As of around 9 am ET on Monday, Miller’s site, Virtualtout.io, gave the Democrats a 90% chance of winning at least 51 seats. In recent days, the GOP’s chances of keeping control of the upper chamber have virtually collapsed, dropping from 23% on Oct. 26 to under 10% now.

A major reason from the quick downdraft, and a sign of the blue wave’s strength: Democrats now hold the lead in longtime red stronghold Kansas, and have turned Iowa into a virtual tossup.

Miller, a professor at Northwestern University, won kudos for correctly forecasting that Democrats would sweep the two Georgia Senate runoffs in early 2021, when the polls and pundits posited GOP wins in both. He scored again by calling a Trump win in 2024.

His methodology: Miller starts by incorporating the “prices” posted on the two predictions markets boasting the highest volumes, Kalshi and Polymarket, and translates those numbers into probabilities of a Democrat or Republican. He then deploys multiple time series that examine 10-12 underlying factors that drive the prediction markets, and continually crunches the numbers to forecast their impact going forward.

Miller updates the probabilities every 10 minutes, and the numbers stream across the top of Virtualtout.io in a continuous band resembling a stock ticker. The “NowCast” also displays the change over the last 24 hours. Hence, his figures are far more current than the polls.

“They all relate to voters’ perception of the performance of the Trump administration,” Miller told Fortune. “That’s the thread running through all the races. The factors include affordability, the economy, and the war in Iran, the issues that motivate voters. The states don’t move independently of one another. When voters’ view of those issues turns more negative, the Republican numbers drop on all the races.”

In other words, Trump has effectively nationalized the battle for the U.S. Senate.

Miller added that the impact of Trump’s tireless stumping is harder to assess. “People have come to expect certain things from Trump,” he said. “The way he behaves is not a surprise, it’s a continuous thing.”

Most of what the president says on the Senate campaign trail, according to Miller, simply reinforces the view of Trump and his party that voters already hold. “Still, there’s no evidence he’s helping,” Miller said. “It doesn’t appear that his appearances are a positive.”

One instance where his model found that Trump’s rhetoric backfired: The POTUS’s defense of the ad Vivek Ramaswamy ran in Ohio featuring the mother of his Democratic opponent Amy Acton severely criticizing her daughter. “Following Trump’s praise of the ad, we saw a huge drop in the Republican percentage,” Miller noted.

Republican numbers have dropped big time in almost all the key races

He pointed out that in the last 10 days, “a lot of contests are moving a little bit” in the Democratic direction from day to day, but the cumulative gains are substantial. In the five elections where the Democrats have long harbored chances for a flip, the GOP stands below 30% in all of them, from a high of 28.4% in Alaska to a low of 2% in North Carolina.

The Republicans’ best chance comes in Maine, where incumbent Susan Collins sits at 34.7%. In Michigan, the probability of a Republican flip via beating Abdul El-Sayed has faded from over one-third in mid-September to a current count of 24.5%.

The Miller model now points to high possibility of Democratic wins in states that, until a few days ago, the GOP counted on as surely in their column. “Those are the races I’ve been watching closely and the move towards the Democrats in the last few days is surprising a lot of people,” says Miller. In the Iowa run for an open seat, Republican Josh Turek held an almost 62% shot at victory as of Sept. 17; now, he’s dropped to 51.6% as his opponent Ashley Hinson pulls almost even.

In Kansas, Sen. Roger Marshall looked on track for easy re-election just two months ago at around 75%. His number’s dipped to 43.9%, and suffered a 7.1% drop just in the past 24 hours.

Miller highlighted that Virtualtout.io and its NowCast will keep running all through Election Day, and will be first to call races, mainly based on its analysis of the swings in the betting odds.

Of course, the Republicans could hang on by keeping Kansas and Iowa, winning Maine where their odds are still decent, and defying the probabilities in Texas or one other key state. It’s a narrow path that in recent days now looks far narrower as Midwestern strongholds swing blue.