One weekend into March Madness and the play has been almost universally outstanding. If you’re a fan of the Big East, Big 12, or Big 10, the tournament has to be a special relief. After months of increasingly desultory play, college basketball exploded back on the national scene with a kaleidoscope of contrasting styles.
Last week, I previewed the tournament for RotoViz through the lens of the March Madness Simulator. So in this week’s edition of Small Sample Size Theater – obviously Gonzaga was overrated; Wichita State would never have caught fire against a real No. 1 seed – I answer the question on everyone’s mind.
How Did the App Do?
The app suggested offensively-oriented teams would outperform. Result: Correct
I focused on Ole Miss and Iowa State in the original previews. Mississippi pulled the upset against Wisconsin before falling to similarly-constructed LaSalle. Iowa State deserved to advance and become a true Cinderella out of the West.
Oregon, whose poor projection may have been fueled by shockingly abysmal play during the temporary absence of Dominic Artis, thoroughly humiliated the shutdown defenses of Oklahoma State and St. Louis.
The app suggested slow burn teams would struggle. Result: Correct
This could be interpreted as the anti-Big East slant. While not foreseeing Pittsburgh’s defeat, the algorithm didn’t like Georgetown. Of Pomeroy’s four highest-ranked teams to be eliminated, three sit outside the top 300 in tempo.
The app did not agree with the RPI on the Mountain West’s prospects. Result: Correct
The Mountain West finished as the No. 1 RPI conference, but their statistical profiles did not hold up to careful scrutiny. The app gave New Mexico a shockingly low chance of defeating Arizona, but they fell to Harvard in the Round of 64 instead. It also hated UNLV, making Cal an automatic upset pick. Colorado State was the bright spot, upsetting a Missouri team that was a chic sleeper pick according to various advanced metrics. Unfortunately, the Rams were playing 5 on 8 in the first half against Louisville and quickly succumbed to the favorites.
The app loved Duke and Florida. Result: So far, so good.
Even though the NCAA tournament committee is probably more reactionary than Greg Blache or Ned Yost, it’s somewhat difficult to completely air ball the No. 1 seeds. Therefore, the app’s selection of a 2-seed and a 3-seed as the favorites becomes interesting. Both squads slashed through the first two rounds. For a variety of reasons, Creighton and Minnesota are tougher matchups than their seeds indicated. Neither put up much of a fight, although Duke benefitted from a truly awful shooting day from the Blue Jays.
Most commentators like to give the defense credit for 3pt% allowed, but it’s almost certainly luck. Consider that most of Creighton’s misses were wide open bricks, whereas the Shockers’ extraordinary 3-point run featured a steady stream of long, contested shots. Luck matters. Creighton shot 41% from three on the season while Wichita State only mustered 34%.
The app picked UCLA as an automatic upset victim. Result: That’s how you get fired despite winning a conference title.
This might be a little unfair, since the loss of Jordan Adams dramatically altered their chances. Still, it often felt like the remaining Bruins stars were more likely to declare for the NBA Draft mid-sequence than get back in defensive transition.
Where did the app go wrong?
Well, despite also having Charles Barkley on board, Memphis did not defeat Michigan State. Chuck, hoarse from repeatedly commentating an epic matchup between fourth-graders, felt that Memphis would win ‘if they could get to 75.’ In this day and age of the diving, leaping, flopping charge, of defenders climbing all over offensive players until they earn the Flagrant 1, of entire conferences trying to drain the shot clock, 75 points might as well be a million. (Unless you’re facing Iowa State.)
Charles was right. Memphis would have won had they scored 75 (assuming you don’t subscribe to the butterfly effect). Unfortunately, they made it to 48.
What about the 1 vs. 8/9 matchups?
The app also liked Missouri, Pittsburgh, and North Carolina to give the top seeds a run for their money. Missouri quickly reinforced the narrative perspective over the statistical one in bowing out to Colorado State.
Pittsburgh, a team beloved by statistical measures everywhere, confirmed that football on a basketball court is not aesthetically pleasing.
The app seemed prescient in picking a tossup between Kansas and North Carolina . . . until Jeff Withey and Travis Releford emerged from the halftime locker room playing as men possessed. Operating with flagrant disregard for the narrative fallacy, I’m going to suggest the psychological battle between Roy Williams and his former employer does not seem to favor the Tar Heels.
Also, for an updated March Madness app, check this post.