Everyone says the polls blew it. Carl Allen says they didn't, and his research shows candidates averaging 49% or higher in well-polled races have won 98% of the time since 2004. Then we go race by race: Texas, Ohio, Maine, and two red states that could shock everyone on election night.
Recorded live on Substack, this conversation with data analyst Carl Allen starts with why polls aren't predictions and ends with a tour of the 2026 Senate map, with real poll averages and prediction market odds pulled up as we talked. Along the way, Carl explains why a one-point lead can be safer than a three-point lead, and why undecided voters in Minnesota aren't the same as undecided voters in Ohio.
✅ Share this episode with the friend who refreshes the polls every morning
Key takeaways
Polls estimate a candidate's current support, not the final margin. When late deciders break heavily one way, as they did for David Crowley in Wisconsin's governor primary, that isn't a poll error.
Watch the 50% line, not the spread. A candidate near 50% has little room to be overtaken, no matter how many voters are undecided.
Undecided voters aren't a monolith. That's why prediction markets give a small Democratic lead in Minnesota 90% odds while a similar lead in Texas gets about 60%.
Carl's sleeper races are Kansas and Nebraska. If Democrats have a strong night, he expects those two to move together.
About Carl Allen
Carl Allen is a researcher and analyst of sports and political data and the author of The Polls Weren't Wrong (CRC Press, 2024). His research focuses on inefficiencies in political betting markets and improving polling methods, and his background includes quantitative analysis for University of Louisville Baseball and MLB data work for Stats Perform.
Podden och tillhörande omslagsbild på den här sidan tillhör
SCAN Media, LLC. Innehållet i podden är skapat av SCAN Media, LLC och inte av,
eller tillsammans med, Poddtoppen.
