At 9:00 pm eastern time on election night, my model showed a 93% chance of Trump winning, while the New York Times needle showed a 60% chance.
By 10:15 pm, my model showed a 99% chance of Trump winning, while the New York Times was still at 70%.
- Simulations were based on the latest state prediction markets from Polymarket.
- 20,000 elections were simulated by drawing multivariate Gaussians centered at log-odds of betting markets.
- Source code is available here: lambda_function.py.
- I might update this page with more information if I get some free time next weekend.