Wins Shares, glorious Win Shares!!

[offshoot of the “Greg Maddux is the the Bestest Pitcher EVAH!” thread]

Rather than continue being my usual obnoxious self there, I’m continuing it here.
A more general discussion of Win Shares should be segregated anyway. I love 'em, think they’re a very useful to assess players, especially across their entire careers, across generations, leagues, etc… I realize that Bill James is still working out some kinks, but there are principles here that make them useful for certain discussions. RickJay and I probably need to settle some WS issues (I thin he misunderstands some things about WS, and I always learn lots from him) but right now John DiFool raises an issue here:

You’re mixing up Win Shares (which as I said James doesn’t award a single extra one of to Ted Williams, despite verbally acknowledging that Williams deserves all the credit in the world for serving in two wars) with his listing of the 100 best players at each position that appears in the New Historical Abstract. On the latter lists, James is NOT defending the ways he arrives at his rankings, and he certainly doesn’t justify them by win share ratings. He simply has a category, heading the quote from p. 344 above, called The Subjective Element, in which he adjusts the rankings according to his quirky decisions, (like listing between #17 and #18 at “shortstop” number xxx so he could work Nomar and Jeter into the book, among his other quirky choices.) He’s all over the map sometimes, it’s true, but in the main I think his position is mine: it’s better to rate, say, Williams with 555 WS as the second best LF of all time, behind Stan Musial (with 604 WS) with a note that Williams missed more of his career to fight wars, and probably would rate ahead of Musial if neither man had been drafted. (Though he did n’t choose that option in the New Abstract, as you correctly note: he just invented 50 Bullshit Points that let Williams edge him out.) His position is usually–hey, you get credit for what you do on the field, and that’s the right tack to take.

I’m not sure what you are arguing here, PRR. Note that James has always been of the position that the stats he generates are merely the starting point, not the ending point; I believe that he said (paraphrasing) that a bad statistical analyst takes the stats he sees/generates at face value, and leaves it at that, and a good one will then take the next step of asking himself how much these stats distort the true picture of what he is trying to measure, what has been left out, etc. The subjective element he uses as a final adjustment in his player rankings is an attempt to balance these more subjective factors which aren’t so easily quantifiable. Don’t mistake the numbers (be it Win Shares, VORP, or whatever) as representing some sort of hard eternal truth, nothing more needs to be said, and he most certainly did not invent Win Shares to be used like that. My position is that you need to carefully evaluate these more elusive factors, and if it is players from different eras that you are comparing, differences between the time periods in which they played must be considered, for fairness’s sake if nothing else.

Let’s talk about why he invented Win Shares, and what they’re meant to do.

He’s usually argued that RAW stats are the starting point, not anything necessarily meaningful in themselves. But the more refined stats (especially the ones he invents, and tinkers with, and is satisfied with) are an end point to some questions and I think WIn Shares qualifies.
The weirdest thing about Win Shares that James has to struggle against is why there are three of them for every win. People ask him all the time why doesn’t award them partially (i.e. give Schmidt 10 instead of 30, so the 1981 Phillies’ Win Shares total will add up to exactly the number of games the team actually won, instead of three times that figure) and he says: because that smaller number would be too inclusive. IOW, it would need to include partial integers to be accurate. James would have to list a fractional value with every number to give meaningful distinction, like listing Schmidt’s 1981 contribution at 9.83 instead of rounding it off to 10. Otherwise, he’s list a whole bunch of guys as “6” when in reality there are meaningful differences between 5.60 and 5.90 and 6.44. So James is creating additional problems for himself, and enduring those problems, precisely because there is a meaningful distinction between a 15 WS player and a 16 WS player.

Is that distinction SO meaningful as to be an absolute guarantee that every 30 WS player definitely has had a better year than every 29 WS player? No, James is unwilling to go that far, mostly because there are some tiny gray areas in the Win Shares determination values (the stuff he’s tinkering with over the years) to make it arguable that someone with a 29.51 WS value (Schmidt’s actual unrounded figure, btw) is clearly better than someone with a 29.49. So because of rare and tiny exceptions like that, he’s unwilling to issue a fiat declaring WS values to be unfailing correct, but he does NOT feel that they are unreliable, and neither do I, the vast majority of the time. If he did, he certainly would have placed them at the 1 WS = 1 game level and called them “Win Share Approximations.”

He actually did have something called “Approximated Value,” for a while in the very earliest stages of developing this idea. It was an arbitrary number, didn’t actually stand for a win, or any part of a win, but just estimated the value of a player’s year, as accurately as James could figure. So Yaz (his first example) could have his 1967 breakthrough season represented by a two-digit number (I forget what scale he used–let’s say Yaz’ first big year was 22) and all his previous seasons would have been in the low teens. This arbitary number was simply to allow whole seasons to be compared to wach other, quick and dirtyy style, so you needn’t get bogged down in “Yeah, But” argument (“Yeah, but he was hitting in Fenway, in 1967, in the AL, in a strong lineup” bbbyyyy) --you’d simply have a meaningful number that could be compared accurately to other meanngful stats. Over the years, as he refined it, the number became meaningful and useful. But accuracy has always been James’ goal, and he’s been refining that accuracy for a few decades now.

The best thing about WS numbers, to me, is that they’re objective summaries. That is, they can’t be criticized for favoring one player or one type of player over another. When we argue whether Maddux has a better peak than Koufax, we might still be defining some terms, such as what constitutes a “peak” but when we accept WS as a standard (which some people refuse to do, of course) we are accepting that James’ system does NOT favor pitchers from the 1960s or the the NL. If we waish to dispute WS numbers we must argue the components of how were derived, and James has set up his WS system (which is very complex but available for scrutiny) to evaluate thoroughly the stats and to correct for all sorts of influences on the raw numbers, so it’s really hard to argue for any bias, and impossible IMO to argue for any significant bias. Besides, James has no reason to skew the formulae in any way, since he uses them to compare players in future years when he might wish he had skewed them otherwise. They’re designed to be accurate, and they are for the most part.

What you and I were discussing was the subjective element, and I admit (as will Bill) that he vacillates on this. Sometimes he wants to credit players like Williams or Jackie Robinson with non-quantiable achievements, fighting wars, fighting racism, etc. and elevate them above what the numbers say. This is Bill being a fan, and a reasonable person, and a human above being a stat-head. But in the main, he’s far more comfortable saying the numbers are what they are, and you can offer mitigations for the numbers if you like, but you can;t change what they say.

Take Elston Howard, for example: one problem in giving Jackie Robinson extra points for missing years before breaking the color line, noble as that may be, is that you’d really have to then credit Elston Howard with extra points for missing a few years, too: Howard, who was in his late teens by the late 1940s, and so played his entire career with the color line broken, didn’t play a major league game until 1955, depite being older than Mickey Mantle who debuted in 1951. Was there racism involved here? Probably. But how much credit do we give to Howard to compensate for it, as long as we’re awarding points for the seasons that Jackie Robinson was ineligible to play MLB? I have no idea and neither does anyone else.

There’s a further problem, again using Howard as our example: unlike Mantle, Howard didn’t become a regular until he had been in the big leagues for five years, because he was playing behind Yogi Berra in his prime. This is analogous to players whose careers get thwarted because they’ve wrongly wasted years in the minors–do you want to award them extra points because this waste of their talents wasn’t under their own control? If you want to argue that Howard was capable of being a star catcher from the early 1950s (but racism prevented that) and that the time he spent on the Yankees’ bench and minor leagues prevented him from reaching capacity, I can;t argue but do you really want to conclude that Howard deserves bonus points for what he may welll have been able to do for an entire decade? If we’re award him a few hundred bonus points (which would be fair) for his missing decade, then Win Shares just becomes a meaningless stat in which every underutilized player, some for racist reasons, other because of the stupidity of management, and others just through poor luck (I’m sure the Yankees kept Howard on the bench for insurance if Berra went down, which Yogi wouldn’t do) is a cause to recalibrate on an individual basis, most of which will be subjective bullshit. Then we’re back where we started, you arguing “Yeah, but” this and me arguing “Yeah, but” that.

Leafing through the New Historical book (which I havent done for years, so thanks!) there are plenty of stories about “subjective” judgments: read the section on Morrie Rath, who played mostly minor league ball but starred whenever he got a chance to play in the majors, at least by modern statistical standards of stardom that were unrecognized in his own time. James loves pointing out such unherolded stars, but there’s always a counterargument, that he’s also glad to make: “Yeah but” the guy DIDN’T DO what you make an excellent case him being capable of doing. Are we really going to judge whether every player who couldn’t find playing time was deprived of his chance by foolish management (Rath), racism (Robinson), a glut of players ahead of him (Mike Easler), playing a position too difficult for him (Cliff Johnson), a war to fight (Williams), a family problem (Edd Roush), an injury that couldn’t be treated properly at the time, etc. Of course, there are combinations of all these, all perfectly subjective and all antithetical to the central principle of Win Shares, which is that performance can be quantified accurately. Not exactly, not beyond all discussion, but as a starting point for discussion that needs some thoughtful points to justify changing. If we were to add credit to Mike Easler for all the needless seasons in the minors because the Pirates just had too damned many good outfielders, why aren’t we taking some credit away from Willie Stargell and Manny Sanguillen and other star players? Better to raise the point that such players might have accrued more WS points if they’d gotten the chance than to monkey with the ratings of what they actually did on the field of play.

I’m not exactly sure who you’re arguing with, PRR. Bill James?

Well, there seems to be some disagreement over how to (or if to) credit players for stuff that doesn’t show up in Win Shares, particularly as regards missed games for injuries, military service, racial barriers, misjudgment by managers, changed use standards across eras, etc., some of which I felt you were making a case for, no?

Umm… “yes.”

Or possibly “no.”

Every case is different. There are no hard and fast rules.

“Greatness” in a player can itself be subjectively defined. Who’s a greater pitcher, Sandy Koufax or Don Sutton? I’m sure most people, including you, would say Koufax, but Sutton of course had many more Win Shares, the main reason for that being that Koufax got injured and Sutton did not. Any argument that Koufax is greater must rely on the assumption that his greatness is defined by his incredible peak run, as opposed to Sutton’s consistent pretty-goodness. I personally don’t think Koufax deserves “Credit” for getting injured; staying healthy is a part of an athlete’s skill set. But if you were to argue Koufax was greater because of his huge peak value, that’s a perfectly valid argument.

Do some players deserve “credit” for being barred by racial segregation? I think in some cases that’s really obvious, such as Satchel Paige; the only reason we don’t have him at 550 Win Shares or whatever it would have been is that he was playing baseball in leagues that didn’t leave behind accurate stats. I think it would be disgraceful to leave Satchel Paige out of the Hall of Fame, or deny that he was a superior pitcher, just because you have numbers for the AL and NL and not for the NNL. But in the case of Elston Howard I’m skeptical; the colour line was broken before he started, he was after all 26 when he broke in, didn’t actually get stuck on the bench as long as people say he did, had a reasonably long career for a catcher, and as a hitter he wasn’t THAT fantastic.

In every case the question of a player’s greatness or quality is a function of the objective evidence and an honest assessment of context. Greg Maddux getting 30 Win Shares in 1995 (if I recall the total correctly) is just a fact. You can’t pretend it was 33, or 28, or 31, but you can make an honest argument about what it means - whether it’s superior to 30 Win Shares in the unintegrated 1927 NL, or to wartimes win shares, or inferior to win shares gathered in the slightly tougher 1992 NL, or what have you.

Ted Williams got 555 Win Shares. There’s no point in pretending he should have 625 because of his wartime service, but it’s perfectly reasonable to conclude he was actually a greater player than his 555 Win Shares would suggest (not that he needs the help) because of his being a Marine in two wars. It’s perfectly reasonable to say Jackie Robinson was a greater player than his win shares initially suggest because of his place in history, and to say Elston Howard’s place in history is not really equivalent. You may not agree, but I think the argument can be made. These aren’t hard and fast answers, though, and I think each case deserves its own examination and an honest argument.

I agree with almost every point, except I’d like to discuss the WS a bit further, and I think your summary of Howard’s career is a little dismissive. His stats are here and

  1. I chose him BECAUSE Jackie Robinson’s mitigating circumstances (of the color line wiping out some of his best years) don’t apply, so I don’t see why you’re raising it as if to contradict me. I want to discuss OTHER circumstances that can muddy the discussion, and show that it’s best to set all of them aside and talk about what someone actually did on the field. Stronger cases can be made for white players, such as Hank Sauer or Morrie Rath.

  2. He did break in at age 26. That’s very late. As with Vic Power, who was earlier, some teams, such as the Yankees, were using any excuse to keep good young black players down in the minors for obscenely long times (though in their defense, they also did the same to some white players, like Bill Skowron, so this may be more poor judgment than simple racism.) In any case the fact remains that he got a very late start. Compared to Mantle, who was two and a half years younger and got started four seasons sooner, Howard lost seven baseball seasons, By the time Mantle was Howard’s age at the start of the 1955 season, 26 and a few months, he had won two MVP awards and had hit over 200 HRs. Not to say Howard could have done the same, of course, but he did get off to a late start.

  3. Howard played in the late 1950s, but not very much. He averaged 108 games per season in his first six years, batting over 400 times only once, in 1959. He got his first 500+ AB season in 1964 when he was 32. By the point Mantle-- who walked a lot of course and was injured some years-- was 32 , he had averaged about 500 at-bats per year for thirteen seasons. I’d say Howard missed a few at-bats there, wouldn’t you?

  4. Catchers don’t have to have the same offensive numbers as other players do, as you know, and his stats would look much better with another 1000 at-bats, or even 1500 more (that figure, while unlikely, is only about 100 more at-bats per year to his MLB career–it assumes he wouldn’t even break in to the majors until the age he did actually break in.) He did stay healthy very late, but he was playing ball from ages 19-25, just in the minors, so there’s no reason to think he couldn’t have gotten some portion of the well-over-3000 at bats Mantle had gotten by age 26.

In any case, I DON’T want to argue that Howard is anything more than he was, just to point out that plenty of players have good cases for being more than their numbers tell us. Nevertheless, I’d be reluctant to aggrandize any player’s career beyond his actual performance, precisely because mitigating circumstances can be made for many players, most of which should be met with a shrug and “That’s a shame.” Howard could have been a HOFer. He wasn’t.

Didn;t get around to discussing WS further there, nor why I place such emphasis on peak, but we can talk about those things some other time, unless you would like to go there now.

pss, I’m just not really all that clear on what you’re trying to say. What exactly is the thesis you’re defending?

In post 8, or generally.? In post 8 I’m taking minor issue with your skeptical characterization of Howard’s career–I’m arguing that he got shorted of many opportunies and as a result his career numbers are way short of what he showed he was capable of achieving.

Generally, though, I’m agreeing with most of your points in post #7–while intriguing, such speculations are just that, and I don’t want to credit (other than verbally acknowledging the speculations) what players like Howard, Jackie Robinson, Ted Williams, and probably the majority of MLB players could claim–that they would have enjoyed better careers under optimal conditions. Since we can never know what would have happened in an alternate universe in which such conditions apply, I’m far more comfortable arguing strictly what players did achieve under the circumstances they actually played in.

This is not to say that my rankings of players’ greatness is simply a listing of career Win Shares (or any other stat) because we place different emphasis on various stats’ importance.
I stress peak stats, especially in discussing HOF considerations, because of the importance of each season as a separate unit. The goal of each season is a World’s Championship, so seasons in which certain players make major contributions towards that focused goal get a lot of importance in my view. That is, a player who has win shares of 10-10-30-15-10 (total=75) is better than another, far more consistent player whose WS numbers are 15-15-15-15-15 (also totalling 75). The first would get some consideration in an HOF discussion (he’d be a ‘No," of course, but we’d have to think about that one “30” Win Shares year, while the other guy is just a flat-out "No,’ with no argument at all, no matter how long his string of 15 Win Shares went on.) There are those who’d argue for the greatness of an ungodly streak of 15- or 20- type seasons, but I’m not one of those.

How many shares would a guy like Herb Score get. He was looking like a phenom until Gil McDougals line drive ended his career. You could extrapolate him into anything you wanted but the sad fact is he was just a blazing fire that went out too soon.
Williams very well could have beaten Ruths 714. But he did not. He went to the service and that is that.

Score had 44 Win Shares his first two seasons, then 14 in the rest of his career. Tom Seaver, to give you a neat comparison, also had 44 Win Shares his first two years. He went on to get another 344 Win Shares over the rest of his career.

Sure, I’ll buy that.

Well, let me ask you this; was Satchel Paige a great player?

Paige has neither career nor peak Win Shares (or any other stats) that would rank him among the greats. An argument that Paige is a Hall of Famer must proceed from the assumption that he was great even in absence of supporting statistics.

I don’t believe there are any realistic Hall of Fame candidates who just rolled up twenty years’ worth of 18-share seasons. Don Sutton’s about as close as you’ll get to that. Players who last a really long time almost invariably peak higher than that, and guys who didn’t have exceptional peaks, like Buddy Bell or what have you, don’t get seriously talked about as Hall of Famers. And this certainkly has very little to do with Greg Maddux, who had one hell of a run of great seasons.

I agree peak dominance must count for something, but who ever said it didn’t? Even James ran a simulation - in fact, I think Sutton and Steve Carlton were his models - and found players with higher peaks but equivalent career value will help their teams win slightly more pennants (it wasn’t a huge difference, but it was there.)

Of course. But since there’s very little reliable documention to support that, Satchel’s greatness is literally legendary. A classic case of “Not represented by stats” greatness. Of course, if I tell you that Johnny Aleo, who lived across the street from me in Brooklyn, was a far greater hitter than any professional hitter, you may just roll your eyes and says “Sez who?” So I’ll only argue for Satchel’s so far. If someone wants to exclude Satchel from the discussion of great pitchers, I’d have to say he has a basis for the exlusion, though I might disagreewith his conclusion.

Maddux was genuinely great for many years, and no question but that’s he’s going in the HOF and deservedly so, but the HOF does have some compilers, guys who put up decent numbers (but no great numbers) for a long time. Billy Williams is marginal like that, some defensive wizards like Aparicio and Mazeroski, too. Good long careers, consistent as hell, excellent in some ways, but not the total contributions that I expect from HOFers.

I don’t think you’ll get much argument that the Hall has made some poorly considered choices. Billy Williams might not be a great example, though; his numbers in his peak run were pretty darned good for the 1960s. It’s interesting to compare him to Jim Rice, who some consider a HoF candidate; Williams’s peak was just as good, he played much longer, and he was a better fielder.

I enjoy focused discussions like Williams vs. Rice, since they tend to stay on track.

Wms Rice
33 36
30 28
29 26

Hit submit too soon. Here are their top five WS years.

Wms Rice
33 36
30 28
29 26
28 24
28 20

Looks like a wash to me, small edge to Williams after the third year, but of course Rice isn’t in the Hall.

I’d like to work on a WS method for HOF status, which would involve multiplying the top year in WS by a certain amount, the next top-year by a lesser amount, etc., which would stress the element I find lacking in simple compiled stats.

To be simplistic, say you multilplied the top year by 5, the next year by 4, the next by 3, the next by 2 and every succeeding year would be face value (this is just for HOF voting, understand)–you might get results that I felt balanced peak and career length. With Rice that would be 180+106+78+48+20, which would =432 and Williams would be 165+120+87+56+28 which would = 456, plus all of Williams’ subsequent points. Both look like near-misses to me, especially considering Rice’s post-sason contributions and WIlliams’ lack thereof. Rice was fairly decent defensively early in his career in Fenway.

I’ve got no idea about about Williams vs. Rice insofar as real world stats or value, but according to these numbers, I’d have to give a small but definite edge to Williams, considering that his 4th and 5th best seasons are equivalent (insofar as WinShares are accurate) to Rice’s 2nd best season.

Your idea for weighting seasons is interesting, but I think you’d have to take larger samples to weight, say, top 3 (or 4 or 5) seasons are grouped (averaged?) and weighted, second 3 (or 4 or 5) are grouped (or averaged) and weighted…Not trying to be contrary here (it seems plain you place more stock in brief excellence than I do), just saying I think taking a person’s (possibly anomalous) pinnacle and further weighting it would lead to problematic distortions of overall worth. Perhaps it wouldn’t effect Hall-of-Famer’s and near Hall-of-Famer’s numbers so much, but I think it would muddy the waters for the top 25% tier.

The explanation for 3X Winshares given upthread seems wrongheaded to me, but I’m uninitiated. I’m getting James’ book for Christmas (even if I have to buy it myself), so maybe I can come back with a better idea of the value of WinShares as a metric in the future.

And RickJay, thanks for the lead in the ‘Maddux is Christ’ thread.

Maybe the wiki explanation is clearer than mine.