Our Dumb Geo-Bracket Did Better Than Your Bracket
| April 3, 2012 | Posted by ReubenFB under Pointless |
Right before the start of NCAA Tournament a few weeks ago we examined the geographic placement of teams in the bracket, identifying squads the selection committee had granted a quasi-home court advantage (Kentucky, Duke, Kansas) and the squads who’d be traveling thousands of miles to play their games (Cal, Davidson, Gonzaga).
Since we’d already done they legwork, we went ahead and put together a purely geography-based bracket based on the simple, silly rule that the team closer to the site of the game will always win. Now that it’s all said and done, how’d it do?
As it turns out, shockingly well.
This bracket, which among other crazy things picked (12) LBSU to go to the Final Four, finished in the 89.8th percentile among all ESPN brackets. Let’s take a look at what it did right:
- Kentucky over Kansas in the Championship Game. Kentucky had the best geographic placement in the tournament having to travel just a total of 1,011 miles from home to its three playing locations, but for obvious reasons we can’t take too much credit for picking them to win since 35.1% of ESPN bracketeers did the same. Kansas was the better call, as they had the 4th-best geographic placement but were only picked by 18.2% to be in the final. Notably, the Jayhawks were 367 miles closer to the site of their Regional Final than (1) UNC, who they defeated 80-67.
- Ohio! This year’s Cinderella had the 13th-best geographic draw of the tournament, the best of all 11-13 seeds (the Cinderella sweet spot). This bracket picked them over Michigan (117 miles closer), USF (452 miles closer), and UNC (204 miles closer). So close to true glory, but still only 4.2% of brackets had the Bobcats in the Sweet 16.
- The Opening Round. Speaking of USF, this bracket picked 7 out of 10 of the second round upsets correctly as (11) Colorado, (10) Xavier, (9) Saint Louis, (10) Purdue, (11) NC State, (13) Ohio, and (12) USF were all closer to their game sites than their opponents.
Here’s what it did not so good:
- The West Region. A down year for the Pac-12 made the West a mess, as the Phoenix-based region featured a 1-seed from Michigan, a 2-seed from Missouri, and a 3-seed from Wisconsin. As a result, the straight-geography rule picked nearby 12-seed LBSU to win the region. They did not, and the bracket only picked 6 of these 15 games correctly.
- A Dumb Sweet Sixteen. This bracket had (13) New Mexico St., (12) LBSU, (10) West Virginia, and (13) Montana riding favorable geographic draws to the Sweet Sixteen. With the exception of LBSU, all of them lost their first game by at least 13 points.
- Bad “Locks.” (10) West Virginia was 1,851 miles closer to its first game than (7) Gonzaga, and yet lost by 33 points. (5) Wichita State was 961 miles closer than (12) VCU, and yet lost by 3.
If you’re having the trouble grasping just how crazy it is that this bracket did so well consider this: if (16) WKU had been seeded as a (15) instead of a (16), this model would have picked them to go to the Sweet Sixteen. They had the 3rd-best geographic placement in the bracket but drew Kentucky in the opening round. Likewise if (14) Belmont was just 11 miles closer to Columbus and 3 miles closer to St. Louis, this model would have picked them to win the ENTIRE FUCKING TOURNAMENT.
The second, more conservative bracket we made followed the rule that the higher-seeded team will always win, unless it’s more than 350 miles further away from the site of the game than the lower-seeded team. This bracket did even better, finishing in the 95.7th percentile of all ESPN brackets by also picking Kentucky over Kansas in the Championship (let’s not kid ourselves, that’s 46% of my points right there) while being otherwise really boring and only picking 9 upsets. In a year that was basically chalk with a few insanely unpredictable results, the conservative route would have won our pool, but the insane geography bracket will always be closer to our hearts.
