Captain K’s Orientation to 2026 Midterm Forecasting

Under the hood of the most successful model for the 2024 presidential election and how I’ll adapt it to account for greater uncertainty

While some prefer to overcomplicate things, simplicity is my speciality, and I have the accolades to prove my forecasting know-how. For the 2024 presidential election, I divided the races into three categories:

• SAFE - Race certain to go to either candidate, with no chance of flipping. I settled on a margin of >8% points for these races, and only one state snuck inside that point - New Jersey at Harris +5.9%. Called all 43 winners correctly.

• LEANER - Race in which the likely winner was known to a high degree of certainty, but with the right circumstances, a flip would be possible. Margin of victory <8%. Called all 6 correctly:

• DECISIVE - 81 electoral votes most likely to decide the presidential winner, taking into account allocation of safe and leaner races. Called all 7 correctly:

Disclaimer: Presidential races are far easier to predict because races are much more partisan and rigidly follow voter registration by party shifts every four years. Midterm, off-year, and special elections are subject to far more variables and decades-old political science dynamics that has more to do with which party controls the White House than anything else, but can also boil down to which issues are hot at the right time. If the electorate is looking for Santa Claus, Democrats are tailor-made to run on handing stuff out. If they’re looking for law and order, Republicans take the cake.

The 2026 midterms inherently favor the Democrats, speaking in terms of the U.S. House only. The Senate, as you’ll continue to see, favors a Republican hold because the map is a bad one for Democrats. The House majority, as elected in 2024, is just 220 to 215 Republican, meaning they can lose just a net of two seats and still hold the chamber. The average loss to the President’s party since 1934 is 27 seats, with just 3 of 23 midterms going in his favor.

I will not be able to present full projections until all maps are set. This may take a while, because yesterday a rogue judge threw out Texas’s 2026 map, in which Republicans were looking to flip up to five seats. I think this sets up for a battle in which California’s redistricting process, which is extremely sketchy, will be called into scrutiny. I don’t foresee the 2026 elections happening with Texas running on the old map and California on a new one. I could see it the other way around, but most likely, I think they’ll either both have new maps, or both have to keep the old ones.

With 435 House seats up for grabs, I’d be stunned if I batted 1.000 this time around. Many of these races will be decided by thousands, if not hundreds, of votes (or “ballots” in states like California), and late-breaking news in the early voting period could swing them away from expectations.

Here is how I am planning on fielding my projections:

---

SAFE SEATS

For now, margins of >10% in 2024. Subject to change based on new boundaries, retiring incumbents, or signs of a significant wave favoring Democrats that could jeopardize Republican seats previously thought safe. I think polarization and the continuing Republican registration edge is likely to prevent a blue tide, if it manifests, from reaching these seats.

Examples:

WA-1 is SAFE for Democrats (hold). No Republican can win this seat in any election as currently drawn.

IA-4 is SAFE for Republicans (hold). No Democrat can win this seat in any election as currently drawn.

---

LEANER SEATS

Seats mostly decided within 10%, but not close, in 2024. Some held by Democrats may drift over 10% margins with a Democrat tide, and some held by Republicans may get close or flip, but I feel more confident in who will win these seats than I do over the final category, which will be listed next.

IN-1 is a LEANER favoring Democrats (hold). As currently drawn, and with a Republican President’s midterm looming, I would expect it to remain out of reach. It would take a red wave election, which is most likely in a Democrat President’s midterm, to flip it under normal circumstances, though the margin is close enough it can’t be ruled out as a SAFE seat. Indiana’s legislature could draw this into a Republican flip if they had the willpower to do so.

AZ-2 is a LEANER favoring Republicans (hold). With partisan loyalties strong in this district, it would take a serious blue wave to flip, which isn’t likely (if at all) to be strong enough to take AZ-2 even if Democrats have a strong performance next fall. Still, the 2024 margin, concurrent with a presidential, requires a campaign to stave off complacency.

---

DECISIVE SEATS

With very few, if any, misses likely on SAFE seats, and a high degree of confidence in the outcomes of LEANER seats, the majority will come down to how the DECISIVE seats play out. The good news is, in my initial pass over the map, the Democrats are on defense in these seats far more than Republicans are (hence the narrow majority).

Here are some textbook examples:

ME-2 is a DECISIVE seat currently held by a Democrat. Golden is retiring from Congress, and former Governor Paul LePage is running for the seat. I like the odds for a Republican flip as long as Democrats don’t pick up wave momentum, even though Maine has, by far, the worst elections (including Ranked Choice Voting) in all New England. Trump carried this split electoral district in all three of his runs.

NE-2 is a DECISIVE seat currently held by a Republican. Bacon is also retiring, leaving this important seat, which the idiotic Nebraska legislature has left competitive in a state Trump won by more than 20 points, without an incumbent. It gives me KS-3 vibes and I wouldn’t be surprised to see this seat flip next fall even if Republicans play solid defense. Democrats will be all over it like flies on a cow patty.

---

CONCLUSION

If the U.S. House consisted of just 6 seats, I’d have this as my forecast today:

• 2 Republican (1 Safe/1 Leaner)

• 2 Democrat (1 Safe/1 Leaner)

• 2 Toss-Up

I’d probably flip ME-2 to Republican and NE-2 to Democrat, giving it a 5-5 prediction for the slate. Bucketing all 435 races, plus the 33 Senate races, into three categories will help make sense of all the chaos and polling noise you’ll hear between now and November 3, 2026. This is a moving target, given so many maps up in the air, retirements yet to be announced, and radical candidates yet to be field through primaries, so all assessments are subject to change. I plan to assess the SAFE seats first (in states not redrawing maps) and may begin trickling those out soon.

If you feel bold, you can make a lot of money putting a pro-GOP bet out there on Polymarket, which currently gives them better odds than history does for holding the House.

---

Seth Keshel, MBA, is a former Army Captain of Military Intelligence and Afghanistan veteran. His analytical method of election forecasting and analytics is known worldwide, and he has been commended by President Donald J. Trump for his work in the field.