​Where the model nailed it

• Myrtle Beach Bowl (Kennesaw State vs. Western Michigan): The model finished within ~1% of actual attendance by correctly capturing Myrtle Beach’s destination-bowl dynamic, consistent tourist-driven demand, and a venue size that naturally limits variance, resulting in turnout that closely tracked historical norms regardless of participant travel distance.

• Wasabi Fenway Bowl (UConn vs. Army): Attendance landed within 3.55% of projection as the model accurately reflected the combined impact of UConn’s near-home proximity, Army’s nationally distributed alumni base, and Fenway Park’s novelty appeal, which sustained demand despite cold-weather conditions.

Where it was directionally right

• StaffDNA Cure Bowl (Old Dominion vs. USF): The model correctly identified strong regional interest for a highly drivable matchup, though actual turnout showed that regional enthusiasm does not always convert proportionally into paid attendance.

• Idaho Potato Bowl (Washington State vs. Utah State): Attendance exceeded both the historical average and the projection, underscoring how strong regional proximity can meaningfully elevate demand in cold-weather bowls beyond typical baseline expectations.

Where it missed (and why that matters)

• Gasparilla Bowl (Memphis vs. NC State): Attendance fell well below expectations, illustrating that brand strength and Power Conference affiliation alone are insufficient to drive turnout in saturated Florida markets without a clear local or near-local fanbase anchor.

• Birmingham Bowl (Georgia Southern vs. Appalachian State): Despite apparent regional proximity, limited travel accessibility and overlapping fanbase footprints constrained attendance, highlighting that practical ease of travel and local market appeal matter more than distance alone for neutral-site bowls.