On August 18, the Kalshi contract on Darline Graham winning South Carolina's Republican Senate runoff closed at 90 cents. Five days later it closed at 66. Something moved 24 points in a state where almost nothing moves.
That same week, on the same exchange, the market on which party ends up holding the seat closed at 89 cents Republican on August 18 and trades at 89 cents Republican today. Between those two dates it never traded outside a 1.6-cent band.
Two boards, one Senate seat, one news cycle. One of them tore itself apart and the other did not blink. Traders are saying something specific with that gap, and it is worth knowing whether they are right before November 3.
New to event markets? Start with how election prediction markets work.
Stay ahead of the markets.
Daily insights and expert picks on Kalshi, Polymarket, and what's moving markets.
Free forever. Unsubscribe anytime.
The Seat That Changed Hands Without An Election
Lindsey Graham won his Republican primary in June with 56.8 percent of 465,076 votes. He died on July 11, after 23 years in the Senate. Because he died after winning a primary, South Carolina law required the party to pick a substitute nominee through a whole second primary rather than by committee vote, which is why a state that had already settled this race in June found itself running a compressed three-week Senate campaign in August.
Two days after the death, on July 13, Governor Henry McMaster appointed the late senator's sister, Darline Graham. President Trump had publicly recommended her that same day. She was sworn in on July 14 as the first woman ever to represent South Carolina in the Senate, having never held public office before.
The structural picture she inherited is one of the most one-sided in the country. Trump carried South Carolina by 17.87 points in 2024, 58.23 to 40.36. No Democrat has won a South Carolina Senate race since 1998. Republicans hold every statewide office except comptroller, supermajorities in both legislative chambers, both Senate seats, and all but one House seat.
And there is a precedent here that every forecaster on our panel reached for independently. In 2020, Democrat Jaime Harrison raised record-shattering money against Lindsey Graham, late polling showed a dead heat, and Graham won by 10.2 points, 54.4 to 44.2. A well-funded Democrat plus close polling plus a nationalized red-state electorate produced a double-digit loss. Hold that comparison; it is doing most of the work below.
The case for each side of the South Carolina (SC) Senate race, priced on Kalshi: Why Darline Graham (R) Will Win South Carolina's Senate Race · Why Annie Andrews (D) Will Win South Carolina's Senate Race
Darline Graham And The Debate That Moved The Price
Nobody cleared a majority on August 11. In a ten-candidate field on 335,555 votes, Graham took 32.8 percent and Representative Ralph Norman took 24.6, sending the two of them to an August 25 runoff. Russell Fry finished third at 19.7 and Mark Sanford fourth at 14.9.
For a week, the runoff looked settled. Graham's contract climbed from 44 cents on primary night to 90 cents on August 18, carrying Trump's recommendation plus endorsements from Senator Tim Scott, who chairs the Senate Republicans' campaign arm, and Majority Leader John Thune.
Then came the debate. Asked about United States policy toward Taiwan and the South China Sea, Graham answered "I'm not on national security," and added that "national security is not my thing." In the same debate she put South Carolina's population at 50 million people; the Census Bureau put it at 5,570,274 as of July 2025. The Hill reported that she was pilloried for saying she was not that informed on national security. NBC News covered her defending her résumé afterward. Three days later CNN reported South Carolina Republicans split over the performance.
The backlash ran through her own party rather than around it. Graham was criticized less for a position than for not appearing to have one, and the CNN report describes South Carolina Republicans arguing publicly about whether an appointed senator who had never run for anything should be their nominee ten weeks before a general election.
The price record is unusually clean, because early voting in the runoff opened on August 19, the morning after. Graham went 90, 82, 80, 71, 69, 66 on consecutive daily closes and sits at 67 cents now. Norman went the other way over the identical window, from 11 cents to 35. No other campaign event in that window was reported at anything like the same scale.
So the exchange rendered a verdict on the debate, and it was a heavy one: roughly a quarter of the probability of the Republican nomination changed hands in under a week.
Ralph Norman And The Consolidation Math
Norman's path is arithmetic. He is a sitting congressman from the 5th District, in office since 2017, a House Freedom Caucus member campaigning on the SAVE Act. His problem on August 11 was that he finished eight points behind. His opportunity is that Fry and Sanford together took 34.6 percent of the first round, and neither of them is on the ballot on August 25. Runoffs are decided by which finalist inherits more of that orphaned vote, on much lower turnout.
His own record is a fact in this race the same way Graham's debate is. Norman voted on January 6, 2021 to omit at least one state's presidential electors from the count, and afterward urged President Trump to suspend civil law in order to remain in office. That record has been litigated in front of South Carolina electorates before; he has held the seat through every election since.
Here is the part that matters for the market on this page: the party contract does not care which of them wins. It pays on the party of whoever is sworn in. So the entire runoff, the debate, the 24-point repricing, all of it is only relevant to this board to the extent that the identity of the nominee changes the odds of a Republican holding the seat in November. Our panel was asked that question directly, and the answer is the finding of this piece.
Annie Andrews And The Only Polls That Exist
The Democratic nominee has been set since June 9, when Charleston pediatrician Annie Andrews won her primary with 61.5 percent over Brandon Brown's 30.2 and Kyle Freeman's 8.3. Andrews ran for the 1st District in 2022 and lost to Nancy Mace, 42.5 to 56.5. Through June 30 she had raised $10,804,041, spent $6,684,746, and held $4,119,295 in cash. Pete Buttigieg endorsed her on August 13.
Now the honest part about the polling, which is the single most important paragraph on this page.
Every general-election head-to-head poll on the public record for this race was paid for by Democrats. The most recent is a Public Policy Polling survey of 577 voters taken July 16 and 17, sponsored by the Andrews campaign. Within it: Norman 43, Andrews 42. Trey Gowdy 44, Andrews 41. Pamela Evette 41, Andrews 42. Sanford 40, Andrews 41. Fry 39, Andrews 42. Mace 39, Andrews 42. Mark Lynch 38, Andrews 41. John Warren 36, Andrews 42.
There is no independent poll. There is no Republican-sponsored poll. And there is no poll at all testing Darline Graham, the runoff favorite, against Annie Andrews. The candidate most likely to be the Republican nominee has never been measured against the Democrat in public.
But the most useful number in that survey is one the sponsor did not commission for a headline. Inside the same sample, a generic Republican beat Andrews 47 to 39. Every named Republican tested ran between three points ahead and six points behind. Naming the Republican cost between five and 14 points against a generic ballot, inside one poll, with one house effect applied evenly to all of it. That is a real measurement of candidate drag, and it is the strongest argument Andrews has.
Against it sits the ratings board. All ten public handicappers have this race on the Republican side, and the two freshest ratings both post-date the primary: Race to the WH at Likely R on August 18, and Sabato's Crystal Ball at Safe R on August 19, the day after the debate. The professionals watched the same week the market did and did not move.
What The Panel Says
We handed eight AI models the whole file: the certified ten-candidate primary result, the runoff rules and date, the sourced debate record, all ten handicapper ratings with their dates, the structural baseline back to 1998, the fundraising, the full polling list with its sponsorship problem stated plainly, and the national environment. No model saw a market price. Each priced the entire board in one pass, then read the other seven anonymized and revised.
| Outcome | Kalshi | Panel blend | Gap |
|---|---|---|---|
| Republican Hold | 89¢ | 88.0% | −1.0 |
| Democratic Flip | 12¢ | 11.5% | −0.5 |
| Neither Major Party | not listed | 1.0% | — |
These are model estimates, not predictions of fact and not financial advice. For anyone translating: 89 cents is roughly a −800 moneyline and 12 cents is roughly +730, before fees and the spread.
The panel and the market agree, and that agreement is the story. Eight models that never saw a price, reasoning from a file that contains the debate transcript quotes and a poll showing the race within one point, landed within a cent of a market that spent the same week violently repricing the other board. The exchange's judgment that Republican infighting is not a Republican loss survives an independent test.
Every Seat's Number
The blend is the median of the eight seats' round-two numbers, taken after each model read the others' anonymized reasoning and could revise.
| Outcome | Fable | Opus | Sonnet | ChatGPT | Gemini | Kimi | GLM | DeepSeek | Blend |
|---|---|---|---|---|---|---|---|---|---|
| Republican Hold | 86.5 | 89.0 | 88.0 | 88.5 | 88.0 | 88.0 | 87.0 | 87.0 | 88.0% |
| Democratic Flip | 12.5 | 10.5 | 11.0 | 10.8 | 12.0 | 11.0 | 12.0 | 12.0 | 11.5% |
| Neither Major Party | 1.0 | 0.5 | 1.0 | 0.7 | 0.0 | 1.0 | 1.0 | 1.0 | 1.0% |
These are model estimates, not predictions of fact and not financial advice. All eight seats returned in both rounds, price-blind, on August 24, 2026. Every number in this piece gets graded in public once the market settles; the full graded scoreboard carries the running record.
Every seat on this panel is graded against real market settlements — records to date: Gemini 86% on 4,746 graded calls · GLM 81% on 2,504 graded calls · Kimi 83% on 2,494 graded calls · DeepSeek 80% on 2,542 graded calls. Recomputed daily; the full scoreboard is public.
More live boards from the same panel: South Carolina's Republican runoff · Graham 67¢, Norman 35¢ · Georgia Senate 2026 · Ossoff 92¢ · Senate control 2026 · Republicans 54¢. Prices fetched August 24, 2026.
Worth saying plainly what the third row is. Kasie Whitener is on the November ballot for the Libertarians, Mark Hackett for the Constitution Party, and Catherine Fleming Bruce is running as a write-in. If any of them won, both listed contracts would resolve No, as would a party switch before the swearing-in. The panel prices that branch at about one percent, and seven of eight seats put something on it rather than zero, which is the correct treatment of a tail that exists.
Where The Panel Changed Its Mind
Round one spread from 85 to 92 on the Republican side. Round two spread from 86.5 to 89. The seats moved toward each other from both directions, which is the useful kind of convergence.
The biggest single revision came from Sonnet, which cut its Republican number from 92 to 88 after reading the others:
But all ten handicappers sit R, and the two freshest (Sabato Safe R 8/19, RttWH Likely R 8/18) post-date both the primary and the debate without moving. Honest band: DEM 7-17. — Sonnet
ChatGPT moved the opposite way, from 85.5 up to 88.5, after weighing the sponsorship problem harder:
I do not discard the polls, but every public head-to-head is Democratic-sponsored, so I shift them roughly 4 points toward the GOP and allocate undecideds with a Republican lean. That turns Norman +1 into a mid-single-digit R lead, while the generic Republican +8 in the same sample is a strong baseline warning. — ChatGPT
GLM went the other way again, from 85 to 87, and was the most explicit about what pulled it away from the structural prior:
What moves me off the structural prior toward DEM 12: the national environment is genuinely worse than 2020 (D+6.3 generic ballot, approval net -21.7, midterm base rate with no 1998/2002-style rally condition), the nominee will exit a runoff broke with ten weeks left against Andrews' $4.1M cash, the runoff favorite has never been polled against Andrews, and the named-Republican penalty (5-14 points below generic) signals real candidate drag. — GLM
Kimi barely moved, from 87 to 88, and said why the round did not persuade it:
Nothing in A-G was new evidence; the group's range brackets mine and G's near-zero-midterm-error point is already priced in. — Kimi
Every seat, without exception, named the 2020 Harrison result as its anchor. Notably, all eight also judged Norman the modestly stronger November candidate than Graham, and all eight put that difference at only a few points of party probability. That is the panel's real answer to the runoff: the identity of the nominee matters, and it does not matter nearly as much as the state does.
What Would Change The Panel's Mind
Four things, each checkable, each with the direction it pushes.
- Any Independent Or Republican-Sponsored Head-To-Head Poll. Seven of eight seats named this first. One showing Andrews tied or ahead pushes the Democratic number up by four to ten points; one showing the Republican up five or better pushes it down. Right now the entire public polling record has one sponsor.
- A Poll Testing Darline Graham Specifically. If she wins Tuesday, she becomes the first nominee in this race never measured against the opponent. Her downside is unmeasured rather than absent, which is exactly why several seats held probability back.
- A Handicapper Downgrade After August 25. Sabato and Race to the WH both held after the debate. If either moves to Lean R once a nominee exists, the panel moves with it.
- The National Environment. The generic ballot sits at Democrats +6.3 and presidential approval at net −21.7, with no rally condition of the kind that produced the 1998 and 2002 exceptions to the midterm base rate. Approval recovering above net −10 pushes the Democratic number down two to three points; the generic ballot widening past +10 pushes it up.
Settlement Timeline
| Date | What happens |
|---|---|
| August 25, 2026 | Republican runoff, Graham vs Norman. The winner is the nominee. This board settles on party, so it does not resolve here. |
| Early September 2026 | First realistic window for a general-election poll naming the actual nominee. |
| October 15, 2026 | Third-quarter FEC filings, the first look at what the nominee has to spend against Andrews' cash. |
| November 3, 2026 | Election day. |
| January 2027 | The winner is sworn in and the contract settles. |
This page re-scores when the story moves, starting with Tuesday's runoff result. Verdicts and prices above are as of August 24, 2026.
Hottest Prediction Markets Right Now
- 2028 Democratic presidential nominee · $191M traded
- 2027 Pro Football Champion · $71M traded
- 2028 U.S. Presidential Election winner? · $62M traded
- 2028 Republican presidential nominee · $60M traded
- Pro Baseball Champion · $59M traded
- Bitcoin price at the end of 2026 · $33M 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.
More on this: 2026 Senate Races: Every Kalshi Board We Track · South Carolina Senate Runoff Turnout: The 375,000-Voter Question · Why Darline Graham (R) Will Win South Carolina's Senate Race · Kalshi Odds: Why Darline Graham Will Win The SC Senate Runoff · When Is The SC Senate Runoff? Kalshi Odds On Graham-Norman
The Bottom Line
South Carolina spent a week having a loud argument about which Republican it wants, and the market that decides whether it will be a Republican at all did not participate. Our panel, which never saw a price, agrees with that: 88 percent Republican against 89 cents, 11.5 percent Democratic against 12, gaps too small to trade on.
The honest caveat is what nobody has. Annie Andrews has raised $10.8 million and is running against a nominee who will emerge from a runoff on Tuesday with ten weeks to go, and the only polls anyone can read were paid for by her side. If an independent survey lands in September showing this within a few points of either favorite, this page will say so and re-score. Until then, the number that keeps every seat pinned near the market is the same one from 2020: record money, dead-heat polling, and a ten-point loss.
To be explicit one final time: these are model estimates, not predictions of fact and not financial advice, and this page endorses no candidate and takes no side in the race. Kalshi is a CFTC-regulated event-contract exchange; you must be 18+ and in an eligible state, and availability varies by state as of August 2026.
Verdicts generated August 24, 2026, price-blind, revised once after an anonymized cross-read. Prices as of August 24, 2026, 17:48 UTC. Panel: Claude Fable, Claude Opus, Claude Sonnet, ChatGPT, Gemini Pro, Kimi, GLM, DeepSeek.
Prices on this page are Kalshi's book. If you also trade on Polymarket, code OS4 gets new users a $50 trading bonus on a $10 deposit — affiliate link; terms as stated by Polymarket; 18+, availability varies by state.
