concept
The classical test of whether a result is unlikely to be chance. In episode 01 Marvin Barth concedes his forecast never reaches it and argues that, for voters missing from polls, it never could. Start with episode 01, at about 28 minutes.
Across the episodes
01
Enters as the standard Barth admits he cannot meet, and argues no one could
Barth's reason the standard cannot apply: you cannot test for people who are absent from the sample.
"how would I ever statistically test whether something is in the poll or not if I know it's not in the poll?"Marvin, The Grant Williams Podcast: Forecasts and Fortunes, Episode 1 (2 Nov 2026), 27:49
The concession he puts at the front of his piece.
"I say, hey, look, nothing I have here is statistically significant."Marvin, The Grant Williams Podcast: Forecasts and Fortunes, Episode 1 (2 Nov 2026), 35:12
Barth names the academic standard his work does not meet.
"Like this would not pass any peer review process, right?"Marvin, The Grant Williams Podcast: Forecasts and Fortunes, Episode 1 (2 Nov 2026), 35:35
The turn from method to verdict: judgment fills the gap the statistics leave.
"And at some point, you have to recognize you're not going to get statistical significance and you still have to make a call. And so here's my call."Marvin, The Grant Williams Podcast: Forecasts and Fortunes, Episode 1 (2 Nov 2026), 47:28
Connections
limit ofPolling Bias
The bias cannot be shown significant in advance.
contrasted withBayesian Statistics
Barth's process does not reach significance.