Pick 'Em Week 3: How the Pool Shook Out

The Upset That Breaks the Board

Ole Miss's 32–24 win over LSU stands as the week's defining moment not because it was unpredictable, but because it arrived with a 22% pregame probability and still felt narrow. The model assigned LSU a 78% chance. The human field, collectively, was even more bullish on the Tigers. What followed was a game that could have landed either way—a one-score margin that vindicated the contrarians without humbling the favorites entirely. That's the texture of a genuine upset: not a blowout reversal, but a close result that flips the modal expectation.

James Madison's 26–13 road win at San Diego State was the other upset, and it carries a different profile. San Diego State entered at 65% to win. The Aztecs were favored by traditional measure, yet still lost by double digits to a FCS opponent playing away. That game speaks to tier collapse—when a mid-major genuinely outclasses a Group of Five program—rather than variance.

The broader pool picture: GIE posted 43 points on 8 correct picks from 10 games. The human field went 20-for-30 across three players, which works out to 6.7 correct per player. VolsByFiddy led the week at 51 points with 8 correct, edging the model on volume and precision. The gap widens when you account for confidence; VolsByFiddy extracted more value per pick than the black box did, suggesting a sharper read on at least one or two spots where consensus had drifted.

The Trap Signals

Florida's 44–39 road win at Auburn is instructive. Auburn sat at 33% pregame; the Tigers had no edge and got worse as the game progressed. The field heavily favored Florida, and they were right. No trap there—just consensus meeting reality.

Louisville and SMU followed a similar path. Louisville was favored (60%) and delivered convincingly, 41–31. Virginia Tech beat Maryland (35–26) as a road favorite at 61% to win. These weren't close calls; they were the chalk holding up.

The week's tightest games—Vanderbilt over NC State (35–31, 88% pregame for Vandy) and Jacksonville State over Georgia Southern (31–27, 60% for the home team)—split the difference. Favorites won, but the margins were tighter than projection. Neither qualifies as a trap; both fell within the confidence intervals that strong favorites usually occupy.

The Lean Forward

What tilts this week toward GIE's column: the model nailed the high-leverage calls (Florida, Louisville, Vanderbilt) and only suffered on the two genuine upsets. VolsByFiddy's lead points to superior pick-level conviction on one or two middle-tier games where the human edge typically emerges—spots where historical pattern or recent performance data tilts the model's central estimate but not its confidence band far enough.

Week 3 rewards decisiveness. Ole Miss and James Madison punished unanimous favorites. The field that hedged, or hedged smartly, is pulling away from the field that didn't.

Human picker edge still lives in the moments between the model's certainty and the game's actual distribution.