We’re Running Thousands of Midterm Simulations Until Election Day. Here’s How Often Each Party Wins

After a humbling round of misfires in 2016, political forecasters now face an even more daunting challenge: Predicting the results of 470 simultaneous Congressional elections at a time when yesterday’s rules seem to no longer apply.

Not to be discouraged, most election handicappers predict a surge in Democratic turnout that may well flip the House of Representatives to Democratic control. That said, Congressional election outlooks have a sizable margin of error given the sheer number of races, less comprehensive polling and fluctuating turnout rates. With Election Day upon us, simulations using the combined predictions of three major forecasters give Democrats a 98% chance of retaking the House with about 230 seats, and a 3% chance of retaking the Senate, which is an increase over a 1% chance a week ago.

To get a handle on how likely Republicans are to lose either the House or the Senate, TIME is using the forecasts of three major political seers to create 10,000 electoral simulations every day, which you can activate below. In the electoral maps below, you can select predictions by Larry Sabato’s Crystal Ball, The Cook Political Report, Inside Elections, or a combination of all three to see, based on the confidence level of each race’s outlook, how likely it is for each scenario to play out in November. The simulations will run live in your browser each time.

As you can see, Inside Elections is the most bearish on the Democrats’ chances of a takeover in the House, though most simulations still spit out odds of around 83 percent.

But the Senate is a different story. Far more seats held by Democrats are up for grabs this year than those held by Republicans, placing the odds of a Democratic Senate in single digits.

Should the predictions for the House be correct, it would mirror the midterm elections of 2010, when Democrats, who then controlled both chambers, endured a beating that flipped the House to Republican control and took a big bite out of Democrats’ previously healthy majority in the Senate. Likewise, Democrats ceded both chambers in 1994, two years after Bill Clinton was elected.

This sort of exercise is known as a “Monte Carlo” simulation, wherein a large number of variables make it more plausible to use brute force computation than pages of calculations. The simulations above will update every morning up until Nov. 6, the day of the election, as forecasters fine-tune their predictions and the cascade of unforeseeable news from the White House and elsewhere continues.

Methodology

Each of the 10,000 simulations predicts the outcome of every election individually using the following odds: 95% likely to be correct for districts marked “Safe,” 75% for “Likely,” 65% for “Leans,” 55% for “Tilts” (which only Inside Elections uses), and a 50% of either party winning in “Tossup” elections. The combined predictions are a simple average of the odds of each of the three forecasts.

The “cartogram” map of the 435 House races in the first interactive was created by TIME with the generous consultation of astrophysicist J. Richard Gott III.