PressAudit's Pr(D win senate) = 0.63 is in the ballpark with forecasts from The Economist (0.58) & FiftyPlusOne (0.73)
Here's what PressAudit's midterms forecast page says about methodology:
> PressAudit probabilities blend recent polls with Polymarket and Kalshi odds, giving polls more weight as more surveys become available. Chamber-control odds come from 10,000 simulations that account for shared national and state-level shifts. These are experimental estimates, not guarantees; the simulation assumptions have not been historically calibrated.
Hmm. In contrast, FiftyPlusOne's forecasting model doesn't use prediction markets as inputs:
> Other forecasters are increasingly using prediction markets as signals in their forecasting systems. We do not [...] There is a poisonous logical circularity in using prediction markets to update a statistical model when the statistical model is itself being factored into the prediction market. Because of these limitations, our opinion is that if you’re going to inject prediction market data into your forecasting model, you should at least let users view a version that doesn’t rely on such data.
[1] https://www.economist.com/interactive/2026/us-midterms/predi...
[2] https://fiftyplusone.news/
[3] https://blog.fiftyplusone.news/p/2026-forecast-methodology
First of all, the most valuable thing that my tracker has is putting real time media bias analysis on the news. Eg: MI overview: https://pressaudit.org/midterms-2026/senate/mi MI media analysis: https://pressaudit.org/f/midterms-2026/michigan?view=article... The news, its media, bias, analysis, predictions, and polls are all put in one place. You can select maps that are based off of predictions only, or polls or a few other forecasting models. So users have some options. But if there’s one thing that I would want them to get from my tracker that they couldn’t get from any other tracker, it’s not the model, it’s the live dossier with media bias analysis on each race that. I offer alternative maps in the map library, so you can look at it based off of polls or other’s models. But I don’t check them. I don’t even check my own composite rating if I’m being honest, I prefer looking at individual pieces of data: recent polls, news, and markets. Eg I think MI and TX Senate Dem odds are being overestimated, but that’s a gut feel based on the intense ground campaign operations and ads in TX and Jewish Democrat backlash in MI. Harder to model those feels. Onto the model: I think that the 99% chance of taking back the House is not valid unless you are accounting for all uncertainty independently for each individual district. I accounted for uncertainty at national and district level. National uncertainty = one shared influence pushes races across the country toward Democrats or Republicans. District-level would mean that every district operates independently, and in that world, all of your simulations will result in district flips cancelling each other out and being kind of random. Every forecaster will have some fudge factor to account to uncertainty and if you do it at the national-level, you increase the possibility of some unknown national factor moving all elections together. I assume that the possibility of some uncertain national factor moving races nationally is higher than 1%. 99% likelihood gives me flashbacks to the New York Times needle on the morning of election day 2016, versus the evening.
Onto predictions: they take all of the available signals and turn it into one data point. With polls, it’s harder to factor in the effect of campaign financing or local voting impairments. I think any model that is primarily based off of polls is going to significantly underestimate the impact of the money spigot and ground game where Republicans hold an advantage. I think it’s unlikely that that advantage is enough to overcome how deeply unpopular Trump is right now.
Let’s see how wrong we all are after the election!