Dr LIC's Krazy SyEnce Korner Again
More NUMBERS for all y'all. This time courtesy of the brightest minds at Northwestern's Kellogg Business School (some of these guys are friends of mine, which I should both admit and apologize to them for potentially butchering my interpretation of their results below...anyway, on with it!)
In a paper currently under review titled, "Meritocracy, Cooperation, and Performance," Adam Galinsky and colleagues analyze the performance of a sampling of 282 professional NBA teams from 1997 to 2008 and demonstrate a simple formula:
meritocracy leads to cooperation, which in turn improves performance.
What is particularly interesting, however, is how meritocracy is measured, which is by two variables: pay dispersion, starting dispersion, and participation differentiation. Pay dispersion is the extent to which there exists a range of salaries on a given team such that some players are very highly paid and some are very minimally paid (pay dispersion would be low on a team with relatively "flat" salaries--where everyone is pretty much being paid the same). The authors measure this using the standard deviation of player salaries on each team. Starting dispersion, which really constituted a measure of how much status was dispersed across players, was measured by standard deviation of starting positions for each team during a given season. Participation differentiation was measured by the standard deviation of playing time for each team during a given season. High levels of pay dispersion, starting dispersion, and participation differentiation meant that teams were operating more in terms of a meritocracy than in some sort of pseudo-socialist all-for-one system.
Cooperation--intragroup cooperation--was measured (and feel free to debate this in the comments) using a combined measure of assists, defensive rebounds, and field goal percentage. And performance was simply measured as winning percentage.
The takeaway is that using some seriously gritty statistical modeling, the data supported their assumption. Teams that are more meritocratic--those that have greater disparities between stars and nonstars--show more cooperation, and in turn win more. Pay dispersion was the variable that overarchingly accounted for this effect.
What is impressive about these studies is that they use field data to corroborate previous experimental and cross-sectional work in the social sciences showing the benefits of hierarchy on cooperation (i.e. "know your role" is a good strategy for promoting teamwork). What's more, I think these data say something counterintuitive about NBA conventional wisdom, which is that having max salaries on the tab can be a good thing if it establishes a clear pecking order...and the very teams that show the most cooperation don't do so because they are all equal serfs, but because there is deference to the gawd at the top of the roster.
Commenters, what say you?
Labels: syence
posted by Dr. Lawyer IndianChief @ 7/13/2010 07:30:00 PM
19 comments
19 Comments:
Paraphrasing Joe Posnanski, who was "quoting" a Sparky Anderson speech to the 1975 Cincinnati Reds in spring training, it's better to simply acknowledge that there are stars and to treat them accordingly. If any of the "turds" (Sparky's word) want special treatment, play like a superstar. Don't BS grown men. Udonis Haslem knows he's not an All Star, but he also knows that he's potentially one of the support figures that makes All Stars into Hall of Famers. Off the top of my head, I'd guess that this is also very true in the NFL, where teamwork is also necessary, and it has no relationship to MLB.
-- David A., Rufus on Fire
so the nordic countries won't be taking over the world anytime soon?
thinking out loud....players need a certain numbers of consistent minutes to get in the groove and play at their best (say 35 minutes per night). this means you need five backups who are happy to make positive contributions for 13 minutes per night. people who view themselves better than this limited role are not a good fit, people who are happy with that role help the team excel.
but i question whether these results can really be applied to the 'real world' outside hoops.... the CBA is an artificial distortion.....max contracts, salary cap & luxury tax don't exist on the outside.
So if they just measured using the standard deviation of salaries, the strategy of "have 3 max players and bunch of nobodies" could still result in more wins. If I understand correctly, it's only when there are NO stars or NO nobodies that cooperation is low. So this is an explanation of how Adam Morrison came to have a ring?
some partially formed skepticism:
i wonder if this indicative of association rather than causation, with the association specifically being talent with wins.
ie:
if your team lacks talent (ie players who command big salaries) you salary cap will be distributed amongst more mid level players.
if your team has a lot of mediocre players, you will likely need to distribute playing time more evenly amongst the squad.
if your team doesn't have the talent to win, over the course of a season you'll likely have to try different lineups.
so perhaps simply having talent means you're more likely to win. and having talent means you're more likely to have an unequal payroll, unequal playing time and a regular starting lineup, creating the association between disparity and wins.
This seems suspiciously obvious. So much so that it seems like it may not be causative, and could be irrelevant.
To paraphrase one of the other commenters, this is basically saying that if my team is lucky enough to have a super duper star(s), I'm probably going to win. We know that's true because basketball is a 5 player game and having one ridiculously good player can net you 66 and 64 wins, respectively, by having two of the greatest statistical seasons in the last 30 years.
But is it a true statement to say that having a bunch of guys at the $8-$10m per year range is causative towards your team sucking? I would argue that no, that's not the case. That the causative element is having a superstar, and maybe it doesn't matter as much what everybody else is paid, but its likely that everybody else could be paid much, much less.
Ya, I worry Brill/Lux have a point. Without having seen the paper, it looks like what this is saying is having two (or 3) max contracts (which will necessarily mean filling out the roster with low-paid guys because of the soft cap, giving you an uneven salary dispersion) helps you win. I suppose there could be something interesting going on in the correlation between having the stars and the increased cooperation, but I'd need to think more about the factors they chose and whether or not that actually indicated increased cooperation. I think I'd possibly want to see something more along the lines of Assist ratio, maybe even raw # of passes, I'm not sure. But it makes total sense that a team without 5 players good enough to permanently cement themselves in the starting lineup is probably not very good.
This seems like a basketball specific conclusion that can't be extended to other sports. Sports that don't have a salary cap (i'm thinking specifically about soccer here, but baseball works too) are dominated largely by the teams with money to pay multiple superstars a ton. This seems to be a function or those specific sports, though, since it seems like in soccer players generally have more well defined positional roles, and in baseball I don't even really know what "cooperation" would mean between players. There's also the problem of normalization in sports with no salary cap since team's payrolls can vary wildly. The conclusion here seems sort of obvious for basketball because the ideal makeup of a team with a salary cap, historically, has one or two stars and a bunch of players who don't care if the stars take all the shots. I wonder if the conclusions made about the standard deviations of salaries on a team are true of the standard deviations of ppg for all the players. Like, if you have one or two dominant scorers and the rest of your team defers to those guys, are you more likely to be successful? I remember reading somewhere that salry is strongly correlated with ppg in the nba.
I agree with everyone else; having an ultimate player (or two, sometimes three) is what matters, and this is what their dispersions and differentials are really measuring. If they have not already done so, they need to figure out a way to get around this.
The Ewing Effect, Bill Simmons' (very unscientific) observation that the departure of alpha dogs often increases efficiency would seem to be a major counterexample. Given my similar experiences watching the NBA, I am inclined to believe that something has gone wrong with this paper.
wv: nonfan
First, I agree with other folks that the study seems to confirm that having a superstar or two leads to wins, but I have to ask why that would necessarily render it faulty or completely useless? The conclusions themselves may not blow anybody's mind, but that by itself doesn't strip the study of all value. People like making decisions using empirical data, so if the conclusions in the study are solid, then however obvious they may seem, they may convince a coach stuck with a team of "equal serfs" to attempt to establish some sort of artificial "meritocracy." If he simply picks certain players to play more than others and runs more plays for specific players, maybe this "meritocracy" will lead to "cooperation," and result in more victories then if he tried to share playing time. (If the players all relatively equal it can't hurt to try. It may irritate the players not chosen but, as I explain below, "cooperation" isn't an accurate label anyway.). If successful, this will skew the results in any future study because salary dispersion will be lower (perhaps "meritocracy" will need to be measured differently in the future as a result), but that doesn't change the fact that in the present day the study can't be useful. Other possibilities include a GM using it to convince the owner to sign off on a max contract for a top free agent, or a GM working for an owner unwilling to budge may be able to use the compiled data to make the most of the situation and still construct a decent team (think "moneyball" and Daryl Morey). Who knows, I may be stretching with these examples and I haven't seen the study, but I definitely don't think its seemingly obvious conclusions automatically render it worthless.
Second, I initially thought that cooperation was being measured incorrectly, but then I came to think that "cooperation" is just a strange label to apply to the statistical categories they selected. That is, I'm still not sure how field goal % and defensive rebounds infer cooperation, I'd like to hear the explanations for this, but if a meritocracy (genuine or otherwise) necessarily leads to more assists, more defensive rebounds, and a higher a field goal %, then whatever you want to call that group of statistics, I guess you will win if you perform well in those areas.
This comment has been removed by the author.
Roadkiller has a good point: This study would be interesting only if it was based on some natural experiment where teams of equal talent distribution were alternately selected for "equal distribution of salary and minutes" or "unequal distribution of salary and minutes". I wonder if the authors have included some measure of talent as an explanatory variable, such as average draft position, or st.dev. of draft position.
on a macro level, i completely agree with this. if you have a set two or three man nucleus that your team revolves around -- whether psychologically, strategically,or whatever -- it more or less forces everyone to play their part. for example, if you're looking at any of phil jackson's teams the first thing that pops out is the hierarchy: jordan, pippen, grant/rodman, and then everyone else. having the two piranhas to center the team enabled not only toni kukoc to thrive in a lesser role, but also let luminaries like bobby hanson, judd buechler, and randy brown to play important bit parts.
however, i think any attempt at consciously achieving a high dispersion rates may be short-sighted. any success is potentially more indicative of lucky breaks than anything else. for every jordan/kobe/lebron-led team, there are the kg-wolves, soon to be joe john-hawks, and isaiah's knicks; you pay too much to one or two guys and you're completely dependent on not only those individuals' success and health (or lack of it), but also draft positioning and selection to maintain quality.
the key point is that a meritocracy is something that puts the power in those that have DEMONSTRATED ability, talent, and/or intelligence. too often the nba in its current manifestation is too quick to reward potential and this is where a gm could get into serious trouble.
but hey, i'm only a kg fan from mpls scarred by hundred million dollar contracts and secret mid-level exceptions.
Hasn't it been shown that max players (and not the Joe Johnson type) are often a "good value" in basketball terms alone? Since there is an individual salary cap, it's possible to overpay role players and potential stars and grizzled vets, but you cannot overpay Lebron James. If your biggest contract is a good value than it gives you room for the occasional mistake and overpayment with your smaller contracts.
I don't know if this contradicts your buddies' study, but I think you need an approach that makes a determination between pecking order and value.
The thing that jumps out at me, which may be better explained in the paper, is the implicit assumption that salary is a reliable indicator of merit. I'm not saying, however, that anything in the model is wrong; rather, I think I'm saying that they've chosen the wrong word. Particularly so because "meritocracy" is such a loaded term. Perhaps they should pull back a little and instead discuss "income distribution, role definition, cooperation, and performance." Or if they haven't already, they could demonstrate that salary is a reliable indicator of merit, eliminating an otherwise unproven assumption.
Couldn't this study just be interpreted as saying those teams that can afford the best players end up winning? (With the cap, signing a max contract means you have to sign some min contract, creating pay dispersion)
Doesn't look good for the Suns, then.
Thanks for the introduction to the meritocracy and I didn't know more about it. It makes things more interesting when you are learn something more about it. It would make a great post at Sportsbook Applications
Post a Comment
<< Home