagues, from the Integration era were typically populated with older players than they are today. As much as two or three years older. That doesn’t mean, however, that they are as talented as today’s younger players. But it’s a thing. Also, with so many more and more localized leagues back then, we can’t say for sure that something like the Arizona-Texas league wasn’t worse (or better) than other leagues at the same classification. This is also true today in some measure, but the spread must have been much wider back then. Still, despite these issues, we can probably work with some confidence because baseball as a game hasn’t changed much. The minor leagues are minor for a reason, and the big leagues have always used them as a means for procuring and developing talent.
No one has ever said that the translation of Negro Leaguers stats into a major league context will provide highly accurate assessments of performance. Not possible given the limitations of the data. We would instead hope to achieve a reasonably accurate assessment. The definition of accurate remains open in this context, and details like the difference between A and AA ball require attention and a flexible concept of “correct.” But if we go down this path, we could only do our best to arrive an answer that passes the sniff test and doesn’t have any glaring mathematical errors.