Dick Allen was known for being an unpleasant teammate who warred with teammates and divided clubhouses. In "The Politics of Glory," Bill James called him "a manipulator of extraordinary skill," and wrote,
"[Allen] did more to *keep* his teams from winning than anybody else who ever played major league baseball."
That led me to wonder: how would that happen? What would make teams less likely to win with Allen in the dugout? I guess, maybe, if the players don't trust each other, and team morale disappears, the players won't care as much about winning. They won't try as hard, or do the little things they'd normally do. Maybe they wouldn't stay in shape, or they wouldn't study the opposition pitchers and hitters as intently, or they wouldn't be as receptive to coaching. Stuff like that.
In other words: with Dick Allen on the team, their individual performances would be worse than you'd expect otherwise. Can we find evidence of that in the statistics?
Back in 2005, I created a little algorithm to determine if a player had a "lucky" year at the plate. Basically, I looked at his stats two years before, and two years after, and took a weighted average of those four years. Then, I regressed to the mean a little bit, and I figured, that's what the guy "should have" done. Any deviation from that, I attributed to random variation. (I've attached an description at the bottom of this post.)
In my original study, I treated all the discrepancies as luck. Of course, if a player does worse than he "should have," it might also be due to other factors, like Dick Allen.
Of course, you probably wouldn't see Allen's entire influence: if he played with the same players for many years, they'd be consistently worse than they could have been, and the algorithm would have no way of seeing that. But enough players come and go that maybe we could see at least some effect.
So, for each of Dick Allen's seasons, I looked at the luck numbers for all the batters on his teams. (I used only batters, in part because, for some reason, my database was very slow processing pitchers). I omitted players who spent time, that year, on more than one team.
The result: Over 15 seasons, the batters on Dick Allen's teams were 125 runs unlucky—about 8.3 runs per season, or 5/6 of a win.
Of course, that doesn't really mean anything—small sample size, and all that. I think if you did a significance test, you'd find that "-8.3 runs" isn't even a single SD from zero.
But maybe we can try a larger sample, of a larger group of "controversial" players.
This web page lists the 15 "meanest players" in baseball. Ten of them were position players. I ran the same "Dick Allen" test for all ten. (My database includes estimates only up to the 2009 season.)
The results... pretty much completely random. Five had lucky teammates, overall, and five had unlucky teammates.
If you care, Mark Teixeira was the "worst" at -30.9 runs per season, while Prince Fielder was "best" at 21.7.
Then, I found a list of the "nicest" players in baseball (who, strangely, are all position players). I ran the test for those, too. I expected the same, random, non-result. I was surprised.
Of the fifteen nicest guys in baseball, thirteen of them had unlucky teammates. That is, in fifteen tests, "unlucky" had a winning record of 13-2. The probability of something that extreme happening by chance is about 1-in-271.
Here, let me list all fifteen players:
- -14.0 Thome
- -29.0 Ibanez
- -13.1 Damon
- -26.8 Mauer
- -11.1 Granderson
- - 7.9 Jeter
- +17.4 Pujols
- -23.5 Hudson O.
- -20.1 Hunter Torii
- -16.7 Pena Carlos
- +28.1 McCann
- -30.9 Teixeira
- -11.2 Giambi Jason
- -21.7 Young Michael
- - 6.0 Holliday Matt
Strangely, Pujols and Teixeira also made the "mean" list as well as this "nice" list. If you take them both out, you're left with 12 out of 13. That's a 1-in-585 shot.
Could it actually be the case that nice guys make their teammates worse? I suppose maybe "nice guys" means players who are less intense, and that makes the clubhouse a bit too laid back?
Or, it could be that cause-and-effect are backwards. Maybe when you play for a lot of teams that disappoint, and you take it well, you get a reputation for niceness. But I dunno, the average is really only around one win a year. It seems unlikely that would be it.
It seems more likely that it might be some unknown third factor, that explains both (a) players being nice, and (b) their teams being unlucky. Maybe, for instance, nice guys play for nice managers, and it's nice managers that are causing the underachievement? Or something like that. Is there anything in that list those players have in common that could be the answer?
Of course, it could be just random, despite the 1-in-271 odds.
Let's try something else. If there's something real going on, and it actually is common for players to influence their teammates for good or bad, then that's probably some kind of "leadership" characteristic we're looking at. And, you'd think, players who are good leaders are more likely to go on to become managers. So, you'd expect that major-league managers should have been more likely to have "lucky" teams during their playing careers.
And, again, yup, there seems to be something there. As of August 29, twenty-three of the 30 MLB managers had major-league careers as position players. (Only two managers were major-league pitchers, and the remaining five never made it to the bigs.) 15 of the 23 were "positive luck". As a won-lost record, that's 15-8. That probability is about 1-in-9.5.
Here they are. (I've included the number of the player's seasons that were considered.)
- + 8.9 Sandberg (16)
- + 8.7 Ventura (15)
- - 6.0 Mattingly (14)
- +15.2 Baker (19)
- + 9.0 Gibson (17)
- + 7.2 Scioscia (13)
- +17.2 Johnson Dave (12)
- + 1.3 Weiss (14)
- + 5.7 Matheny (13)
- +12.2 Girardi (15)
- - 1.1 Redmond (12)
- -15.7 Hurdle (10)
- +12.5 Bochy (9)
- + 5.5 Roenicke (7)
- + 8.2 Sveum (11)
- - 8.0 Melvin (9)
- -25.5 Washington (10)
- - 7.2 Gardenhire (5)
- - 4.3 Francona (10)
- -39.5 Wedge (4)
- +16.8 Yost (6)
- +53.9 Gibbons (only 2)
- + 7.6 Porter Bo (3)
The managers are in the same order as in this link, which is from most to least illustrious career. The negatives seem to be concentrated near the bottom, so you might think that maybe it's just that good players have more pluses than bad players. But, no,when I looked at everyone, not just managers, there was no such effect.
Also, the "positive" managers seem to have had more years in the majors. But again, I checked, and again there's no general effect like that. Players with longer careers are just as likely to have had unlucky teammates than players with shorter careers.
But, maybe—maybe—all else being equal, you're more likely to be considered for a managerial job if you played on winning teams. Teams with good luck are more likely to have won. So, maybe there's that kind of selective sampling effect going on here.
Or, maybe it's just random.
Finally, I checked an arbitrary bunch of other players I thought of who maybe had negative reputations for something or other. Those results were back to random:
- + 0.4 Reggie Jackson
- - 8.3 Dick Allen
- - 6.9 Rick Cerone
- +20.1 Barry Bonds
- -25.7 John Mayberry
- +11.9 Garry Templeton
- +13.6 Willie Stargell
- - 3.4 Thurman Munson
- - 6.5 Jose Canseco
So, there you have it. Taken at face value, it seems that clubhouse cancers don't seem to affect their teammates. But, nice guys make their fellow batters worse. And, having a future manager on the team makes them better.
I'm not really sure I want to take it at face value, though. Still, I have no idea what's really going on.
Here's the explanation I promised, for how the "luck" is calculated. As an example, I'll estimate Tommy Herr for 1985.
In the four years surrounding 1985, Herr's offensive WARs were:
2.4, 2.1, , 2.4, 1.3
What number would make 1985 fit right in? Maybe, 2.4 or so?
Well, Herr actually had a 6.0 offensive WAR in 1985. That makes him 3.6 wins "lucky," which is 36 runs. My algorithm actually comes up with a luck estimate of 32.7 runs (I didn't use WAR, and my algorithm is a bit more complicated). But that's roughly how it works.
The idea is, it's a way of doing roughly the same thing you'd do if you looked at the record by eye.
And, by the way, this is not exactly the same algorithm as I used in the past, but I'm not sure what the difference is, unfortunately. But the results are very similar to the old ones.
For more details, go to my website and search for "1994 Expos".
Originally published on Sabermetric Research. Reprinted with permission.