Do fans care about the path after they expect a berth?

Available fields are not an analysis-ready sample.

The same ESPN scoreboard request was used for November 16, 23 and 30, 2024. Each returned 25 rows despite the requested limit of 1,000. The returned slice includes classification gaps, including a Delaware–Villanova game. It is not a complete FBS frame or a representative sample. Completion, classification and any feed truncation must be resolved before analysis.

Measure four changes before kickoff.

For the same frozen information and remaining-season simulation draws, compare a focal win with a focal loss. Stakes equal the probability after a win minus the probability after a loss. Record qualification, a single bye, a double bye and hosting at least one game separately. Expected number of home games is a different quantity.

Single and double byes are mutually exclusive entry categories. A host or bye requires qualification. In an actual format without double byes, that category is zero by rule. The proposed eighteen-team arrangement was not in force historically; a simulated BetterCFB path is a counterfactual, not an incentive fans actually faced.

Use archived pregame ratings, schedules, rankings and rules. Do not substitute final CFP rank, postgame win probability or today’s retrospective NRR. Document when other results became known. The intake validator rejects future timestamps, missing provenance and inconsistent probabilities; passing it verifies structure, not the truth of the supplied archive.

Freeze the design before joining outcomes.

1. Sample: propose the final four scheduled regular-season weeks of 2024 and 2025 FBS home games, excluding neutral sites and conference titles. The current three-day slice is only a feasibility probe, not that study sample.
2. Quality and selection: model qualification and seeding stakes together, with team-season effects, past-only form, opponent quality, rivalry, network, time slot and competing games. Check enough within-team variation and overlap before estimating separate effects.
3. Capacity: compare raw attendance, verified fill ratio, documented non-sellouts and historically non-sellout programs. Do not automatically impose a Tobit model: ticket sales, reported counts, capacity changes and true censoring need checking first.
4. Checks: settled-seeding games should have zero modeled seeding stakes. A near-lock sensitivity uses qualification probability at least 0.95; that is not mathematical clinching. Test calibration and guard against final-result leakage.
5. Interpretation: ordinary regression gives an association. Another team’s result can share news, broadcast or schedule shocks; it is not automatically a natural experiment. NFL benchmarking requires its own sample and transfer limits.

Interest, total fan spending, rights revenue and local economic benefit require different outcomes. A full stadium or a positive attention association does not establish any of the latter.