When we began writing Holy Shamoley! Part 1 around the beginning of September, about eight weeks before Election Day, we started looking closely at what the polls and election analysts were showing.

As we began working on Part 2 at about 40 days out, we discovered something particularly useful: we could compare the 2026 national polling environment with where it stood at roughly the same point before the 2018 midterms.

The comparison got our attention. But 40 days is a long time in an election, so we kept checking.

Now we’re about a month out. Individual polls have moved, sometimes considerably. But the broader picture has remained striking.

So let’s look again.

About 40 days out: 2018 versus 2026

We’re comparing 2026 with 2018 because 2018 was the last midterm election during a Trump presidency. That makes it a useful historical comparison, but not a prediction.

At approximately 40 days out in 2018, Democrats held about an eight-point advantage on the generic congressional ballot, the national polling question that asks voters which party they would support for Congress.

At our comparable 2026 checkpoint, the Democratic advantage was also in that general range.

What happened afterward in 2018?

The Democratic advantage remained substantial. By Election Day, the final RealClearPolitics polling average was D+7.3. Democrats ultimately won the national House vote by about 8.4 points in RealClearPolitics’ final-results calculation and took control of the chamber.

That history is useful, but the relationship is not mechanical. District boundaries are different. Candidates are different. Incumbency matters. Where each party’s voters live matters. Polling error matters.

2018 is our benchmark, not our prediction.

2018 and 2026: Checking the Generic Ballot

Checkpoint	2018	2026
About 40 days out	Democratic advantage about 8 points	Democratic advantage in roughly the same 8–9 point range
About one month out	No directly comparable RCP daily average recovered	D+8.3 (50.6%–42.3%)
Election Day	Democrats won the national House vote by about 8.4 points	?

The missing 2018 one-month number is intentional. We were unable to recover a directly comparable RCP average for that date, and we won’t manufacture one from individual polls.

But something very different happened in the Senate

Despite the strong Democratic national environment in 2018, Democrats did not take the Senate.

Republicans increased their majority.

The problem was the map. Democrats were defending an unusually large number of seats, including seats in states Trump had carried.

That is why a strong generic ballot by itself cannot answer the Senate question.

And that is where 2026 has become particularly interesting.

Look at the Senate map now

FiftyPlusOne 2026 Senate forecast via 270toWin. This map updates as the forecast changes.

Republicans currently hold a 53–47 Senate majority, counting the independents who caucus with Democrats on the Democratic side. A change in control therefore requires a net Democratic gain of four seats.

The notable development is the number of Republican-held seats that have become competitive.

That does not make any particular combination of four pickups inevitable. A larger map simply creates more possible combinations.

Democrats also have seats of their own to defend, notably Michigan and New Hampshire.

So the basic Senate problem remains:

Win enough Republican-held seats without losing Democratic-held ones.

Wait. Texas? Iowa? Alaska? And now Kansas?

Yes.

And this is where Holy Shamoley! continues to earn its name.

Texas

Texas has remained competitive.

A Fox News poll released October 2 has Democrat James Talarico at 51% and Republican Ken Paxton at 49% among likely voters. The two-point difference is within the poll’s margin of error.

The significance is not that a two-point poll settles Texas. It clearly does not.

It is that Texas is still appearing as a competitive Senate race a month before Election Day.

Iowa

Iowa still comes with our flashing:

MORE DATA PLEASE.

The newest polls continue to disagree. The Fox News poll released October 2 has Democrat Josh Turek ahead 49% to 47% among likely voters, while another recent likely-voter poll has Republican Ashley Hinson ahead by one.

So Iowa remains competitive, and uncertain.

Alaska

Our earlier advice survives beautifully:

Watch Alaska. Don’t engrave anything.

An October 2 Quantus Insights poll has Republican Sen. Dan Sullivan at 48% and Democrat Mary Peltola at 47%.

That is almost a textbook illustration of why we should call a race competitive rather than pretend a poll tells us who will win it.

Kansas

Kansas is the new eyebrow raiser.

Recent polling has shown an unexpectedly close race between Republican Sen. Roger Marshall and Democrat Adam Hamilton. An October 2 Trafalgar poll, for example, has them tied at 44%.

Then something happened that may be even more revealing than a single poll.

On October 2, the Senate Leadership Fund, a major Republican Senate super PAC, shifted substantial resources into Kansas while pausing further advertising in North Carolina. The group reserved millions of dollars in Kansas advertising to support Marshall.

Campaign spending decisions are not election results either. But they tell us where political organizations believe resources may be needed.

Kansas has therefore joined the broader list of states worth watching.

What are the election models seeing?

This is where we return to something with the wonderfully improbable name FLIPR.

FLIPR is Silver Bulletin’s 2026 election model, created by Nate Silver and Eli McKown-Dawson. It combines polling with other information and runs simulated elections to estimate the range and frequency of possible outcomes.

It is a model, not a crystal ball.

When FLIPR launched in August, its headline model put Democratic Senate-control chances around 57%. In early September, that fell to essentially 50–50. By September 20, the three versions of the model had moved into roughly the 63%–69% range.

FLIPR was updated again October 2. Its polling averages account for factors including likely-versus-registered-voter samples, pollster reliability, sample size, recency and changes in the generic ballot. The current topline control probabilities are behind subscriber access, so we are not going to pretend we have numbers we cannot verify.

The useful lesson from FLIPR remains the same:

Watch why a model moves, not merely what number it produces.

And different forecasters can examine much of the same evidence and assess the uncertainty differently.

Inside Elections, for example, made another round of ratings changes on October 1. Among them, Alaska’s Senate race moved from Tilt Republican to Toss-up, while North Carolina moved from Tilt Democratic to Lean Democratic.

That is precisely why we use several kinds of evidence.

And what about the House?

The House picture has also continued to move.

On September 25, the nonpartisan Cook Political Report changed its ratings in 15 House races.

Every one of those changes was in the Democratic direction.

After those changes, Cook listed 185 seats as Solid Democratic and another 23 as Likely or Lean Democratic, for 208 altogether. On the Republican side, 176 were Solid and 29 Likely or Lean, for 205. The remaining 22 races were rated Toss Ups.

Then Inside Elections made another round of House changes on October 1. Its changes included Iowa’s 1st and 3rd districts moving from Toss-up to Tilt Democratic, California’s 22nd moving from Tilt Republican to Toss-up, and several Republican-held districts becoming more competitive.

Republicans currently hold a 220–215 House majority, so Democrats need a net gain of three seats.

None of those ratings determines what voters will do. They show how independent election analysts have changed their assessments as new information has arrived.

So, did the picture hold?

Broadly, yes.

That does not mean every number improved. It did not.

It does not mean every close race will move in the same direction.

And it certainly does not tell us what the result will be on November 3.

What we can say is narrower and more interesting.

When we checked at about 40 days out, the national polling environment showed a substantial Democratic advantage, the Senate map had expanded, and House ratings were moving.

We kept checking.

About a month out, the national generic-ballot advantage remains in roughly the same range. Texas, Iowa and Alaska remain competitive. Kansas has attracted serious national Republican spending. Other traditionally Republican states are appearing in polling and campaign-resource decisions. House handicappers have continued moving individual races.

So the thing that made us say Holy Shamoley at 40 days out did not disappear during the next week.

But the question mark belongs there too.

Polls have margins of error. Polling errors can be correlated across states. Different pollsters produce different results. Campaign spending changes. Events happen. Turnout is unknown until people actually vote.

And 2018 remains a particularly useful warning: a national political environment can strongly favor one party in the House while producing a very different result in the Senate.

The evidence can tell us what the landscape looks like now. It cannot tell us what voters will ultimately do with it.

On November 3, we get the answer.

Holy Shamoley, indeed.

 

---

About our polling and forecasting sources

RealClearPolling (RCP): A long-running polling aggregator that combines multiple public polls into polling averages. We use averages rather than relying on a single poll because they can provide a broader picture of a race. Its historical data also allows comparison with previous election cycles.

Silver Bulletin / Nate Silver: Silver Bulletin produces polling averages and the FLIPR election model. Its polling methodology considers factors including pollster reliability, sample size, recency and the type of voters surveyed. We use it as another way of examining changes in the polling environment, not as a prediction of what will happen.

Cook Political Report: An independent, nonpartisan publication that has analyzed U.S. elections for decades. Its race ratings incorporate the political characteristics of states and districts, candidates, the national environment and reporting from campaigns. We use Cook primarily as an expert race-rating source rather than as a polling average.

Inside Elections: A nonpartisan election-analysis publication that rates House, Senate and gubernatorial contests. We use its October 1 ratings changes as another independent measure of how analysts’ assessments of individual races have changed.

Individual polls: Where we discuss Texas, Iowa, Alaska or Kansas, we identify individual poll results because they help illustrate why those races are receiving attention. A single poll is a snapshot, not an election result, and differences within a poll’s margin of error should not be treated as establishing which candidate is ahead.

Campaign spending: We also discuss the Senate Leadership Fund’s October 2 shift of advertising resources toward Kansas. Spending decisions can provide evidence about where campaigns and political organizations believe resources are needed, but they do not establish how a race will turn out.