On the morning of August 19, the Kalshi contract on Stefany Shaheen winning the Democratic nomination in New Hampshire's 1st congressional district traded at 82 cents. A contract like this pays one dollar if it comes true and nothing if it doesn't, so 82 cents is the market's way of saying "about an 82 percent chance." As of the fetch behind this page, that same contract last traded at 57 cents.
No result was announced in those five days and nobody left the race. One poll came out, and one attack ad ran. We then handed nine AI models every public poll of this race, the full nine-name ballot, the 2018 result in this same district, and the documented record of the past two weeks, while showing them no prices at all. They came back with a board that says the money is still behind.
The Five Days That Took 25 Cents Off Stefany Shaheen
Saint Anselm College released its August survey on the evening of August 20. It had been in the field on August 17 and 18, and in the Democratic primary for this seat it found Maura Sullivan at 26 percent and Shaheen at 24, with Hampton selectwoman Carleigh Beriont at 12, state Rep. Heath Howard at 7, attorney Christian Urrutia at 3, and 27 percent of voters still undecided. The Democratic subsample was 359 likely voters with a margin of error of plus or minus 5.2 points, which makes 26-24 a tie rather than a lead.
It was still the first poll in the history of this race that Shaheen did not lead.
Two things hit this market inside three days, and it treated them very differently.
The first was the ad. On Tuesday, August 18, Shaheen ran the race's first negative spot of the cycle, which the New Hampshire Union Leader reported that afternoon. The next morning Shaheen's contract cracked, sliding from 82 to 77 by 11am Eastern and touching 71 by mid-afternoon. Then it closed the day back at 80. Traders looked at a front-runner going negative, moved eleven cents, and put nine of them back before the close.
The second was the poll, released Thursday evening. Shaheen's contract closed at 80 at 20:00 UTC on August 20. Twenty-seven hours later it closed at 61, and this time nothing came back. Sullivan's contract did something stranger. It did not trade at all on August 19 or for most of August 20, its last print stuck at 15 cents while the book quoted it 16 bid against a 22 ask. Then 520 contracts went through in the 22:00 UTC hour on August 20, at prices from 22 up to 27, and by the next hour it closed at 35. The slide kept going: Shaheen traded down through the 60s across August 22 and reached 57 by midday on August 23, where it has sat ever since. Over five days the gap between the two on last-traded prices went from 67 cents to 15.
Worth noticing what that ordering means. Saint Anselm was in the field on August 17 and 18 and the ad aired on the afternoon of the 18th, so the survey that moved this market by 19 cents was collecting most of its answers before, or alongside, the spot it now gets discussed with. The poll is not a reading of how the ad landed.
Here is where the board sits at the moment this page was built.
The prices above are last-traded prices, which is also what the tape in this article is made of. Two more numbers are quoted on every contract: the bid, which is the best price someone will pay you for one, and the ask, which is the best price you can buy one at. Shaheen is 57 bid against a 58 ask and Sullivan is 41 against 42, both about as tight as this exchange gets, on roughly 14,100 and 9,700 contracts of lifetime volume. Carleigh Beriont's contract is the thin one: 2 bid against a 4.1 ask on 1,825 contracts, so a round trip there costs more than half the position, and her last trade at 2 sits at the bottom of that spread rather than in the middle of it.
Each contract's rule is one sentence long: it resolves Yes if that candidate wins the Democratic nomination for this seat. Kalshi's contract language does not name the authority it will read the result from; the exchange's own event listing points at the Democratic Party of New Hampshire, and there is no published contract-terms document for this series that says more. The exchange lists an expected expiration date of September 8, the day of the primary itself.
Prices are as of August 24, 2026. Kalshi is a CFTC-regulated event-contract exchange; participation is 18+ and limited to eligible states, and availability varies by state as of August 24, 2026. This page takes no side in any race and analyzes prices only. New to this kind of market? Start with how election prediction markets work.
Free: The Weekly PM Market Brief — the AI panel's graded record, the week's biggest market-vs-model gaps, and what's spiking next. One email, Sundays.
The ad itself is worth reading, because it is the sharpest thing either campaign has said out loud. The 30-second spot accused Sullivan of hiding that "90% of her campaign money comes from outside our state. $300,000 bundled from AIPAC donors, $380,000 from Jeff Bezos's PAC, even from Trump's ICE contractor," and closed on "Let's show them, New Hampshire's not for sale," per the Union Leader's account above.
Sullivan's campaign manager Natalie Silver called the spot desperate and said Shaheen should be "ashamed" of it, telling the paper that Sullivan "doesn't take a dime from corporate PACs, ICE vendors, or AIPAC," and adding that "these desperate attacks are particularly rich coming from Stefany, who has taken over $158,000 from DC lobbyists, including pharma and gun lobbyists." Two days later, Jewish Insider reported that Sullivan spokesperson Emmett Avery said "Maura has not received any money from AIPAC and would not accept its endorsement if offered," and that Shaheen spokesperson Emma Greenberg said Sullivan is "beholden" to AIPAC. Jewish Insider also reported that AIPAC itself has made no endorsement in this race and has not been directly involved, and that the pro-Israel group NORPAC held a fundraiser for Sullivan.
Several of our models read the timing of that ad the same way, and it is worth naming plainly: front-runners with comfortable internal numbers do not usually fire the first negative shot of a primary three weeks out.
The Number That Moved Seven Of Nine Forecasters
Our nine models each priced all nine candidates in a single pass, then read each other's anonymized reasoning and were invited to revise. In that second round, seven of the nine independently landed on the same piece of arithmetic, and none of them got it from us.
Between the June Saint Anselm poll and the August one, the undecided share fell 20 points, from 47 to 27. That is the largest block of votes that moved in this race all year. Here is where it went:
| Candidate | June 24-25 | August 17-18 | Change |
|---|---|---|---|
| Maura Sullivan | 15% | 26% | +11 |
| Carleigh Beriont | 3% | 12% | +9 |
| Stefany Shaheen | 22% | 24% | +2 |
| Heath Howard | 9% | 7% | −2 |
| Christian Urrutia | 3% | 3% | 0 |
| Sarah Chadzynski | 1% | 0% | −1 |
| Undecided | 47% | 27% | −20 |
Twenty points came off the fence. The six candidate lines Saint Anselm breaks out account for nineteen of them; its August release does not itemize where the twentieth went. Shaheen collected two. She has universal name recognition in this district, the endorsement list, and a sitting United States senator for a mother, and when a fifth of the electorate finally made up its mind, she got a tenth of it.
That is the number the panel kept returning to. GPT put it as "those assets have been visible all cycle and have not converted the undecided bloc." Grok called it the fact that "directly weakens the thesis that remaining late and undeclared voters default to the famous name."
The honest caveat, which most of the seats raised themselves: the June poll sampled 398 registered voters and the August poll sampled 359 likely voters. Tightening a screen from "registered" to "likely" changes who is in the sample, so some part of that 20-point release is method rather than mind-changing. Kimi held its Shaheen number partly on this ground, writing that the decomposition "is partly confounded by the RV-to-LV" change. It is a real objection. It is also not big enough to turn +2 into a front-runner's share.
Hottest Prediction Markets Right Now
- 2028 Democratic presidential nominee · $190M traded
- 2027 Pro Football Champion · $70M traded
- 2028 U.S. Presidential Election winner? · $62M traded
- 2028 Republican presidential nominee · $59M traded
- Pro Baseball Champion · $59M traded
- Bitcoin price at the end of 2026 · $32M traded
Every market above links to our full AI model verdict; browse them all on the OddsShopper prediction markets hub, and see how every settled call actually scored on the full graded scoreboard.
Maura Sullivan Has Run This Exact Race Before
There is no runoff in New Hampshire. Whoever gets the most votes on September 8 out of nine names on the ballot is the nominee, even if that is 30 percent of the vote. In a field this fragmented, the winning number is small.
We know roughly how small, because this district ran the same experiment eight years ago with the same person in it. In the 2018 Democratic primary for NH-01, eleven candidates split 63,619 votes. Chris Pappas won with 26,875 votes, or 42.2 percent. Maura Sullivan finished second with 19,313 votes, 30.4 percent. Nobody else cleared 6,500.
Sullivan is the only candidate in the 2026 field who has been through this. She lost that year to Chris Pappas, who had already won three different offices in New Hampshire since 2002, state representative, then Hillsborough County treasurer, then executive councilor from 2013, and who carried the backing of nearly every elected Democrat in the state. Sullivan lost while raising 96.7 percent of her individual contributions from outside the state, $1.37 million against $46,648 raised at home, which InsideSources reported was a higher out-of-state share than any other major U.S. House candidate in the country that cycle. The carpetbagger charge followed her all summer. She still finished with 30.4 percent.
There is a small piece of New Hampshire symmetry buried in that result. Pappas got his own start in politics in 1996, volunteering on the gubernatorial campaign of a state senator named Jeanne Shaheen.
The panel treated that 30.4 as a benchmark rather than a ceiling. GLM went further and called it a "proven 2018 floor in this district," which is a stronger word than one prior result can carry. Either way, a candidate who has already reached the low 30s here once is most of the way to a number that wins a nine-way plurality.
One more mechanic decides who actually shows up. New Hampshire runs a semi-closed primary: voters registered as undeclared, the largest registration bloc in the state, can walk in on September 8, ask for a Democratic ballot, vote, and change back to undeclared before they leave. Anyone already registered with a party had until June 2 to switch. So the size and shape of the Democratic electorate is not fixed in advance, and every "likely Democratic primary voter" screen is a pollster's guess about how many undeclared voters turn up.
Where The Panel Landed
Nine models, none of them shown a price, each pricing all nine candidates from one view of the race, then revising after reading the others. The blend below is the median of the nine seats after that revision round.
| Outcome | Last Trade | AI Blend | Gap (blend% − mkt¢) |
|---|---|---|---|
| Stefany Shaheen | 57¢ | 36.0% | −21.0 |
| Maura Sullivan | 42¢ | 40.0% | −2.0 |
| Carleigh Beriont | 2¢ | 10.0% | +8.0 |
| Everyone else (six names) | not listed | 12.4% | +12.4 |
These are model estimates, not predictions of fact and not financial advice.
The panel and the market have almost exactly the same opinion of Maura Sullivan. They disagree about everything else. Every number in this piece gets graded in public once the market settles, against the full graded scoreboard.
Every Seat's Number
| Model | Shaheen | Sullivan | Beriont | Field |
|---|---|---|---|---|
| Claude Fable | 47.0% | 42.0% | 7.0% | 4.0% |
| Claude Opus | 39.4% | 42.4% | 11.1% | 7.1% |
| Claude Sonnet | 32.3% | 36.4% | 13.1% | 18.2% |
| GPT | 36.0% | 42.0% | 10.0% | 12.0% |
| Gemini | 38.0% | 48.0% | 8.0% | 6.0% |
| Grok | 34.0% | 40.0% | 9.0% | 17.0% |
| Kimi | 39.2% | 37.1% | 11.3% | 12.4% |
| GLM | 34.7% | 37.8% | 10.2% | 17.3% |
| DeepSeek | 30.0% | 36.0% | 10.0% | 24.0% |
Every seat on this panel is graded against real market settlements — records to date: Claude Fable 82% on 616 graded calls · Claude Opus 83% on 686 graded calls · Claude Sonnet 81% on 664 graded calls · GPT 84% on 5,836 graded calls · Gemini 86% on 4,570 graded calls · Grok 83% on 3,928 graded calls · Kimi 82% on 2,442 graded calls · GLM 81% on 2,465 graded calls · DeepSeek 80% on 2,490 graded calls. Recomputed daily; the full scoreboard is public.
These are model estimates, not predictions of fact and not financial advice. Each seat's own nine numbers were required to sum to 100. The blend column is the median across the nine seats candidate by candidate, and a set of column-wise medians does not have to total 100, which is why it comes to 98.4.
Read down the Shaheen column and the disagreement is inside our own panel too. Claude Fable is at 47 and DeepSeek is at 30, a 17-point spread on the same evidence. Not one seat put her where the money has her.
More live boards from the same panel: Chris Pappas in the New Hampshire Senate primary · 96¢ · the New Hampshire Senate general election · Democrats 89-90.9¢ · Alaska's Senate race · Democrats 61-62¢. Prices fetched August 24, 2026.
Carleigh Beriont And The Six Names The Board Leaves Out
Kalshi lists three contracts on this primary. Nine Democrats are on the ballot. Add up the three midpoints, halfway between each bid and ask, and you get 102.05 cents. Three YES prices that already total more than a dollar leave nothing over for the six names nobody can buy, so the only way to back the field here is to buy NO on all three at once and hope none of them wins.
Each seat priced all nine names, so the panel has a number for the six you cannot buy. Here is the median of the nine seats for each of them, next to what Saint Anselm measured in August:
| Unlisted candidate | August poll | Panel median |
|---|---|---|
| Heath Howard | 7% | 6.0% |
| Christian Urrutia | 3% | 3.0% |
| Sarah Chadzynski | 0% | 2.0% |
| Bill Conlin | not reported | 1.0% |
| Matthew Emerson | not reported | 0.5% |
| Sarah Bella Spinosa | not reported | 0.5% |
Heath Howard carries nearly half of that unlisted 12.4 on his own. He polled 7 percent in August, holds endorsements from 31 state legislators, and has run hardest at the party's left flank. If he, or Urrutia, or anyone else outside the three wins on September 8, all three listed contracts settle at zero at once. That outcome has no YES contract of its own on this board; it is priced only in the shadows of the three NO sides.
The other candidate on this board the models call mispriced, and the only one they think is too cheap, is Carleigh Beriont. Her last trade is 2 cents and the midpoint of her book is 3.05. The panel's median is 10, and every seat on the panel is above the market: the lowest is Claude Fable at 7 and the highest is Claude Sonnet at 13.1.
The case for her is one number. She went from 3 percent in June to 12 in August, taking nine of the twenty points that came off the fence, a bigger share than Shaheen took. The case against her is that it is also only one number, from a 359-person sample with a 5.2-point margin, off a 3 percent baseline. One poll can be the first reading of a real move or it can be one poll. Gemini revised her upward in the second round on the view that her 12 percent is "likely real progressive absorption rather than pure LV-screen" artifact. Nobody on the panel made her a favorite; several of them made her the reason Sullivan and Shaheen both stay under 45.
There is also a longer-running item in the file. In February, InDepthNH reported that Shaheen's role as chief strategy officer at ARMI, the federally funded biomanufacturing institute founded by Dean Kamen, had become a campaign issue after Justice Department files showed Kamen had regular email contact with Jeffrey Epstein after Epstein's 2008 conviction. ARMI received $80 million from the Defense Department in 2017 and $44 million from the Economic Development Administration in 2022, and Senator Jeanne Shaheen sits on the appropriations subcommittee overseeing the department that funded it. Stefany Shaheen said "the ARMI Board needs to address this matter independently." Howard said "moral clarity is not optional." That story has been available to voters for six months and Shaheen still led every poll until last week, which is roughly what the market has been pricing.
Where The Panel Changed Its Mind
The second round is where this panel earns its keep. All nine seats revised their numbers, and seven of them named the same cause for it.
"Moved Shaheen 52 to 47 and Sullivan 38 to 42. What moved me was the decomposition of the undecided collapse, which I had not weighed explicitly and which is a ceiling signal independent of the screen-change caveat. I did not flip the top two: the strongest opposing argument, that the first attack ad proves her internals show Sullivan ahead, is consistent with a campaign that is still ahead and knows the trend." — Claude Fable, one of the two seats that still had Shaheen in front after the revision round
"I now make Sullivan the narrow favorite. The decisive new insight is the decomposition of the undecided collapse. This pattern, late deciders breaking toward Sullivan and not toward the best-known name, flips my prior assumption." — DeepSeek, which moved Sullivan from 30 to 36
Claude Opus moved the other way, and for a reason nobody else raised:
"Trimmed Sullivan 45 to 42 and raised Shaheen 37 to 39, narrowing my spread from 8 points to 3. A peer-raised asymmetry I had not priced: an active sitting U.S. Senator can personally campaign for a daughter in the final two weeks, a closing-consolidation lever with no Sullivan equivalent." — Claude Opus, whose stored board rounds to the 42.4 and 39.4 in the table above
What Would Change The Panel's Mind
Four things are checkable between now and September 8, and each one pushes the number a specific way.
- A new public poll fielded after August 18. As of this page there is not one. If a second pollster reproduces Sullivan ahead and Beriont near 12, the panel's read hardens and Shaheen's 57 has further to fall. If a new poll puts Shaheen back up six or more, the August survey looks like a screen artifact and the market's price was right all along.
- Jeanne Shaheen campaigning in-district in the closing fortnight. This is the lever Claude Opus priced and nobody else did. Visible, dated, and easy to check.
- A consolidation move by one of the six unlisted candidates. Howard at 7 percent or Urrutia at 3 dropping out and endorsing changes the plurality math directly, and it is the single event most likely to move Beriont's two-cent contract.
- Whether the ad war escalates or stops. The AIPAC exchange is on the record and dated. A second negative ad from either side inside the last two weeks tells you more about internal polling than any public survey will.
Settlement Timeline
| Date | What happens |
|---|---|
| June 2, 2026 | Party-affiliation change deadline (passed) |
| August 18, 2026 | Last day any public poll of this race was in the field |
| September 8, 2026 | New Hampshire state primary; polls close and the nomination is decided by plurality |
| September 8, 2026 | Contracts are expected to expire and settle |
| November 3, 2026 | General election for the seat |
This board re-scores whenever the story moves. All prices and figures on this page were fetched August 24, 2026.
The Bottom Line
The market did something impressive last week: it took a poll seriously, fast, and moved 25 cents in five days without waiting for anyone to tell it what the poll meant. The question this page is asking is whether it moved far enough.
Nine models that never saw a price say no. Their median puts Stefany Shaheen at 36 percent against a 57-cent contract, and the reason is a single line of arithmetic anyone can check: when 20 points came off the undecided pile between June and August, the best-known candidate in the district collected two of them. The counter-argument is real, and two seats held it. Claude Fable kept Shaheen in front on her name and institutional edge, and read the attack ad as what a campaign that is still ahead does when it sees the trend coming. Claude Opus added the lever nobody else priced, a sitting senator free to campaign for her daughter in the closing fortnight. One poll inside its own margin of error is a thin foundation for a 21-point disagreement.
That is the disagreement the price is carrying into September 8, when the votes settle it. Whatever happens, every number in this piece is already recorded and will be graded in public against what the votes actually say.
The panel's calls are on the record. Our AI panel puts a graded, public verdict on the New Hampshire Senate race the winner of this primary will share a ballot with — see the live model verdict.
