Monday, June 2, 2014

Statistics, Racism, and Stereotyping

This article was originally published in ESPN the Mag in late March. It's an amazing conglomeration of foolishness. It was written by Jeff Phillips, a principal at The Parthenon Group, which is some kind of business consulting firm, so perhaps this isn't exactly sportswriting. But if ESPN the Mag is going to publish it, it qualifies.

FOR 150 YEARS, "clubhouse chemistry" has been impossible to quantify.

Have people been trying to quantify it for 150 years? I don't think there was anyone back in the early days of the American League trying to figure out some way to quantify the clubhouse chemistry. Plus, it would've been totally different back then. All the players were white guys from the good old USA, so the model would probably have to measure things like "willingness to pound moonshine or "ability to cover for me when I need an alibi for the manager's bed checks" or "sleeping ability on a train". I imagine that old time GMs probably had different models for chemistry than their current counterparts.


The only way to measure it, winning teams said, was by gut feeling: Either you have it, the way the Red Sox did with the Idiots of 2004 and 
the Beards of 2013, or you don't.

It's a good thing all the other baseball teams have the RED SOX around to show them examples of good chemistry. God knows the RED SOX are the most charismatic organization playing America's most charismatic sport. This is what makes everyone hate Boston's sports team.

I also find it interesting that the authors don't present the scores of the '04 or '13 Red Sox on their scale. This would've been a really good place to show that their super-duper chemistry metric matches the eye test of the obviously chemically superb Red Sox. But there's no mention.

Until now. 

EMPHATIC SENTENCE. Jeff Phillips is following the advice of his 10th grade English teacher, Mrs. Higgenbotham, who reminded him to use syntax to reflect meaning. Way to go, Mrs. Higgenbotham!

Working with group dynamics experts Katerina Bezrukova, an assistant professor at Santa Clara, and Chester Spell, an associate professor at Rutgers, we built 
a proprietary team-chemistry regression model.
Jeff Phillips probably pitched his article something like this:

Jeff Phillips: We've got an educated baseball audience hungry for some new insights on baseball, and they seem to love statistics. But we also need to appeal to the good old boys who understand clubhouse chemistry. So we're trying to figure out how to quantify this. This is gonna be a killer article!

Katerina Berzrukova: Listen, I'm busy working on getting tenure, and working with you clowns isn't exactly going to boost my academic resume. Plus I am Russian. Call me if you want to quantify the chemistry on your gymnastics team.

Chester Spell: Yeah, publishing in ESPN the Mag is not exactly a peer-reviewed journal. If I'm trying to build my resume, I probably don't want to get published by the same organization that has employed both John Kruk and Joe Morgan as a baseball analyst.

JP: Wait, did I mention I'm writing an article for the World Wide Leader in Sports? The Biggest Sports Journalism Outlet around? You can get a lot of hits on our website.

[KB looks at CS]

CS: So?

KB: Yeah, so? Nobody reads it in Russia.

JP: Also we can pay you handsomely because our coffers are overflowing with money. I'll just grab a few grand from the Bristol Money Bin, where Simmons and the ESPN execs just swim in cash like Uncle Scrooge.

KB: Now that you mention it, I've got a PhD from Moscow State University, where I helped the vaunted Moscow State Dragoons hockey team run off a four-year undefeated streak by concocting a clubhouse chemistry model based on their loyalty to the motherland and their vodka tolerance. So this stuff is my wheelhouse.

CS: Well, I do a lot of research about organizations and businesses, and Major League Baseball is an organization and a business!

JP: Great! I'll throw some bills at you, you throw some random numbers and regressionspeak back at me, we bang out a few thousand words for the Mag, and we call it a day. Whew! I'm gonna hit the pool.

*************

Plus the model is proprietary. Which basically means that article could've been random numbers thrown out of nowhere. That might work if I thought ESPN had the ethos of being honest and trustworthy and devoted to making a contribution to how we understand the complex game of baseball... hahaha, I kill me.

Our algorithm combines three factors -- clubhouse demographics, trait isolation and stratification of performance to pay -- to discover how well MLB teams concoct positive chemistry.
It's a good thing that human relationships in the complex world of a baseball clubhouse can be reduced to three indicators. And it I'm s
till not buying the premise that MLB teams are trying to "concoct" positive chemistry. Sure, I bet now and again a team makes a decision perhaps not entirely based on baseball performance (the Reds did spend a fair amount of coin for an aging Scott Rolen in 2009 probably in part because of his clubhouse presence), but that particular acquisition wouldn't have done much at all for these threefactors. 

Also, the word choice here is putrid. "Algorithm"? "Concoct"? C'mon, Mrs. Higgenbotham! I know you wanted to teach Jeff as much vocabulary as you possibly could, but you have to also teach him how to use it appropriately!

Plus this model doesn't quantify the major factor that would be conducive to positive chemistry: not signing total assholes like Milton Bradley, John Rocker, or Albert Belle When these guys come up with a model that precisely quantifies being as asshole, I'll start listening.

According to the regression model, teams that maximize these factors can produce a four-win swing during a season.
So by overhauling their entire 25-man roster in order to manipulate the overall chemistry of the team, they might be able to get four wins out of it. Sounds like a plan! Actually, maybe they should manipulate the overall quality of their baseball players instead.

According to our projections, such bonding will clinch this year's World Series: The Rays (third in chem) will prevail over the Cardinals (28th) in six games for Tampa Bay's first Series title. Happy players, happy fans.
At least the Cardinals suck at chemistry. I've heard all sorts of bullshit about the Cardinal Way over the years, so I'm glad to see that the BFIB won't latch on to this as evidence of their moral and personal superiority to the rest of the National League. But this model is still stupid. And if chemistry accounts for 4 wins over 162 games, its effect in a 7 game series must be miniscule. So I don't think it makes any sense at all to over-reach and use it to predict the winners of the World Series.

Chemistry breakdown
Demographic factor: The impact from diversity, measured by age, tenure with the team, nationality, race and position. Teams with the highest scores have several overlapping groups based on shared traits and experiences.


Does this mean teams should try to avoid racial integration? Or including foreigners? Or only bring them in if they're in groups? This is such an awful way to look at any kind of group construction - instead of actually working with your players, teaching them not to be assholes to people from other races and countries, and generally working to improve the clubhouse chemistry of the best players they can assemble, it suggests instead that GMs should view their human resources as fixed identities. What a terrible way to view human beings, and what a terrible way to generate organizational culture.

Also, why the hell is position in this factor? Are we really looking at the chemistry impact of the few positional decisions a team makes? OH SHIT IF WE CARRY A THIRD CATCHER THAT TIPS OUR CHEMISTRY FACTOR. CMON UP CORKY!

Isolation factor: The impact from players who are isolated because of a lack of subgroups from these shared demographic traits. Too much diversity can, in fact, produce clubhouse isolation for players who don't have teammates with similar backgrounds or experiences.

If they're assholes, yes, then you'll have a team full of players who don't want to talk to their teammates because they have different backgrounds. God forbid we have TOO MUCH DIVERSITY. I guess that's why the 1927 Yankees were so awesome. They didn't have too much diversity.

Maybe they're thinking of a team like the 1975 Reds - three Latinos, three black guys and too white guys - as their ideal team. That way, every player has his own buddy within his own racial group. I'd be interested to see how they treat race, for example. Perez was Cuban, Concepcion Venezuelan, and Geronimo Dominican; would they they have meaningful shared traits? As a Reds fan I am ashamed that I don't know the answer to that question. As a baseball fan I'm interested to see how they actually categorize demographic traits. Oh wait, this is an article written to appeal to the hordes of fans who just want to throw numbers around without actually thinking about how they were constructed. Shame.

Ego factor: The impact from individuals' differences in performance and monetary status. Too few All-Stars and highly paid players signal a lack of leadership; too many, however, creates conflict. The ideal level falls in the middle.

Actually, too few All-Stars signals a lack of good players. Too many All-Stars signals, generally, a good team. Leaving out the obvious point about All-Star selection being a popularity contest, it's still a stupid-ass thing to say that the IDEAL LEVEL of All Stars is "in the middle". Maybe in Russia an All-Star means something else.

Chris W. once reminded me that Bill James once said that if you find a stat that says some nobody was a better hitter than Babe Ruth, you should probably find a different stat. In this case, if you find a model that says a team should not try to have a maximum number of all star players, you should probably find a different model. Or at least a different proxy variable for "performance". All-Star appearances is a a terrible variable proxy for performance. The current leader in the AL shortstop voting is Derek Jeter, who has the 12th best WAR among AL shortstops.You'd think that with the tons of possible ways to quantify performance , the professors would pick something sensible.

Their point about the possible chemistry effects of income inequality in the clubhouse seems relevant. I can see how that might cause problems in a clubhouse. I'd actually like to see the information on that. BUT NO: the model is proprietary.

The article goes on to discuss every division and how chemistry will affect the standings. I'm not going to go through every division because it's boring, but here's a sample:

The Tigers' clubhouse diversity contributes 0.9 of a win by itself, but the strength of its subgroups is weak due to differences in age and years with the team, leading to an Isolation score of minus-1.6 wins. The Royals' Isolation score, on the other hand, is 4.7 wins higher, boosted by a pitching staff that includes eight Americans among the team's 10 core pitchers.

This is pure junk. We're supposed to draw a meaningful conclusion from the fact that the Royals, are doing their best to keep American jobs at home instead of outsourcing them like those anti-American Pirates? Or that Dave Dombrowski's careful assembling of Tigers' 25-man clubhouse diversity is almost worth one win? It's just so stupid.

Man, that article was bad. Faux academic knowledge? Statistics misused? Terrible writing? Subtly promoting racial stereotyping? Nice work, ESPN The Mag.

4 comments:

CHart said...

Wait, you're not LarryB...

dan-bob said...

Back in the old days this here blog was about more than Bill Simmons.

Jack M said...

I'd just like to point out that at the time of posting this, the soon to be World Series Champion Rays are currently the only team in the Majors with a sub-.400 winning percentage.

Derpsauce said...

Jack -- I think I can ease your concern there. The model probably didn't take into account that the Rays would have Kevin Kiermaier on their roster for this much of the season, and that guy is a complete asshole.

You know, I mean, probably.