Wharton Moneyball The Limits of Prediction in Sports and Elections
Nov 4, 2026
Election polls, prediction markets, and sports models all face uncertainty, especially when errors move together. The discussion explores how early results can reshape forecasts, why NFL parity makes division races unpredictable, and whether playoff teams can overcome roster weaknesses. It also examines tennis rankings, Connor McDavid’s scoring pace, the Sixers’ hot start, and the college football playoff race.
AI Snips
Chapters
Transcript
Episode notes
Correlated Errors Make Landslides Look Impossible
- Treating election errors as independent makes aggregate outcomes look far more certain than they are.
- Correlated polling errors can push many close races in the same direction, as the panel argued happened in 2016.
Why Nate Silver’s 2008 Record Was Misleading
- Nate Silver’s 2008 state-by-state success looked extraordinarily unlikely only if races were independent.
- Shared national conditions meant correctly calling one state made many other calls easier, not one-in-a-thousand extraordinary.
Use Early States To Test Election Models
- Watch early results for forecast errors, not merely winners, to update beliefs about the entire election.
- Eric Bradlow highlighted Maine, the Carolinas, and Florida as early tests for correlated polling error.