America Decides!

Count in progressResults read from NBC News and Fox News on the morning of 4 November. We are updating this page as the states finish.

On Tuesday night, as the returns came in, the map began to look familiar. Georgia, North Carolina, New Hampshire, Minnesota, Maine, Ohio: one battleground after another landed where our synthetic voters had put it the day before.

That call went on the record on 2 November, when our Decision Desk published a number for every Senate and governor race in the country, each one drawn from 200 synthetic voters built to match their state. Of the 70 races decided so far, we named the winner in 60. In those six Senate battlegrounds we came within about two points of the count every time: Jon Ossoff’s 17.8-point win in Georgia against our 16.4, Roy Cooper’s 20 in North Carolina against our 22, Sherrod Brown’s return in Ohio at 4.4 against our 3.4.

The poll average missed all six by more, and always in the same direction. It knew the Democrats would win; it could not see by how much. We saw much more of it.

Fig. 1

The battlegrounds: our call landed on the count

Poll averageOur callCount so far

The map

Here is every call we made, on a map of the states. Each tile is coloured by the party leading the count. The tick says our call named that leader.

Fig. 2

Sixty of seventy, state by state

Democrat or independent leadsRepublican leadsStill counting✓ / ✗ our call

Senate

Governor

Senate

Governor

Democrats have flipped seven Senate seats: Iowa, Kansas, Maine, North Carolina, Ohio, South Carolina and Texas. We called four of the seven flips: Maine, North Carolina, Ohio and Texas. We also called every Democratic hold, from Georgia and Michigan to New Hampshire and Minnesota.

Seventy races, one line

If every call matched the count exactly, every dot below would sit on the dashed line.

Fig. 3

Seventy races, one line

SenateGovernor

They hug it. Across the 35 Senate races, our calls and the count so far have a correlation of 0.94, from Wyoming’s Republican landslide to Rhode Island’s Democratic one.

We saw the wave

The polls were not wrong about who would win most of these races. They were wrong about by how much. In the typical Senate race between a Democrat and a Republican, the poll average under-counted the Democrat by 5.0 points. Our call under-counted them by 1.6. For governor, it was 4.1 against 3.6.

Fig. 4

In the typical race, we saw more of the wave

Poll averageOur calls

In the races that decided the Senate, that was the difference. A synthetic electorate is not a poll and it does not average polls. It is a population built to match the real one, each voter given what they would have known in the last days of the campaign. When the polls all lean the same way, it can come down somewhere else, and on Tuesday, in the battlegrounds, it came down closer to the truth.

Three calls against the polls

In three races, the poll average and our call named different leaders. In all three, the count has gone our way.

Fig. 5

Three calls against the polls

Poll averageOur callCount so far

• Georgia governor. The polls had Rick Jackson narrowly ahead. We had Keisha Lance Bottoms. She won by 8.4.
• Kansas governor. The polls had Ty Masterson. We had Cindy Holscher by 0.4. She leads by 0.6 with 98% counted.
• Alaska Senate. The polls had Mary Peltola. We had Dan Sullivan by 2.5. He leads by about a point, with Alaska still counting.

The closest calls

Race	Our call	Count so far	Off by
Oklahoma Senate	R+19.0	R+18.8	0.2
Kansas governor	D+0.4	D+0.6	0.2
Ohio Senate	D+3.4	D+4.4	1.0
New Jersey Senate	D+28.9	D+27.9	1.0
Michigan governor	D+8.5	D+9.7	1.2
Alaska Senate	R+2.5	R+1.2	1.3
Minnesota Senate	D+11.9	D+10.5	1.4
Georgia Senate	D+16.4	D+17.8	1.4
Wyoming Senate	R+31.8	R+30.4	1.4
New Hampshire Senate	D+16.0	D+14.5	1.5

The two races we called first

Before the Desk, we published two races on this blog, weeks ahead of the vote.

Michigan. On 8 October we published Abdul El-Sayed by 4.4. He leads Mike Rogers by 5.9. Our call was closer than the poll average, and closer than four of the five polls we tracked that day.

Texas. On 22 September, six weeks out, we published James Talarico ahead of Ken Paxton, while one forecaster had Paxton at 77% to win. Talarico has won, the first Democrat to win statewide in Texas in more than thirty years.

The scorecard

	Races decided	Winner named
Senate	34	30
Governor	36	30
House	24	to be scored
Total	70	60

---

Where we fell short

A desk that only shows its wins is not keeping score.

Ten wrong winners. In the Senate we had Iowa, South Carolina and Kansas staying Republican, and Dan Osborn beating Pete Ricketts in Nebraska. For governor we missed Vermont, New Hampshire, Texas, Iowa, Nevada and Florida.

The polls were closer on the margin. We saw more of the wave, but we were noisier. Our calls missed the Senate margin by 7.5 points on average, against 5.9 for the poll average, and we were closer than the polls in 27 of the 62 races both of us called. The gap is widest in close races: where the count is within 15 points, the polls missed by 4.7 on average and we missed by 8.5. Our big misses also pull our average lean back toward the polls’: 3.6 points too Republican in the Senate against their 4.7.

We stretch safe seats. Our electorates push lopsided states further than the voters did. Delaware, Virginia and New Mexico came out about 10 points too Democratic, and West Virginia and Alabama 15 to 21 points too Republican.

Red states that swung, we missed. The largest Democratic gains came where our electorates held firmly Republican: Louisiana (19 points short), Iowa (14), South Carolina (12.5), Mississippi (12) and Kansas (9).

Popular incumbents across party lines. Phil Scott in Vermont and Kelly Ayotte in New Hampshire won with votes from the other party. Our synthetic voters did not cross over for them, the same miss we saw in our pre-election tests.

Black voters, again. Mississippi is level. We had Cindy Hyde-Smith by 12. Scott Colom closed the gap in the Delta and in Hinds County, which are heavily Black. In our Michigan post we called the way our synthetic electorates treat Black voters the largest known defect in this method. Mississippi says it still is.

Two numbers that drifted. The Desk’s final Michigan call, El-Sayed by 10.4, was further out than the 8 October call on this blog. In Texas we named the winner but at 2.4 points; Talarico leads by about 12. The polls were nearly as far out.

---

What comes next

The counting continues. As the states finish we will update every number here, score the 24 House races, and put the results on the Desk beside our calls.

Every voter in these studies is synthetic. The figures describe simulated electorates, not polls of real people, and were published before polls opened as testable estimates. Results are the count as of the morning of 4 November 2026.

---

Sources

Results

• NBC News, Senate results and governor results, read about 3 a.m. ET, 4 Nov 2026.
• Fox News, Senate results and governor results, race calls with times.
• NBC News, Ricketts wins Nebraska, called at 3:27 a.m. ET.
• The Hill, Mississippi Senate race too close to call; Mississippi Free Press, Colom leads by 28 votes.
• Alaska Public Media, Sullivan and Peltola locked in tight race.
• Kansas Reflector, Governor’s race deadlocked.
• Texas Tribune, Talarico defeats Paxton.
• Axios, El-Sayed defeats Rogers.

Our calls and the poll averages

• Decision Desk: United States, 2026, calls published 2 November 2026. Poll averages are the Desk’s, as of 1 November.