"If someone has identical performances against two pitchers in a sample that small, it might be valid information or it might just be a coincidence that in future plate appearances will prove unreliable." -- from Part One of "All Fastball-Happy Families…"

When I noted the above in Part One of this series, I was being a little hyperbolic, or so I thought.  But before investing another week or so into drawing up the several hundred pitcher/batter matchups I promised you, I decided to try matching only a few selected pitchers and batters first.  Working with Bob Gibson and Don Drysdale as my archetypal dominating N.L right-handed power-pitchers whose repertoire, styles, and virtually identical peak periods replicated each other, it occurred to me that Bill has remarked somewhere that another pair of contemporaries of theirs seem also like a perfectly matched set: the Willies Stargell and McCovey, two lefthanded slugging OF/1B-men.  So how did these two pairs fare against each other?

Incredibly closely, is the short answer.

Would you believe, as Maxwell Smart used to ask in these players’ heyday, that against Bob Gibson both Stargell and McCovey batted an identical number of times?  And would you believe that they got an identical number of hits?

It’s true:  both Willies went 38-for-131 against Gibson.

And would you believe that this pattern repeated itself against Drysdale? Well, not with that kind of precision, but yes:  their lifetime OPSes against Drysdale were even higher than they were against Gibson, and remarkably close to each other:  1.108 and 1.117. The career of Stargell, who was the youngest of the four, didn’t align perfectly with the career of Drysdale, who was the oldest of the four. (Basically the first part of Stargell’s career matched up with the last part of Drysdale’s.)  But Stargell still had a significant number of appearances against Drysdale, 76.

I love coincidences, and I can’t get over both Willies going 38-for-131 against Gibson. OK, I’m over it.

The larger subject of coincidences has recently gotten my attention, though.  Everyone has them in his life, most of them trivial, so trivial they really can’t be called coincidences, but some are amazing. In my spare time, which means "all of it," I’ve jotted down 24 weird and unlikely things that have occurred so far during my years on earth (I’ll work on my Martian years at some later date), such as the only hitchhiker I’ve ever picked up in my entire life, on I-95 outside of Boston, hundreds of miles from home, in the pouring rain, turning out to have just finished taking a Master’s degree with my ex-girlfriend, whom I ended up marrying three years afterward.  I mean, he knew her personally, had a crush on her, was astonished that I knew her, etc. We discovered that commonality about two hours into our drive, somewhere around New London, Connecticut. Stuff like that. I came up with two dozen such improbable events, excluding Stargell’s and McCovey’s record facing Bob Gibson, and I’m sure you could come up with a few strange ones of your own once you set your mind to it. 

"Sheer coincidence" could explain why the Willies hit Drysdale so well, and so much better than they hit Gibson, or the issue could, even in these relatively large samples, be open for discussion. Of the four matchups, the one that seems the most significant is McCovey’s record facing Drysdale, in that it’s a fairly significant number of appearances and it’s the strongest performance of the four, so even if you found a number of uncounted appearances (say, Spring Training at-bats) between the two and they brought down that 1.117 OPS, McCovey’s numbers against Drysdale would still be very impressive overall. There’s no question but that McCovey hit Drysdale exceptionally well, but we’re not going to enter the issue of why that was, and certainly not before we ask whether these numbers are meaningful.

I don’t remember where Bill first commented on the similarity between the two Willies, or when, or in what context, so I obviously don’t remember fine details about the observation, but it’s either essential to the observation or a very strange coincidence that two Willies shared not only a first name, a pair of positions, an offensive game and a hat size, but also an OPS. Stargell and McCovey both had a lifetime OPS of .889.

I just looked at Bill’s 2008 article on the 96 Families of Hitters and found that Bill and I are defining "families of hitters" differently. Also, that I voiced the same main complaint about his system then that I feel now: Bill disregards handedness, which I think is key to any family of hitters. My next chief complaint would be that he is including players’ entire careers in his calculations, which I think is dangerous: a player who has tremendous speed when young, let’s say, and loses it traumatically in mid-career will have similar lifetime steals and triples to another player who has moderate speed for his entire career. He is using OPS in one very strict way: one of his components is to classify players by their OPS, so Stan Musial is a "A" player because he has an OPS of better than .900,  Stargell is a "B" because his is between .800 and . 899, and so on. The problem here is the problem with all cutoffs, that they essentially equate an .800 and an .899 OPS.

You can start by identifying other points of similarity, and then fine-tuning by using OPS-cutoffs, as Bill does, or you could begin with the OPS matches and then look for other things to fine-tune by, as I’m doing.  I think you can find truer "families" (at least in the sense that I mean families: players who resemble each other) by regarding their overall offensive effectiveness, which is what OPS is designed to measure. If someone can’t get on base but has power, and another player is the opposite to the same degree, they’re going to have roughly equal OPSes, meaning that they’re roughly equally effective.

We can then use our subjective memories to weed out the dissimilar family members. For example, Tom Haller and Lou Brock were contemporary NLers who had OPSes of .758 and .753, respectively, but we’ll regard that pairing as less than useful here, starting with the obvious difference being Brock moved as quick as a comet and Haller moved as quick as a continent.  They may have been equally effective, but they achieved that effectiveness in vastly different ways. By using contemporaries who batted from the same side in the same league at the same time, we can eliminate some, not all, of the potentially contaminating factors in our considerations and find batters who may (or may not) have faced the same pitchers with similar results.

In the comparison of Stargell to McCovey, I decided to abbreviate my first list of comparable pitchers by eliminating both Juan Marichal and Bob Bolin, because McCovey was a teammate of both for virtually all three players’ entire careers, but the other two pitchers on my list, Jim Maloney and Tony Cloninger, both yielded similar results to Drysdale and Gibson. If you think the two Willies hit the two Hall of Famers unusually well, you’ll be blown away by how they batted against Maloney and Cloninger.               

In other words, in all eight of these matchups so far, McCovey against each of the four and Stargell against each, none of which were selected knowing the outcomes in advance, the worst showing has been McCovey’s .823 against Gibson, and the average being over a 1.000 OPS against the four pitchers, all of whom were chosen for their overwhelming ability to strike out batters.  (BTW, the total of the eight matchups is 146 Ks in 836 PAs, by my count, which seems at a glance to be below these strikeout specialists’ K/PA ratio, especially considering the two Willies were themselves high K/PA guys.) All sorts of caveats are still in order, of course:

1.       836 PAs still isn’t an overwhelming lot, about 10% over the record for PAs in a full season, which Bill has assured us is not quite enough to form a stable body of results. You can have fluke seasons, and this isn’t much more than one full season.

2.       Over .100 OPS points improvement over Pops’ and Stretch’s lifetime .889 OPS figure isn’t impossible by chance

3.       The rest of the stats are still impressive, but the question of whether they’re overwhelmingly impressive is still open.

I should take a moment to discuss the adverbs and adjectives, like "overwhelmingly impressive," I’m using to modify these stats:  there are a few OPS figures coming that are identical, which is purely coincidental, since raw OPS is itself a cobbled-together stat (SLG + OBP) that isn’t rounded off (that is, the stat is often slightly imprecise, since SLG and OBP are usually irrational fractions that need rounding off, so when they’re added together you would expect the totals wouldn’t add up right all the time, but no: listed OPSes seem consistently to be simply the total of the two rounded-off fractions, and is thus approximate, on a small scale. If one guy has an OPS of .800 and another guy has a .799, the distinction is not only too small to be worth noting but truly meaningless, in that if we bothered to add up the fractions rather than the rounded-off fractions, the two might each add up to .800.) So when I’m saying that two OPS results are virtually identical, I mean that they’re very close.  In these sample sizes, just one single or one out, or a fraction of either, could make what appears to be a real difference in two performances. I don’t know what it would require, for example, for Stargell’s OPS against Drysdale, 1.108, to surpass McCovey’s OPS of 1.117, but it ain’t a lot. Something like a single walk would tie them, or even push Stargell a little higher, so I’m calling those two virtually identical, and any result that has two OPSes within .05 of each other "extremely close." OPSes within .100 (.800 and .900) are reasonably close, too close to make anything out of in these sample sizes, and .200 is about the point where we can consider a difference to begin to bear meaning. A .300 difference is clearly meaningful (that is, a batter who normally has an .850 OPS but has a .550 OPS against one particular pitcher is someone we can think of, even at this sample size, to be performing poorly against that pitcher) and OPSes of .400 or more are considered "significant." Not that this is important, but I don’t want you to think I’m fudging these numbers by referring to close and significant distinctions more loosely than I am.

Still, you would expect (I would expect) that any batters facing such quality pitching (Marichal, Gibson, Drysdale, Maloney, Cloninger and Bolin) would do worse, not better, than they did facing average pitching, so to me the only question is: have I generated a broad enough range of statistics that they approach significance?

Not yet, it seems, and rather than reach further in terms of pitchers, which would mean relaxing my standard for power-pitchers to study here (i.e. a lower K/IP ratio, or fewer IP), I’ll try to find other OPS twins against this same body of pitchers:

In casting around for another pair of twins from the 1960s, I first lit upon Vada Pinson and Billy Williams, which seemed promising in a variety of ways: both contemporary left-handed hitters mostly known as high-BA hitters but with considerable HR pop, just a full step down from the two Willies in power. But not, as it turned out, especially close in OPS, so I decided to just use that as my standard of comparability: OPS twins in the NL from the 1960s, and I found a few pairs to match against Drysdale and Gibson (and to the other four pitchers in my study).

I had to eliminate pairs who mostly played for the Dodgers or Cardinals, since they wouldn’t have a significant number of PAs against either Drysdale or Gibson. Tommy Davis and Donn Clendenon posted identical .771 OPSes in the NL, and Mike Shannon and Tommy Harper came very close, .698 and .695 but Davis’ years as a Dodger and Shannon’s as a Cardinal rendered those moot. Some of these OPS figures are a little off from their lifetime OPS marks since I decided to look for their OPSes only when they played in NL. Playing almost exclusively in the NL, of course, eliminates problems with facing different pitching, as similar birth-years eliminates era-bias. What I was looking for, primarily, was players who faced each other a lot, and batter and pitchers whose stats would be directly comparable.

The non-Dodger, non-Cardinal pairs I was left with were Aaron/Mays (.944/.941), Santo/Clemente (.838/.834), and Mazeroski/Taylor(.670/.667). Mays and Aaron have obvious similarities, as do Taylor and Mazeroski, at least positionally.  Santo and Clemente might seem a little off, but if you can imagine a situation where the Cubs had a logjam at 3b and the Pirates had had one at RF, the two might have been swapped even up for one another in a fair trade.  Santo had a little more consistent power, Clemente had a similar advantage in BA, and it all comes out in the wash.

For these, I’ll simplify and just print the OPSes, side-by-side:

We will dispense with demonstrating that these batters hit often against each pitcher (Mays and Aaron played the length of these NL opponents’ careers and never missed a game), with the exception of Bob Bolin, whom Mays never faced, Marichal, whom Mays rarely faced, and Tony Cloninger who spent a lot of years as Aaron’s Braves teammate. The results are nothing as eye-popping as the Stretch-and-Pops show, but still intriguing:

Bob Gibson pitched well against both Mays and Aaron, holding them well below their lifetime OPS marks, and Drysdale pitched adequately against them both, keeping them around their lifetime OPS mark, a few singles above Mays’ OPS of .941, a few singles below Aaron’s mark of .944.  In the only large inconsistency so far, Maloney "held" Aaron to .985 but really shut Mays down (5 hits in 58 ABs, with 1 HR) with a .583 OPS. If Maloney’s inconsistent performance against the two sluggers were closer to the norm, and Drysdale’s and Gibson’s consistency were the anomalies, I’d have to assess this experiment as a failure, but so far, I’m not there yet. Let’s try Santo/Clemente:

And Tony Taylor and Bill Mazeroski:

The one conclusion I’m able to take away here is that Mazeroski seems to have needed some more off-days facing right-handed power pitching—those are pretty uniformly awful numbers there, only once even approaching his lifetime .667 OPS. But comparing each batter to his twin, only a few times do the OPSes add up very well: Santo and Clemente did pretty similarly facing Drysdale, Gibson, Maloney and Marichal, and Taylor and Mazeroski likewise facing Drysdale, Gibson, and Marichal. Otherwise, there’s no real pattern forming here.

There are several ways to compare these figures, all of them interesting, none of them dispositive:  you could compare each pair of batters horizontally; i.e., how closely their OPSes against one pitcher match up. You could compare them vertically; i.e., how closely each batter did against the entire group. And you could compare them cumulatively; i.e., add up their stats to see how they did against each pitcher, compared to how they did lifetime against the entire NL.

If what we have is a complete mish-mosh of random numbers, then I’ve wasted your time-- or I’ve effectively shown that there is no such thing as a family of pitchers in any real sense that we can demonstrate, which would itself be a useful thing to show:  MarisFan61, in the "Comments" section of  Part One, linked to above, remarks of this

thing that is already very much done, only I think it's done mostly the old-fashioned way: by looking. Do you not think so? What makes me say that is simply that we hear about it all the time. I hear it so commonly -- it's so familiar ("he's sitting because he doesn't hit this type of pitcher very well" sometimes with an addition like "and he can use the day off anyway") -- that I couldn't even tell you if I've ever heard it from managers or coaches or just from announcers. I realize that what you're doing is looking for a systematic way of doing it, rather than just through observation -- but I think the idea is already in common practice.

If it’s in common practice, which is true, but isn’t supported by any empirical testing, then maybe it’s an example of practices that don’t make much sense?   In that same "Comments" section, Brock Hanke pointed out, that I may have misselected the families entirely—Marichal threw a curveball (I thought a screwball, too, and several other pitches), while Gibson’s secondary pitch was a slider, and Gibson’s fastball was more overpowering than Marichal’s, etc.  This is true, too, I have no doubt—every pitcher’s repertoire is bound to differ from everyone else’s—but if we’re going to discuss pitcher families and batter families, we need some objective method to determine who’s in the family and who’s out, with a certain degree of latitude for family membership (just like real-life families).

There’s also the problem of stat-contamination, as noted: Bill’s original article on pitcher-families put Fergie Jenkins, Catfish Hunter and Robin Roberts, as I recall, into one family, which was great for defining a family composed of moderately high K, very low BB, very high HR pitchers. (Have I got that right? Seems to me I’ve just defined Bert Blyleven.)  But Jenkins, Hunter, and Roberts all pitched to different groups of batters, in different eras, in different leagues for the most part, so it’s hard to test the theory by seeing how they did against the same batters in a decent sample size.

Does my thesis hold water? The thesis being: if you can define a family, and find some members who pitched to the same batters enough times, that family should show some similarity of results. If it doesn’t, if it’s just a jumble of stats, full of sound and fury, then there isn’t much to the concept of pitcher or batter families. So: do we find patterns emerging here?

I’m finding enough to keep me going.

Let’s try another pair of contemporary lefties from the same period: Bill White (.806 OPS) and Vada Pinson (.801 OPS).

Two of the outcomes with comparable pitchers are remarkably close, Drysdale and Marichal, and they are the pitchers whom both White and Pinson faced the most (within this group) by far.                White was Gibson’s teammate for most of their careers, so we can’t make too much of that one, but I’d have to say the vertical comparison shows White doing poorly against this family of pitchers, considering especially that it is all righties and he was a middle-of-the-order lefty power guy.  (I’d like to expand this one outward a little and see if it holds against other, more loosely related family members.) Pinson’s vertical  showing is much closer to the norm here—he hit close to, or over, his overall .801 OPS, and this is in line with what I’d expect a good left-handed batter to do.

Let’s try another, this time with two smallish outfielders who played on the same NL team throughout the 1960s, Johnny Callison and Tony Gonzalez, both debuting with the Phillies in 1960.  Callison was more of a power hitter, small as he was, and Gonzalez, even smaller, was better at getting on base, but their NL OPSes are similar, .787 for Johnny C. and .774 for Tony G. Since none of the pitchers played for the Phillies, we should have good sample sizes here:

Callison hit Bolin and Cloninger well, Gonzalez hit both of them fairly poorly—but the other four comparisons are pretty close. You’ll never beat the pair of .832s against Maloney. (I double-checked to see that I hadn’t copied the same stat twice by mistake, then checked again with Gonzalez vs. Drysdale.)     

How about comparing OPS twins Denis Menke (.713) and Ron Hunt (.715), signed a year apart by the Braves, born six months apart in adjoining states, both All-Star NL infielders who got to their virtually identical OPS marks by different routes, Menke by power, Hunt by getting plunked.

I’ve highlighted in yellow the two pairs of matchups here that are unusually close to each other. I think it’s a fair conclusion that neither Hunt nor Menke seemed to hit very well against right-handed power pitchers. There are no atrocious outcomes here (the lowest OPS is Hunt vs. Drysdale, .201 OPS points below his lifetime .715) but almost all are lower than their lifetime NL OPS marks (Menke about .640, Hunt about .620 by the all-seeing eyeball method, weighting approximately the number of PAs). This certainly seems about what you’d expect—they’d lose some OPS points due to the righty-righty matchup, and these pitchers are of course an especially intimidating group, so it’s not weird or anything that Hunt and Menke would lose close to .100 OPS points facing them. It seems pretty consistent at the same time:  their performances against Marichal and Maloney are eerily close, and their performances against Gibson are just .041 OPS points apart. Aside from Cloninger vs. Menke, which had only  21 PAs, as they were teammates for much of the 1960s, the total range is a nice tight .290 range, meaning that it only once (the difference between Hunt’s high OPS and his low) came close to my .300 standard of significant difference. Cloninger aside, Menke stayed within .597 and .733.

Instead of continuing these pairs of OPS twins, let’s try the opposite tack. If we can identify the pitchers who were the least like Drysdale, Gibson, Marichal, Maloney, Cloninger and Bolin, their opposite numbers, will these same batters show roughly similar results? I found a group of lefty pitchers in the 1960s who were low-K/IP guys: Cal Koonce, Al Jackson, Bob Hendley, Dick Ellsworth, and Claude Osteen. Let’s see the results of those lefty pitchers vs. the righty batters:

Compare this group of twins to the lefty twins’ performances vs. the same young, lefty NL low K/IP guys:

I’ve highlighted the two lists, lefties and righties, according to this somewhat arbitrary system: within each pair (McCovey-Stargell, Pinson-White, etc.), I’ve highlighted in yellow those OPSes closer than .100. and the OPSes greater than .300 in green, leaving the ones in between those two extremes in gray. If you just want the summary, of the 35 comparisons between "twins,"  there are 16 yellow instances of OPSes between OPS twins under .100,  9 gray instances of OPSes between .100 and .300 POS points apart, and 10 green instances of OPSes greater than .300 points apart.

I remind you, not for the first time, that ALL of these comparisons are tainted by small sample size (and not for the first time, that they’re the biggest sample sizes I can find), but I still find it fairly remarkable that so often these OPS "twins" replicate each other’s results against the same pitchers as closely as they do. Aaron and Mays, for example, against Koonce (.801/.804) are not only virtually identical, but I remind you those OPS marks are also quite different from (about 140 points below) their overall OPSes, which suggests that Koonce was pretty effective against both, as well as pretty similar.  (Koonce faced Aaron 34 times and Mays 41 times, putting them 24^th and 15^th on Koonce’s most-frequently faced list.) Now of course there are almost (but not quite) as many results that are wildly disparate among the twin-sets as are remarkably close, and of course I’m using gross and arbitrary approximations in setting standards for "close," "neutral," "quite different" and "very-far-apart" OPS results, but these are only preliminary findings, in which I’m just looking to see if this sort of study is even feasible and even worth doing more closely. So far, I must say I’m not completely discouraged. If I had found, for example, that the majority of OPS results were widely disparate for the majority of OPS twins, I’d probably have called this a poor study of ill-defined subject, and given it up weeks ago.

More of this stuff to come, though not as swiftly as you might like, as this is a busy part of my year and I’ve got some other BJOL columns in the works that I’d like to finish before I forget completely what I was trying to say there.