Researcher ORCID Identifier

0009-0006-7222-2958

Graduation Year

2026

Date of Submission

4-2026

Document Type

Campus Only Senior Thesis

Degree Name

Bachelor of Arts

Department

Economics

Reader 1

Laura Grant

Rights Information

© 2026 Kevin Daniel White

Abstract

This paper tests whether preseason uncertainty causes sportsbooks to make larger prediction errors in early-season NFL games than in late-season games, using game-level OLS regression on 6,185 regular season games from 2002 to 2025, where prediction error is measured as the absolute difference between the closing point spread and the actual game margin. Across the full sample, prediction error holds steady at roughly 10 points regardless of where in the season a game falls, and coaching changes do not amplify early-season error. Rather than amplifying error, larger draft capital gaps between teams are associated with slightly less prediction error, suggesting that observable roster investment signals are priced in rather than overlooked. A notable exception emerges in a subsample from 2014 to 2022, where games in the opening two weeks involving a new head coach show meaningfully higher prediction error than comparable late-season games — a localized pattern that the full-sample results obscure. Together, the findings are consistent with a market that is broadly efficient on average, where any inefficiencies appear too localized in sample and too narrow in scope to support systematic out-of-sample exploitation.

Download

This thesis is restricted to the Claremont Colleges current faculty, students, and staff.

Share

COinS