Can Managers Influence Pythagorean Over/Under Achievement
November 7, 2009 at 8:00 am by Capitol Avenue Club under Atlanta Braves
With the end of Bobby’s career looming and the discussion about possible replacements already taking place, I’ve been thinking about managers a lot, lately. What makes a good manager? I don’t know. People generally cite two things when debating whether or not a manager is a good one. 1) The opinions of players, ‘experts’, and various other baseball people. If the sabermetric community has learned anything, it’s that you should not believe something just because someone says so. Things we thought were right 10 years ago we now know are 100% wrong. How do we know these people aren’t making similar errors of judgment? Unless they present evidence, we don’t. So I don’t buy that. I don’t buy that players know what makes a good manager. Hell, players didn’t have a complete understanding as to what made a good player 40 years ago (some would argue they still don’t. I won’t). And I don’t buy that the same guys who, well, let’s just do the Poz quote:
Pitching was 90 percent of baseball because we said so. Managers needed to bunt more because we said so. Pitchers needed to go nine innings and pitch through pain because we said so. You judged a hitter on his batting average, a pitcher on his victories, a fielder on the number of errors he made, a player on his ability to perform when the chips are down — all because we said so. You know what? We were pretty stupid.
Yeah, those people. I don’t believe they know how to evaluate a manager better than anyone else with a simple understanding of the game. Because, like Poz said, for years they preached concepts as dogma without any evidence and they were wrong. It’s one thing to be misinformed and have a flawed understanding of something, to convey that misinformation as dogma on a massive scale is a whole ‘nother issue. It’s clear they were either a) in the dark or b) making things up.
My point is, research has proven the media wrong before. And I don’t see any reason why they wouldn’t do this again if we were able to properly study the question (which I’m not convinced we’re capable of at this time. Hit F/X will drastically improve our understanding of the game–I truly believe it, perhaps enabling us to study this question better.)
Maybe the media and players have it right. Maybe the media and players do know what makes a good manager. I see no reason why we should presently accept it if we don’t have any data to support it, though. To be able to say “I don’t know the answer to that” is empowering (and it also separates smart, competent people from stupid, ignorant people). That’s why my stance on what makes a good manager is, “I don’t know”. It will be so until I see some real, scientific work on the subject. If much attention is given to this subject and multiple powerful, scientific studies conclude that the media and players were right all the time, I will gladly tip my hat to them. By the same token, I won’t be attempting to rub their incorrectness in their face if they’re proven wrong. It’s not a personal thing. It’s not that I don’t trust the media and the players, it’s that they haven’t given me any reasons (data) to believe them.
So, for me, that’s not a very strong argument. Because Ken Rosenthal said so or because T. R. Sullivan said so or because Mark Teixeira’s kiss-ass says Joe Torre Girardi (brain fart) is the best manager he’s ever played for even after playing for Bobby Cox (real fuckin’ classy, Mark) isn’t a valid argument.
Which brings me to the second argument, 2) A manager’s Pythagorean over achievement. “X manager’s teams frequently over perform their Pythagorean expectations” is a line often uttered. On the surface I like this one more. I mean, Pythagorean record is largely a function of your players’ production (we’re assuming H1 = Managers do influence Pythagorean luck), but if you ‘get more actual wins out of your team than their Pythagorean record, it must be the manager’s doing’.
Again, is this true? Is a manager capable of getting his team to perform beyond their Pythagorean expectation?
I’ve designed a study that attempts to answer this question.
H0 = Managers do not influence Pythagorean luck.
H1 = Managers do influence Pythagorean luck.
I selected 20 managerial stints. These stints had certain properties. 1) At least 10 continuous years. 2) No player-managers. 3) No 1981 (the split-season). They are:
| Manager – Team | Years |
| Connie Mack – Philadelphia Athletics | 1901-1936 |
| John McGraw – NY Baseball Giants | 1907-1923 |
| Tony LaRussa – St. Louis Cardinals | 1996-2009 |
| Bobby Cox – Atlanta Braves | 1991-2009 |
| Joe Torre – NY Yankees | 1996-2007 |
| Walter Alston – Brooklyn/LA Dodgers | 1954-1975 |
| Lou Piniella – Seattle Mariners | 1993-2002 |
| Dusty Baker – San Francisco Giants | 1993-2002 |
| Jim Leyland – Pittsburgh Pirates | 1986-1996 |
| Bruce Bochy – San Diego Padres | 1995-2006 |
| Mike Scioscia – Anaheim/LA Angles | 2000-2009 |
| Tommy Lasorda – LA Dodgers | 1982-1995 |
| Earl Weaver – Baltimore Orioles | 1969-1980 |
| Red Schoendienst – St. Louis Cardinals | 1965-1975 |
| Tom Kelly – Minnesota Twins | 1987-2001 |
| Miller Huggins – NY Yankees | 1918-1928 |
| Wilbert Robinson – Brooklyn Robins | 1914-1931 |
| Joe McCarthy – NY Yankees | 1931-1945 |
| Casey Stengel – NY Yankees | 1949-1960 |
| Sparky Anderson – Detroit Tigers | 1982-1995 |
Take the 20 managers, write down their Pythagorean Over/Under for every year. For each manager, take their average Pythagorean Over/Under in the even years and the odd years. If managers could really influence Pythagorean Over Achievement, the even years and odd years should correlate. I then created a scatter plot using the 20 manager’s even/odd average splits as the 20 data points:
NotNot only do we see no clear pattern, the r (-0.245) and rsquared (0.060) values indicate that there’s isn’t much reason to believe that managers can have any influence on Pythagorean over/under achievement. The sample size (20) isn’t as big as I’d like, but I sacrificed some sample size to get the quietest data I possibly could. There isn’t enough here to draw any sweeping conclusions, but it’s a fairly good piece of evidence that managers have virtually zero impact on Pythagorean over/under achievement. And until I see a scientific study that concludes otherwise, that will be my position.
The data for this study, as always, is available to the public in OpenOffice format per request. To request the data, email PWHjort@capitolavenueclub.com with “Data Request” in the subject and which study you’re requesting in the body.
Loading ...
[...] Can Managers Influence Pythagorean Over/Under Achievement … [...]
first of all Tex said that about Girardi not Torre–but speaking of Joe Torre vis a vis this topic you have posted today. he once said that “baseball has a heartbeat” something you sabremetricians seem to ignore. not everything can be “proven” with piles of sometimes conflicting data.
aside from that mark teixeira seems about as stand up a guy as any one would want. i don’t know why you dislike him so intensely. it’s not his fault that your team traded for him then traded him away. is it?
I meant Girardi, not Torre. My bad. I’ll correct it next time I’m around a computer.
Interesting and insightful analysis here. Well done.
P. W.,
I am not sure how all this r squared stuff works (I do have a clue, but this is all, a clue). However, if I read your chart right, the chart does suggest something, thought the effect is possibly below a statistically acceptable threshhold.
The long tenured managers should represent a slightly better than average manager. Connie Mack owned the team, so he probably should be excluded. If the rest seemed incompetent to GM’s and owners, they wouldn’t have kept jobs for long stretches. It doesn’t mean they WERE better, but it would not be unlikely that they were slightly better than average.
Your chart shows the lowest performance in all grouped years of any single manager as a little below 2 games below pythag. However, at least 3 (of 20) plot points show + wins at 3 or more and 2 of those are 5 or more. That would seem to say that long tenured managers tend to outperform pythag slightly over long periods of time. 8 of the 20 were positive in both even years and in odd years. None were negative in both even years and in odd years. Only 4 managers experiencednegatives greater than 1 in EITHER axis. So, this group is not evidencing equal tendencies toward the plus and minus outcomes (nor, grouped as it was, should we expect it to). We would expect the grouping of all managers to show that somehow (maybe not in pythag, but where else?) the short tenured managers underperformed their counterparts.
Another way to look at the same issue is how often the plus pythags cluster toward the “expected managerial acheivement” years of 46 to 56 and how often the long tenured managers tended to start at lower pythags, peak in those years at plus pythags and finish at low pythags all over again.
Longer tenured managers are usually on teams that win a lot. Teams that win a lot tend to outperform their Pythagorean expectation. What this study shows is that managers over performing their Pythagorean expectation is more of a systematic phenomenon than an attribute (for the reasons I just mentioned).
Very interesting, this gives me hope that Bob Geren can’t screw up Oakland’s 2011 season! :D