My predictions in June 2022

I bought into Galen Metzger’s theory of continuing suburban realignment and persuasion early on. Then as the NY-19 and NY-23 special rolled along, it was shaping up to a roughly even to R+3 fundamentals-based environment.

My homebrewed model was focused on “fundamentals” in a general sense, choosing the races with the highest residuals. I calculated I had a ~20pp edge over predictit odds with these picks. (pretty good return when considering the exorbiant fees)

MI-GOV: 85% D
NV-SEN: 65% R
WI-GOV: 60% D
PA-GOV: 80% D
NH-SEN: 80% D

Pretty good, except that I doubted Ralston who only missed 1 race in the entirety of the state of Nevada (and even he tried to joke-delete his predictions). But there were always good chances to sell for a decent profit on election night.

Pundits had a bad record in 2022

Pundits were victims of insane herding based on nothing more than vibes and dubious quality polling the last week before the election. A prime example is Pennsylvania’s senate race. The 538 deluxe model showed a 15pp increase for Dr. Oz in the last week, almost entirely driven by expert ratings and Republican pollsters. In particular, the reverse-engineering by Sabato’s Crystal Ball and Split Ticket was not grounded in reality. These two justified their picks by feeling that Republicans would win 51 Senate seats.

In contrast, traditional pollsters such as Ann Selzer and NYT were vindicated. With the exception of Emerson who polled the wrong NM districts. In particular that KS-03 poll which showed the Davids leading by 14 points, ultimately dummymandering the Kansas map that Republicans needed to wheel their members out of the hospital to pass.

Conclusion

Don’t herd like pundits. The conventional Twitter wisdom is entirely based on vibes. Focus on the data, and you too can predict that Lauren Boebert is in trouble a week before everyone else does due to early vote data.