Monday, April 27, 2009

Cognitive Bias potentially discovered

I was thinking about the concept of upsets.  Of teams being either better or worse than their record.  Essentially, teams taking days off and giving up a game.  

As far as I know only really good or really bad teams take games off.  Everyone else is scrabbling for everything they can.

But I've never studied the playoffs and seeding to see how many lower-seed upsetting higher-seed situations there are.  The NBA is uniquely suited among the major sports to see how match-ups effect playoff results because of the linearity of scoring, and the seven game series format.  This allows for a LOT of basketball to be played between two teams in an effort to sort out who is the best.  

One of the problems with analyzing it is the seeding.  The NBA uses a 1 vs 8, 2 vs 7, 3vs 6, and 4 vs 5 seeding.  An alternative seeding could be 1 vs 2, 3 vs 4, 5 vs 6, and 7 vs 8 format.  This would let the 1 and 2 seeds duke it out right away, without risk of players getting injured while getting the scrubs out of the way.  The winner would likely face a moderate test in the 3 vs 4 winner, and then celebrate the title early as they beat the winner of the 5 vs 6 vs 7 vs 8 scrum.  

The last series would be akin to a victory lap.  

If we used that seeding, we'd have lots of instances of teams with very close records playing seven games series and we'd see what sort of strength there is to a W-L record, i.e. how accurately it predicts the outcome.  

Just a random thought I had about seeding.  

Also, when I research seeding upsets, check to see the expected wins and losses because of the disparity in homecourt advantage for the higher seeded team.  What I mean is that if two identical teams played, and one had home court advantage in a seven game series in the current format, what percentage of the time would they win.  And what percentage of the time does that actually happen.  

No comments: