Man Bites Blog

The original Gladwell David and Goliath has a bit of a moralistic tone:

But let’s remember who made that rule: Goliath. And let’s remember why Goliath made that rule: when the world has to play on Goliath’s terms, Goliath wins.

The games are ultimately fixed.   If David comes up with a novel strategy for Goliath to win, Goliath will change the rules to exert his dominance.

This is unfair, Gladwell leads us to believe.  After all, all David wants to do is substitute effort for talent.  Shouldn’t that be celebrated?

I’m not so sure.

You see, I kinda like basketball.  And part of what I like about basketball is that it provides a wonderful forum for talented athletes like LeBron James and Kobe Bryant to display their skills.  Watching them curse in frustration while a bunch of hungry players with floorburns on their knees presses their teams into futility might be fun once.  The Pistons victory over the Lakers in 2004 was enjoyable.  But I want to see these guys driving toward the basket, making great shots, and exhibiting their talent.

If a sport makes has a wide path where an underdog can compensate for a talent deficit with effort and discipline, that sport is doomed.  Want proof?  See NHL, 1995-2000,the New Jersey Devils, Florida Panthers, and the neutral zone trap.  David beating Goliath makes for great stories, and great sports movies, but ultimately boring sports.

The N.C. State and Villanova victories are part of basketball lore, but the NCAA realized that less talented teams being able to beat more talented teams by taking the air out of the ball was a bug, not a feature, and put in the shot clock.  Not because it needed to consolidate Goliath’s power, but because standing around dribbling down the clock isn’t the game of basketball.  And sinking your own ships isn’t sea battle.

The same goes for “unpredictable”  postseasons.  The baseball postseason in the wild card era is essentially determined by luck and if you have a dominant starting pitcher.  If you have Cole Hamels, Chris Carpenter, Josh Beckett, or Curt Schilling, you can ride them to a World Series victory, even with average talent around them.  This does not seem to have resulted in a surge in popularity for baseball.

Obviously there are  are some insurgent innovations that should be embraced.  And the games will continue to evolve in part based on these.  But games need to strike a balance between being open to novel tactics, and remaining true to what the sport is, and allowing the best players to be the stars.   And I think this balance should be tilted more toward the latter than the former.

Malcolm Gladwell was back exchanging e-mails with Bill Simmons reiterating his hypothesis that the full court press offers and underdog the best chance at victory.  For football, he proffered a no-huddle offense.

I’m unconvinced.

It seems that if your team faces a talent disparity, what you want to do is lower the sample size.  The lower the sample size, the greater the chance for an anomalous result (i.e. your team winning).  In basketball, this means running down the shot clock to minimize the total number of possessions.  In football, this would mean running the football to keep the clock moving and minimize the number of snaps.  The greater the sample size, the greater the chance that Goliath’s innate advantage will assert itself.  The lower the sample size, the more likely strange things will happen.  On any given Sunday, any team can beat any other team, and this is particularly true if the Sunday you’re given contains a small number of plays.

If you think back over the most notable underdog victories — N.C. State over Houston, Villanova over Georgetown, Giants over Patriots, Giants over Bills, Patriots over Rams,  Hickory over South Bend Central, that is the path to success.  It is often recalled that Villanova shot 78.6% that night, but that was 22-29, and 9-10 in the second half.  If you double the sample size, that percentage is much less likely.  By lowering the sample size, Villanova put this anomalous result in play.  I can’t think of upsets of a similar scale driven by hurry-up tactics.

Now, you still need to win those possessions, and Gladwell would argue that speed-up tactics like the full court press and no-huddle offense may result in more data points, but each one is skewed to the underdog’s advantage.  But in my opinion, unless the tactic targets a specific weakness of the dominant team, such as only having one ball handler, or a pass rusher who’s out of shape, this is more than offset by the expansion of the sample size.

I think Gladwell’s definition of “underdog” is a bit broader than mine, such that it includes this year’s Louisville team, which entered the NCAA tournament #1 overall.

But I think that definition is so broad as to be meaningless.  Gladwell defends including the 1996 Kentucky team on the basis that it had only one NBA star, Antoine Walker.  But it’s hard to think of may college teams that had multiple NBA stars.  Phi Slamma Jamma and the early 1980’s North Carolina team are the only ones that sping to mind.  Chris Webber is the only member of the Fab Five to really go on to stardom.  The great UNLV team of the early 1990’s produced Larry Johnson.  Duke is famous for the star-crossed history of its graduates.  How about that North Carolina team that won the 2005 championship?  If the 1996 Kentucky Wildcats, with Rick Pitino as coach, can be considered “underdogs,” so can anybody.  The closest may be the 2006-7 Florida team coached by Pitino protege’ Billy Donovan.

Rick Pitino has his choice of what kind of players he want to recruit.  He chooses to go with players who are more coachable than talented, and that’s his prerogative.  Mike Krzyzewski goes after a different type of player, as does Jim Calhoun and Bill Self and Roy Williams and Tom Izzo (there’s a guy who wins with less talent!).  Each has had success, but doesn’t make their approach prescriptive for everybody.

Gladwell may be working on different time horizons than a single game.  Yes, Louisville may have entered the tournament #1, but they only had #30 or so talent.  Ok.  But I think what this demonstrates is that these tactics only take you so far.  The high school team in Gladwell’s article lost the championship game.  The article blames the refs; I think they just hit their ceiling.

The odd part is that Gladwell should have known this.  He wrote an article last fall about how difficult it was to tell if Missouri quarterback Chase Daniel would be an effective NFL quarterback because he played the spread offense at Missouri, and Gladwell recognized he wouldn’t be able to do that in the NFL.   Indeed, Missouri couldn’t win with it once they started playing the more talented teams in the Big 12 South.   The Missouri basketball team found out the same thing when they ran into a talented, disciplined team in Connecticut.

“Let’s lower the sample size, so we have a chance of winning” may not be a very inspirational rallying cry, but it gives the underdog the best shot.

There’s been a lot of discussion about Malcolm Gladwell’s article drawing a parallel between the difficulty in identifying good NFL quarterbacks and the difficulty in identifying good teachers.  Most has been about the validity of the particular parallel.

Closer to my heart, I think there is a very strong parallel to identifying good software engineers.

The quarterback narrative looks at QB Chase Daniel, who has been very successful in the “spread” offense at Missouri.  Does that mean he will be a good NFL quarterback?  Maybe, maybe not.  Daniel is fiercely competitive, and makes good decisions.  But the conventional wisdom is that the “spread” offense wouldn’t work in the NFL.  Would Daniel thrive in a more conventional offense?  It’s hard to tell, and QB’s taken with early draft picks have a success rate around 50%.

Coaches put in sytems like a “spread” offense in part to mitigate the unevenness of talent at the college level.  You can’t count on having an elite quarterback, especially at a school like Missouri that doesn’t have the same recruiting pull as a school like USC or Oklahoma.  Better to have a system that is not dependent on having elite talent, or is based on talent that is currently undervalued.

This puts a definite floor on how bad the team can be, but it also puts a ceiling on how good it can be.  Mizzou beats up on the Baylors and Iowa States of the world, but they get blown out by Oklahoma and Texas, who have NFL-type talent on defense.

This doesn’t meant that Daniel lacks talent, or that anyone can be a “spread” quarterback.   But the system limits what Daniel can exhibit.

We’ve tried to do something similar with computer science.  We were faced with a problem — fewer graduates than industry was demanding.  People found programming intimidating, were scared off by the possiblity of offshore outsourcing, and were not pursuing computer science degrees.   And those that did were from a narrow demographic — very few non-Asian minorities or women.

So we tried to make things simple for the students, similar to how coaches tried to make the offense simple for the players and the quarterback.  Out went C and C++, in came Visual Basic and Java.  Students could get their first program up and running in a matter of minutes.  No need to write your own data collections, memory management, or user interface — the API will take care of that for you.

Then these students came to work, and in a somewhat famous article, Joel Spolsky articulated similar frustrations to what NFL scouts are articulating about quarterbacks.  Not only did these students not have skills that he, as a hiring manager was looking for, they had not even demonstrated the ability to learn these skills.  Could this candidate who’s only developed in Java ever “get” recursion and pointers?   You’d have better luck trying to figure out if Chase Daniel could run an NFL-Style offense.  What to do?

One option is to run the software-engineering equivalent of a spread offense.  Use technologies that make things simple for developers, put in lots of policies and procedures, and limit how much damage a poor developer can do.  Such a system may be sufficient for an IT shop that just wants to get the data in the database and produce reports, but it won’t get to the BCS of software engineering, which is what Spolsky is striving for with FogCreek.

The problem in software developement is more acute than college football.  A college football’s coach’s job is to win football games, and if the clearest path to that is to run something like the spread offense, or Georgia Tech’s Triple Option offense, then they are doing the right thing.  But a computer science program’s task is to prepare the next generation of software engineers.  Simplifying the program may bring in and retain more students, but fails at the basic mission, as some are beginning to notice.

Malcolm Gladwell highlights two instances in last week’s Sports Illustrated of athletes (legally) using other tools to enhance their performance, and concludes:

It’s such a relief “performance enhancing drugs” are banned from professional sports, isn’t it? We have no idea what their long-term health consequences are, and there’s a real possibility they offer users an “unfair” advantage.

Oh, dang! You’ve got us all nailed, Mr. Gladwell. We’re all just hypocrites who drummed up the steroids controversy because we really were out to get Fernando Vina.

There will always be athletes looking for an edge, and they will always probably be a half step ahead of the rules and enforcement.

I don’t think that implies, as Gladwell seems to be suggesting, that we must just throw our hands up in the air and say everything goes. There will likely always be poverty, racism, homophobia, violence, and abuse in the world. That doesn’t mean we don’t try to stop what we can out of fear of being inconsistent because we didn’t get rid of it all.

UPDATE: Gladwell has a new post up in which he says he is not advocating legalizing steroids, but pointing out the incoherence of the current culture around it.  Fine, but not terribly interesting.