The Field Stumbled Into Tuesday
Ole Miss beat LSU 32–24 as a 22% pregame underdog. That's the Week 3 upset that mattered—not in abstract terms, but in how it cleaved the pick pool open. VolsByFiddy, the week's leader, nailed it. The model did not. GIE finished 8-for-10 with 43 points; the human field went 20-for-30 across three players, which means the pool absorbed the LSU miss and moved on. But that one game, on that night, separated the sharp from the reactive.
The rest of the slate cooperated. Florida held off Auburn 44–39 despite entering as a narrow favorite (67% pregame win probability). Vanderbilt demolished NC State 35–31 as heavily favored (88% WP). Louisville beat SMU 41–31. Virginia Tech covered at Maryland, 35–26. Smaller spots broke as expected: South Alabama over Ohio, Jacksonville State over Georgia Southern, Delaware over Coastal Carolina. UConn destroyed Southern Miss 48–20. James Madison's upset—a 65% favorite losing to a lower FBS program—felt noisier than Ole Miss's, but it moved less money and fewer confidence points. San Diego State fell at home.
What made Ole Miss different was the gap between public expectation and result. A 22% underdog is functionally a statement: Vegas saw LSU as a decided-favourite team. The Tigers had been favored in pregame models across the industry. Then the game happened, and Ole Miss won by eight. Not a competitive loss. A win. That's the kind of move that costs a pick pool real points, because it forces a choice: follow the pregame consensus or hunt for the angle.
VolsByFiddy hunted. Eight correct picks on a ten-game slate—that's 80% hit rate, and it was enough to score 51 points and lead the pool. GIE hit 80% as well, but scored lower because confidence allocations were elsewhere. The model trusted Florida and Vanderbilt more. Both won. Neither was a trap. The difference between a week-leading week and a solid week was one game at 22%.
The broader field—Cole at 27 points with 6 correct, g-cat at 29 with 6—took bigger hits chasing confidence elsewhere. Over ten games, one miss doesn't erase a slate. But one upset at sub-40% pregame probability does. Week 3 was less about scheming or execution and more about betting your confidence on one horse: did you take Ole Miss, and how much did you allocate to it.
Three weeks in, VolsByFiddy leads outright. The model sits second. The gap widens when one voice finds an edge the others miss—and in a ten-game week, that edge often has a team name attached to it, not a metric.