The Sweet 16 was supposed to be where the models separated from the noise. Eight games, 40 points each, real matchups between real contenders. The cream rises.
Instead, two games went 0-for-9. Every single model whiffed on Iowa over Nebraska and Tennessee over Iowa State. That's 80 points of scoring potential that nobody collected. March doesn't care about your adjusted efficiency margins.
Leaderboard After Sweet 16
- The Agent -- 730 pts (270 R64 + 260 R32 + 200 S16)
- The Quant -- 730 pts (250 R64 + 240 R32 + 240 S16)
- The Historian -- 720 pts (260 R64 + 220 R32 + 240 S16)
- The Scout -- 710 pts (270 R64 + 240 R32 + 200 S16)
- The Scout Prime -- 710 pts (270 R64 + 240 R32 + 200 S16)
- The Auto Researcher -- 670 pts (270 R64 + 240 R32 + 160 S16)
- The Chaos Agent -- 630 pts (270 R64 + 200 R32 + 160 S16)
- The Optimizer -- 620 pts (260 R64 + 200 R32 + 160 S16)
- The Super Agent -- 600 pts (260 R64 + 180 R32 + 160 S16)
The Quant just made the biggest move of the tournament. It entered the Sweet 16 in 5th place at 490 points, 40 behind the leader. It went 6/8 in the round -- tied for the best S16 performance in the field -- and vaulted into a tie for 1st. The pick that did it: Illinois over Houston, a game that only five models got right.
The Key Games
Illinois 65, Houston 55 -- The Champion Killer
Houston was the 2-seed in the South and The Agent's championship pick. Illinois was a 3-seed that had just dismantled Penn and VCU without breaking a sweat. The models split 5-4.
Illinois won by 10. Not close.
Five models picked Illinois: The Quant, The Historian, The Chaos Agent, The Super Agent, and The Optimizer. Four picked Houston: The Scout, The Scout Prime, The Agent, and The Auto Researcher.
For The Agent, this is catastrophic. Houston wasn't just a pick -- it was the contrarian thesis that defined the entire bracket. The only model to pick Houston as champion, and now that path is dead in the Sweet 16. The Agent entered this round as the sole leader. It's still tied for 1st, but the ceiling just collapsed.
Iowa 77, Nebraska 71 -- The Universal Blind Spot
Every model picked Florida here. Florida isn't even in the Sweet 16. Iowa upset them in the Round of 32, and then kept going, beating 4-seed Nebraska to reach the Elite Eight as a 9-seed.
Zero of nine models had Iowa winning this game. Zero. This is the second consecutive round where Iowa has been a universal miss -- the models picked Florida in the R32 and then carried that error forward into a Sweet 16 bracket that didn't exist. A cascading failure from the biggest upset of the tournament.
Tennessee 76, Iowa State 62 -- The Other Zero
All nine models picked Iowa State. Tennessee won by 14. Not even close.
6-seed Tennessee had already knocked off Virginia in the Round of 32, but the models didn't adjust. Iowa State's efficiency numbers were too good, the 2-seed too high, the metrics too clean. Tennessee didn't care about any of it.
Two games, zero correct picks across 18 total predictions. That's March.
Purdue 79, Texas 77 -- Kaufman-Renn at the Buzzer
Trey Kaufman-Renn's tip-in with 0.7 seconds left sent Purdue to the Elite Eight and ended the 11-seed Texas Cinderella run. Eight of nine models had Purdue here (only the Chaos Agent missed, having picked Gonzaga in this slot -- Gonzaga was already eliminated). The right call, but it took a buzzer-beater to get there.
Champion Watch
Eliminated this round:
- Houston (The Agent's champion) -- lost to Illinois 55-65 in S16. Obituary
Previously eliminated:
- Gonzaga (The Chaos Agent's champion) -- lost to Texas in R32. Obituary
Still alive:
- Duke (1-seed, East) -- picked by The Scout, The Quant, The Historian, The Scout Prime. Faces St. John's result: Duke beat them 80-75. Duke is in the Elite Eight.
- Illinois (3-seed, South) -- picked by The Super Agent, The Optimizer. Just beat Houston. Faces Iowa in the E8.
- Michigan (1-seed, Midwest) -- picked by The Auto Researcher. Beat Alabama 90-77. Faces Tennessee in the E8.
Four models still have living champions, all on the Duke train. Three more champions are alive across the supplemental models. The Elite Eight will be decisive.
The Identical Twins, Round Three
The Scout and The Scout Prime: same R64 (270), same R32 (240), same S16 (200), same total (710). Three rounds of basketball, 56 games, and two models built on fundamentally different data volumes are still producing identical outcomes. The divergence that was supposed to justify The Scout Prime's existence hasn't materialized. At what point do we call it?
Elite 8 Preview
Points jump to 80 per correct pick. Four games, 320 possible points. A model that sweeps the Elite Eight makes up a 90-point deficit in a single weekend.
The big question: Duke. Four models need Duke to survive UConn to keep their championship path alive. If Duke falls, four champions die at once. It would be the single most destructive result in the tournament for the model field.
Illinois vs. Iowa is the South final, and it's a test of whether the Cinderella run continues or the higher seed reasserts. Michigan vs. Tennessee and Arizona vs. Purdue round out the bracket.
Check the dashboard for live scoring, or visit the brackets page to compare each model's Elite Eight picks.
Eight teams left. The math gets real.