Japanese HOF Project

[nat]

Sadaharu Oh

1. Preliminary Remarks

There are some figures that are so humongous that they are hard to comprehend. There are 225,622 miles between here and the moon. The Pacific Ocean weighs 1.483×10^21 pounds. That sort of thing. Here is another: Sadaharu Oh hit 868 homeruns.

Eight hundred and sixty-eight.

There are a number of ways you can try to make sense of this record. One is to ask what a similar record in the US would look like. Another, and one that American baseball fans inevitably ask when they start thinking about Oh, is to ask how many home runs he would have hit if he had played in the US. The “what if he had played in the US?” question is partly asked out of sheer parochialism, and part of it is a disbelief that someone could accomplish a feat that seems to be impossible in the American game, but part of it is hoping that there is some way to comprehend this record, something that would make it intelligible. Like when Bill Nye builds a scale model of the solar system.

This is my second post about Oh. You can read the first one here. There’s been enough written about Oh, including by Oh himself, that you don’t need me to tell much of his story. So in this post I’m going to play with some numbers. In particular, I’m going to take a swing (if you’ll forgive the baseball metaphor) at answering the questions posed in the last paragraph. The first question—what would a similar US record look like?—is relatively straightforward. And I’ll answer that one in a minute. The other—what would he have done if he had played in the US?—is utterly unanswerable. A bit later I’ll explain why, and then I’ll make an effort at answering some other related questions.


2. Oh’s production in an American context

Oh played from 1959 to 1980. He spent his entire career with the Yomiuri Giants. I am going to use the most straightforward way that I can think of to translate his production into an American context. Basically, what I’m going to do is compare Oh’s homerun rate to league average, and then use the resulting figure to calculate what a comparable homerun rate would be in the target league (I’m going to use the NL from 1959 to 1980). Then I’ll adjust Oh’s plate appearances by the difference in league schedules, and multiply the adjusted homerun rate by this number of plate appearances.

The math is just arithmetic, but I’ll explain each step and then provide the formula. Since we want to know how much better Oh was at hitting homeruns than his contemporaries, we need to know how good they were at hitting homeruns. So the first step is to divide the number of plate appearances that batters in the Central League recorded by the number of homeruns that they hit. That gives us:

(LgPA/LgHR)

We’re going to want to compare this number to Oh’s homerun rate, so we then divide Oh’s PA by his HRs. And we get the comparison between Oh and the league by dividing the league homerun rate by Oh’s. So far we’ve got:

(LgPA/LgHR)/(Oh’s PA/Oh’s HR)

So, for example, in 1964 the Central League hit a homerun every 38.6 PA, and Oh hit one every 10.8 PA. Divide the first number by the second number, and you find that Oh was about 3.5x more efficient at hitting homeruns than was the 1964 Central League. Now, we are interested in translating Oh’s performance into MLB context, so we need to calculate MLB homerun rates:

MLB_PA/MLB_HR

Then you divide the latter number by the former number. This tells you how many plate appearances an MLB player would go between home runs, if he was as far above league average at hitting home runs as Oh was:

(MLB_PA / MLB_HR) / ((LgPA/LgHR) / (Oh’s PA/Oh’s HR))

In 1964 National League players hit a home run about once ever 50.5 plate appearances. If you divide that by 3.5 you get 14.42. So, to be as far above NL average as Oh was above CL average in 1964, you would need to hit a homerun once every 14.42 plate appearances.

That gives us a rate stat, but we want a counting stat: a number of home runs. So we need to do three things. We need to adjust Oh’s plate appearances by the difference in the schedule length between MLB and Japan. We then need to adjust this number by the percentage of league games that Oh actually appeared in. Finally, we divide this adjusted number of plate appearances by the rate that we calculated above. The number of games the Central League played varied by year, but was usually 130. MLB season length switched in 1962 from 154 games to 162, so there will be some differences in the calculations from year-to-year, but the general idea is this:

(Oh’s PA)*((#NL games/#CL games)*(Oh’s games played/#CL games))

Let’s go through this part slowly. “Oh’s PA” is just his actual number of plate appearances. “#NL games” is the number of games scheduled in the National League, either 154 or 162 depending on the year. “#CL games” is the number of games in the Central League, either 130 or 140. Dividing NL games by CL games tells us how much longer, in percentage terms, the NL season is. We’re going to multiply Oh’s actual PA in order to account for the fact that an American version of Oh would have played in more games, but he doesn’t get credit for the entire difference in the schedule, because Oh missed some games here and there. That’s what “(Oh’s games played/#CL games)” is all about. It tells us what percentage of Central League games Oh actually played in. By multiplying the (percentage) difference in the league schedules by the percentage of games that Oh actually played in, we ensure that our Oh clone doesn’t get too much credit for playing in a league with a longer season.

So, this whole figure gives us a number of MLB-adjusted plate appearances for our Oh clone. If we divide this number by the average number of plate appearances per homerun (calculated above), we get our MLB translation for Oh’s homerun production. Here’s the final formula:

((Oh’s PA)*((#NL games/#CL games)*(Oh’s games played/#CL games))) / ((MLB_PA / MLB_HR) / ((LgPA/LgHR) / (Oh’s PA/Oh’s HR)))

Feel free to check my math, but I think that works. I did this calculation for each season of Oh’s career, and then summed the number of homeruns hit in each season (rounded to the nearest whole, it looks odd to have him hitting 45.2 homeruns). Here’s the table:

Year HRs
1959 9
1960 27
1961 25
1962 65
1963 45
1964 56
1965 62
1966 65
1967 43
1968 33
1969 40
1970 52
1971 37
1972 43
1973 56
1974 36
1975 26
1976 26
1977 38
1978 27
1979 23
1980 21
Total 855

In a curious twist, the differences in season length almost completely make up for differences in homerun rates. Translated into the National League, 868 homeruns in the 1959-1980 Central League is… 855 homeruns. Yowza. The top figure is 65, a number achieved in both 1962 and 1966. (Years in which Oh actually hit 38 and 47 homeruns.) Ten of these figures would have led the league: 1962-1967, 1970, and 1972-1974.

(Incidentally, does anyone know how to insert a table into a Net54 post? I tried HTML, and it accepted the code but yielded weird results. Obvious possibility is that I was doing it wrong, but any advice is appreciated.)


3. The counterfactual question

It’s important to see that the foregoing does not tell you what Oh himself would have done if he had played in the National League. It tells you what a player who was as much better than NL average as Oh was better than CL average would have done. If the counterfactual question ‘how many home runs would Oh have hit if he had played in the National League?’ is taken literally, I don’t think that there’s any way to answer it. A 19 year old Sadaharu Oh might have gotten homesick during spring training, gone home, and joined Yomiuri. Or he might not have been able to keep up with a 162 game schedule and succumbed to chronic injuries. Or he might have been just fine. There’s no way to tell.

So the counterfactual question is unanswerable. We can, however, answer another question that might be found in the vicinity of that one. To explain that question, I think it would be best if I were to talk about MLEs for a while.

‘MLE’ stands for Major League Equivalence. You use MLEs to evaluate minor league performance. The idea is that you’ve got Joe McMinorLeaguer and you want to know what to expect from him. You’ve got his minor league numbers, but they were put up against minor league pitching, in minor league ballparks, and so on. And it’s hard to know what they tell you about his potential major league performance. So what you do is you find a bunch of players who have appeared in the same minor league as Joe and also appeared in the major leagues, and you see how strong the correlation was between their minor league performance and their major league performance. (It’s obviously a lot better if they played in the majors during the same year that they played in Joe’s league.) You then assume that Joe’s numbers would translate as well as this comparison class, and you adjust his minor league numbers accordingly. Those major league equivalences are not what Joe would have done if he had been in the major leagues—that’s unknowable—but they do give you some idea about what he would have done, and they can be fed into a projection system with some degree of confidence that the projection it will give you isn’t just nonsense.

The crucial bit for my purposes is that you can do this with Japanese statistics too. The guy to look to here is Clay Davenport, co-founder of Baseball Prospectus and guru of baseball statistics. His Davenport Translations give us just what we want. He’s got two sets of translations, a normal one and a ‘peak’ translation. I’m not clear on the differences between the two models, although it is obvious that the latter is more forgiving for hitters. Whether this is due to different comparison classes, or regressing numbers to the mean less (or more) aggressively, or what, I don’t know. In any case, here are the Davenport translations for Oh’s home runs:

Normal Translation

Year HRs
1959 3
1960 9
1961 7
1962 19
1963 18
1964 25
1965 22
1966 23
1967 20
1968 20
1969 18
1970 24
1971 19
1972 23
1973 26
1974 21
1975 16
1976 19
1977 19
1978 16
1979 13
1980 13
Total 393


Peak Translation

Year HRs
1959 5
1960 11
1961 8
1962 21
1963 19
1964 25
1965 22
1966 23
1967 20
1968 20
1969 18
1970 24
1971 19
1972 24
1973 27
1974 23
1975 18
1976 22
1977 24
1978 22
1979 19
1980 22
Total 436

Both versions of the Davenport Translation see Oh as a mid-range slugger in MLB. Perhaps his biggest strength was his consistency, so if an MLB team saw fit to keep a player with 20ish HR power at 1B, it’s not unreasonable to think that he could have had a very long career. Davenport uses the 1992 American League as his target league, so the statistics in the above tables are translated into that context.

However, the Davenport Translations don't account for differences in the league schedule. Once you work that in, the picture changes:


Peak

Year HRs
1959 5
1960 14
1961 10
1962 27
1963 25
1964 34
1965 28
1966 28
1967 25
1968 25
1969 22
1970 30
1971 24
1972 30
1973 34
1974 29
1975 22
1976 26
1977 30
1978 27
1979 22
1980 27
Total 544

That total would place him 17th all-time, just between David Ortiz and Mike Schmidt. Under the normal translation, adjusted for league schedule, he ends up with 486 home runs, good for 30th all time, above Adrian Beltre and Miguel Cabrera, and below Lou Gehrig and Fred McGriff.


4. The card

This is a post to a baseball card website, and so is ostensibly about a baseball card. So here’s a baseball card. It’s a 1977 Calbee. In 1977 Calbee released many sets; at least, Engel catalogues them as separate although closely related sets. (How they were actually distributed I don’t know.) I have been unable to determine which of the many ’77 Calbee sets this one belongs to. There is one promising candidate (although I don’t have my copy of Engel handy and I don’t remember which one it is), but the book says that the back of this particular set is framed by ‘weeds’—a leaf motif that turns up both on Calbee cards and on some menko sets. And my card doesn’t have a frame around the text on the back. So I can’t say any more than that it is a ’77 Calbee.

Meikyukai: Yes – Hall of Fame: Yes