South Carolina Senate Runoff Odds: Graham Vs Norman, Kalshi

Updated September 3, 2026 · 19 min read by Jake Hari

Darline Graham (the ballot name; national outlets often write Darline Graham Nordone) beat Ralph Norman for the Republican nomination in South Carolina's special U.S. Senate election on Tuesday, August 25, 52.5 percent to 47.5 on the reported count. The runoff market this page tracked all month is settled and paid, and the live race is now Tuesday, November 3: Graham against Democrat Annie Andrews for the seat itself, a Kalshi market that trades on party and prices the Republican side at 86 to 87 cents as of September 3, 2026.

The result also answered the question this page spent August asking: whether two independent reads landing on the same number means the number is right. The final pre-vote read published here two days before the runoff had the Kalshi market at 67 cents and our price-blind eight-model panel at 66.5 percent, and both were on the right side. The panel's grade runs closer than right side: its modal arithmetic gave Graham roughly 53 percent of the two-way vote, a four-to-seven-point win, and she won by five with 52.5 percent.

The graded record — what the crowd, the panel, the margin boards and the one poll of the race each said, and how each landed — is directly below, and the pre-runoff analysis itself, the board and every panel table, is preserved beneath it behind a frozen-record line so every grade on this page can be checked against what we actually wrote before anyone voted.

The Quick Answer

Darline Graham won the August 25 South Carolina Republican Senate runoff, 52.5 percent to Ralph Norman's 47.5 (193,396 votes to 175,327 on the reported count), and Kalshi's runoff market is finalized: her contract settled YES and paid a dollar, Norman's settled NO. In our final pre-vote read, published two days out, the market had her at 67 cents and our eight-model panel, scored price-blind, had her at 66.5 percent (the price history itself printed 55 cents in the last hour before polls closed; that hour is graded below); both called the winner, and the panel's central margin case, a four-to-seven-point win, hit at five. The live market is now the November 3 general election against Democrat Annie Andrews, which trades on the party sworn in rather than a named person: Republican 86 to 87 cents, Democratic 13 to 14, as of September 3, 2026, with the panel's post-runoff read at 90 percent. The full graded scorecard, the hour the runoff price cracked on election night, the November board, and the frozen pre-runoff record are all below.

New to event markets? Start with our 2026 Senate odds guide on Kalshi before you read the board.

The Market

Field	Detail
Venue	Kalshi
The Event	The November 3 South Carolina U.S. Senate general election — one yes/no contract per party
The Contract	Each market resolves YES if that party's candidate is sworn in as Senator for the term beginning in 2027. It settles on party, not on a named person, and closes early once the winner is sworn in
The Nominees	Republican: Senator Darline Graham (won the August 25 runoff). Democratic: Dr. Annie Andrews (won the June 9 primary outright)
Prices Below As Of	September 3, 2026

The August 25 runoff market that this page tracked through August is closed: Graham's contract is finalized YES, Norman's NO, along with every margin contract on the runoff's banded board. Every price in the live sections of this page is the November general-election market, stamped September 3, 2026; everything below the frozen-record line further down carries its original August stamps.

Hottest Prediction Markets Right Now

• 2028 Democratic presidential nominee · $190M traded
• 2027 Pro Football Champion · $69M 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.

The Result: Graham Won, By Five

The count as reported stands at 193,396 votes to 175,327, 52.5 percent to 47.5, a margin of 18,069 votes out of 368,723 cast. The Associated Press called the race for Graham on runoff night and CBS News projected the same result. Norman conceded plainly: "I don't like it, but I accept it," per Live 5 News. These remain reported rather than certified totals as of September 3; certification can move them by a handful of provisional ballots, not by 18,069.

One number in that paragraph deserves a second look before the grading starts, because it broke an assumption this page had carried for two weeks. The first round on August 11 drew 335,555 voters. The runoff drew 368,723, an increase of 9.9 percent, in a format that is supposed to shrink. Our runoff turnout page tracked Kalshi's turnout ladder through election week: at 8 p.m. on runoff night the 350,000-vote rung was bid at 97 cents, the 375,000 rung was quoted at 14 to 19 cents and falling as the first returns arrived, and 400,000 sat at 2 to 3 cents, a central read of 350,000 to 375,000 that the 368,723 count landed inside. That quote is an hour after polls closed, so read it as the market absorbing the count, not calling it. The conventional wisdom that a runoff shrinks the electorate, including the version printed lower on this page, was wrong.

How The Final Read Graded

This page existed to track one gap: the distance between the Kalshi price on Graham and what a price-blind panel of eight models thought the number should be. The gap closed to half a point two days before the vote, and then the vote graded both. Here is the scorecard, read against the pre-vote record preserved below the frozen line.

Final Pre-Vote Read	The Number	What Happened	Grade
Kalshi Head-To-Head	Graham 67¢	Graham won	Right side
Eight-Model Panel, Seat Median	Graham 66.5%	Graham won	Right side, half a point under the price
The Panel's Margin Arithmetic	~53% two-way share, a 4-to-7-point win	52.5%, won by 5.0	Hit, within half a point
Banded Margin Board, Priciest Contract	Graham by 0-3 (17¢ to 19¢)	Margin was 5.0 points	Missed by one band
Banded Margin Board, Second-Priciest	Graham by 3-6 (15¢ to 19¢)	Settled YES	Hit
The Only Public Poll (Trafalgar, Aug 22-24)	Graham 47.9, Norman 47.8, 4.4% undecided	Graham by 5	Missed by five, understating the favorite

The row worth narrating is the third one. When the panel re-scored on August 23, its seats' modal arithmetic gave Graham 55 to 60 percent of Russell Fry's eliminated bloc, conceded Norman 65 to 70 percent of Mark Sanford's, and landed her two-way share "around 53 percent, a four-to-seven-point win." The exact sentence is still in the frozen record below. She took 52.5 percent and won by five. The Trafalgar Group's final tracking survey of 1,031 respondents, in the field August 22 to 24 and the only genuine head-to-head poll anyone published, had Graham at 47.9 percent, Norman at 47.8 and 4.4 percent undecided, a dead heat in which the undecided slice was the whole margin; the transfer math built from first-round shares beat the poll.

The one-line grade: the crowd and the models converged at 67 cents and 66.5 percent, the count came back 52.5, and the panel's four-to-seven-point margin window caught the five-point result. Right winner, right size, and one wrong assumption about turnout, on the record below.

The frozen repricing section below asks the reader to "remember that 74-to-90-to-67 arc" in Graham's price. The arc got one more number: 99. On election day her contract traded 64 to 67 cents through the morning, touched 73 in the afternoon, and then slid to a print of 55 cents in the final hour before polls closed, her cheapest price of the entire day, on heavy volume, with the tied Trafalgar poll the only fresh input on the price history. An hour later, with real returns on the board, the same contract was at 99 cents, with 3.3 million contracts crossing in the hour. The full hour-by-hour account is on our runoff result page, and the one-line lesson travels: the hour before the count is the least informative hour of the cycle, and it is when a market's most nervous holders set the price.

Two honest misses belong on the record too. The margin board's single priciest contract, a Graham win by zero to three, settled NO: the crowd's modal outcome was one band too narrow, though its second-priciest band caught the result. And this page's own "How To Read A Runoff Market" section, preserved below, leaned on the premise that runoffs draw fewer voters than first rounds. Turnout rose 9.9 percent instead, and Graham won anyway. The mechanism was partly wrong even where the direction was right. The panel's run-by-run trajectory tells the same story from the other side: its August 12 read of 55.3 percent was too cautious, the market's climb to a 90-cent close in the week after the first round was too confident, and the number they converged on two days out was the one the count validated. For how the market handled the first round itself, including the night Kalshi's board flipped ahead of the AP call, see the price history page.

The Live Board: November 3, Graham Vs Andrews

The general election is Tuesday, November 3, 2026, and it is now the live race on this page. Graham, appointed in July to the seat of her late brother Lindsey Graham and now the Republican nominee for the full term beginning in 2027, faces Dr. Annie Andrews, a Charleston pediatrician who won the June 9 Democratic primary outright with 61.5 percent of the vote. Andrews has run federally before: she was the 2022 Democratic nominee in the 1st District, losing to Nancy Mace by about 14 points, and her fundraising is real, $10.8 million raised with $4.1 million cash on hand as of June 30, per FEC filings. No adverse record for Andrews surfaced in the searches behind this page's verified data card. Her reaction to the runoff result was one line, per Live 5 News: "The field is set and I'm ready to bring this election home." The same report notes Libertarian Kasie Whitener is also on the November ballot, which matters for the contract below: it settles on the party sworn in, so a third-party winner would settle both major-party markets NO.

November 3 General Election (Settles On Party)	Kalshi Price	Sportsbook Equivalent
Republican (Nominee: Darline Graham)	86¢ bid, 87¢ ask	-614 to -669
Democratic (Nominee: Annie Andrews)	13¢ bid, 14¢ ask	+669 to +614

Prices pulled from Kalshi on September 3, 2026, quoted as bid and ask, so the ask is the cost to buy. About 1.05 million contracts have traded across the pair lifetime, roughly 10,600 of them in the latest 24 hours: a real book, though nothing like the 4.9 million that crossed the runoff market on election day alone. The shape of this board's fortnight is its own small verdict: on Kalshi's daily candles the Republican side closed between 89 and 91 cents every day of the week before the runoff, closed at 88 on runoff night, printed as low as 85 the day after Graham won before closing at 87, and has closed between 86 and 89 every day since. A runoff that removed the race's biggest open variable moved the November price by roughly two to three cents, which is the market saying the seat, not the nominee, was always the thing it was pricing. Kalshi's November margin ladder says the same thing from the other side: the "Republicans by 7 or more points" contract sits at 47 cents bid, 53 ask, and the "5 or more" rung at 59 bid, 65 ask, so the crowd's median expected margin is about seven points, in a state Lindsey Graham carried by 10.2 in 2020.

Our panel re-scored the November question price-blind on August 26, the morning after the runoff result entered the record, when the Republican contract was trading at 86 cents. All eight seats returned a number and none saw a price. In the final revision round: DeepSeek 96 percent for a Republican hold, Claude Fable 94, Kimi, Gemini and Claude Sonnet 90, GPT, GLM and Claude Opus 89. (One first-round answer, Sonnet's, came back as 5 percent, an inverted read of the question that the seat corrected to 90 on revision; the first-round median without it was 89.) That is a seat median of 90 percent against a market at 86 to 87 cents, a gap of three to four points. What the high seat prices that an 86-cent market does not, in its own revision-round words: "South Carolina's R+8 PVI, Trump's 2024 +18 margin, and the 28-year Democratic Senate drought make a Republican hold overwhelmingly likely despite Darline Graham's weakness and a Democratic-leaning national environment." The 89 floor, from Opus, concedes the same structure and names what shaves it: "a novice nominee's debate damage and a D+6 national environment shave the margin but rarely flip a seat this red." That is a small disagreement. The cautionary number from our graded ledger applies to the big splits: on the 293 graded markets where the blend has left the market by ten points or more, the market has been right 204 times and the panel 89, about seven in ten for the crowd. A three-to-four-point gap is not one of those splits. Within the same day (the evening of August 25, Eastern time), six of the eight seats also scored the Andrews contract from her side of the board, price-blind, and put her at a median of 7 percent (the low seat at 5, the high at 12), against a market at 13 to 14 cents: the same read from the other direction. The election-night tape lives on our runoff result page; the case pages for both nominees, each carrying its own price-blind panel read, are the case for Graham in November and the case for Andrews.

The structural facts under the 87-cent price are the ones every seat cited. South Carolina has not elected a Democrat to the U.S. Senate since 1998. Every public handicapper has this race on the Republican side of the line, spread between Lean R and Safe/Solid R, and the freshest rating, Sabato's Safe R dated August 19, was issued the day after the debate performance that damaged Graham in the primary electorate. And the state's most instructive precedent cuts against reading Andrews' fundraising as a signal by itself: Jaime Harrison set a national fundraising record against Lindsey Graham in 2020, with late polling showing a dead heat, and lost by ten.

What would actually move this board, in order of force: the first independent poll of Graham against Andrews, which does not exist yet; every published head-to-head in this race was Democratic-sponsored and none tested Graham as the nominee; Andrews' third-quarter fundraising report in October; any handicapper moving the race to Tossup; and a Graham debate moment like August 18 in front of a November electorate rather than a runoff one. This page re-scores the panel when one of those lands.

The case for each side of the November 3 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 The settled runoff record, preserved for grading: Why Ralph Norman Will Win The South Carolina Senate Primary · Kalshi Odds: Why Darline Graham Will Win The SC Senate Runoff

Every number in this piece gets graded in public once the market settles, so you can hold the panel to its record. See the full graded scoreboard for how the calls have landed so far, and the chamber-wide 2026 Senate control board for where this seat sits in the majority math, alongside every 2026 Senate board we track.

Settlement, Dates And Refresh

Item	Detail
Runoff	Held Tuesday, August 25, 2026 — Graham won 52.5% to 47.5%; Kalshi's runoff and margin markets are finalized
Certification	Totals still reported, not certified, as of September 3, 2026; provisional ballots can move them by a handful
Q3 Fundraising Reports	Due October 2026 (the first FEC look at Graham as a general-election nominee)
General Election	Tuesday, November 3, 2026
Contract Settles	On the party of the Senator sworn in for the term beginning January 2027; the market closes early at swearing-in
Re-Scored	The panel re-scored the November question on August 26, 2026; the next re-score comes when a general-election poll, a rating change, or the October fundraising reports land. Prices as of September 3, 2026

The Bottom Line

This page opened as a gap tracker and closes its runoff chapter as a graded one. Across four panel runs the models walked from 42 to 66.5 percent while the market walked from 48 cents to 90 and back to 67, and the count settled the argument in a way neither side can spin: right winner, right margin window for the panel, one band narrow for the crowd's modal outcome, and this page's own wrong assumption about turnout that the result exposed in public. That is what holding estimates to a record looks like, and it is the whole reason the pre-runoff analysis below is frozen rather than rewritten. The same apparatus now points at November 3: an 86-to-87-cent market, a 90 percent panel read, a nominee who has never been tested in an independent general-election poll, and a state that has not sent a Democrat to this chamber in 28 years. If you would rather sweat games than election returns between now and November, our free expert picks today are open to everyone, no account needed. Here, the next number that moves the board is not ours to write: nobody has fielded an independent poll of Graham against Andrews, and the first one to land inside single digits would drag 86 cents into the low 80s before the October fundraising reports even post. That poll, or that report, is when the panel re-scores and this page gets its next graded line.

---

The Pre-Runoff Record, Frozen As Published

Editor's note: the analysis from here to the link footer is preserved from the page as it stood on August 23, 2026, two days before the vote — the board, the run-by-run panel table, both candidates' sourced records, the margin boards and the panel's own stated uncertainty, verbatim, so the grades above can be checked against what we actually wrote. (The pre-runoff FAQ, dates table and closing summary were replaced by the updated versions above, and the graded-scoreboard line and sibling case-page links now live in the November section; every analytical section is kept word for word.) Prices, tenses and dates below are as of August 23, 2026, and are deliberately not updated.

The Result: The Primary Went To A Runoff

No candidate cleared the 50 percent needed to win outright, so the Associated Press projected a runoff on the night of August 11. On the final count, Darline Graham led with 32.7 percent, Ralph Norman took 24.7 percent, and Russell Fry finished third at 19.6 percent, with Mark Sanford fourth just under 15 percent and the rest of the field in single digits. Norman finished five points clear of Fry on the final count for the second runoff slot.

That result reshaped the board. As late as the final Sunday, Kalshi's nominee board had Fry at 33 cents and Norman at 24; the order flipped on election-day morning, and the count confirmed the flip: Fry is out. Norman is in. What is left is a straight two-way contest between the appointed incumbent who led the first round and the sitting congressman who consolidated enough of the conservative vote to finish second. Everything below prices that head-to-head.

The Board

The two runoff candidates, with the certified first-round share, the current market price, and the panel's blended estimate after its revision round.

Candidate	First Round	Market	AI Blend
Darline Graham	32.7% (led)	67¢	66.5%
Ralph Norman	24.7%	34¢	33.5%

Market prices are live as of August 23, 2026; the AI estimates were scored price-blind on August 23, 2026, after the endorsements, the debate and the close of early voting, and the runoff is August 25, so the race can still change. The blend is the median of the eight seats, not an average. Put the two columns side by side and the gap that was the whole story of this page a week ago is gone: half a point, inside the bid-ask spread. That is not the panel copying the price, because it never sees one. It is two independent reads of the same runoff arriving at the same place by different roads, and the section below walks the road.

More live boards from the same panel: Georgia Senate 2026 has Jon Ossoff a heavy 92¢ favorite over Mike Collins at 8¢; North Carolina Senate 2026 has Roy Cooper at 91¢ over Michael Whatley at 8¢; and whichever Republican emerges here feeds straight into the chamber-wide 2026 Senate control board, where Republicans hold at 53¢. Prices fetched August 22, 2026.

How The Panel's Number Has Moved

One row per time the panel actually re-scored this race, with the market price on the same day and the reason the run happened. The panel figure is the blend as published that day (the seat mean through August 12, the seat median since). The gap column is what this page exists to track.

Panel Run	Question Scored	Panel (Graham)	Market (Graham)	Gap	Why It Re-Ran
August 7	Win the full eleven-name primary	42%	48¢	6 pts under	Pre-primary field verdict, four days before the vote
August 12	Win the runoff	55.3%	74¢	19 pts under	Runoff set the night before; first two-way read, six seats
August 19	Win the runoff	62.5%	78¢	16 pts under	Post-debate re-score, first run with all eight seats
August 23	Win the runoff	66.5%	67¢	0.5 pts under	Endorsements, rally and the full early-vote count in hand, two days out

Read the table top to bottom and two lines converge. The market's line runs 48, 74, 78, 67: it repriced hard on the first-round result, kept climbing through Fry's endorsement to a 90-cent close that no panel run happened to catch, then gave most of that back after the debate. The panel's line runs 42, 55, 62.5, 66.5: it never saw a price, and it climbed every time it was handed a new sourced fact, the certified eight-point lead, then Fry crossing to Graham, then Trump in person and a falling early-vote curve. The August 19 row is the one to notice. The panel had moved to 62.5 percent that morning and the market was still at 78, so the 16-point gap was real, and it closed from the market's side over the next three sessions, not the panel's. The August 23 row shows what happens when both have the same information: they agree, and the rest of this page is about why, and about the one scenario in which both could be wrong together.

Darline Graham: The Front-Runner The Panel Now Agrees On

Graham enters the final 48 hours ahead on every visible count. She is the sitting appointed senator and the late Lindsey Graham's sister, appointed to the very seat he held, which is why the Graham name on this ballot carries so much weight. On July 17 President Trump gave her his "Complete and Total Endorsement" and reiterated it after the first round, and she led that first round by exactly eight points. The week after added the most valuable name available: Russell Fry, the man she and Norman knocked out, endorsed her on August 14, joining Senators Tim Scott and Tom Cotton, Senate Majority Leader John Thune and Lt. Governor Pamela Evette. That package is why the market pushed her as high as a 90-cent close. What it could not survive intact was the debate: her contract fell 24 cents from that close to 66 in the sessions after the August 18 head-to-head in Newberry, and has sat at 66 or 67 since, through Trump's Friday rally with her in Myrtle Beach and the close of early voting.

A week ago the panel would not round her up past the mid-50s. Now all eight seats make her the favorite, at a median of 66.5 percent, and every one of them gave the same first reason.

"Norman needs roughly 59 percent of the eliminated 43 percent while the largest bloc, Fry's 19.6 percent, and Trump's in-person rally both point to Graham." — Claude Fable, scored August 23

The seats do not think she wins big. The modal arithmetic across the panel gives her 55 to 60 percent of Fry's voters, concedes Norman 65 to 70 percent of Sanford's, and lands her two-way share around 53 percent, a four-to-seven-point win if turnout holds evenly across blocs. What moved the number from 55 to 66 was not a change in that math; it was that three of the four things the panel said would change its mind, a Fry endorsement, a Trump appearance and a turnout signal, all landed, and two of the three landed on her side. The market had already priced that. Now so has the panel.

Ralph Norman: The Underdog The Panel Rated Above The Market, Until The Market Caught Up

Norman is the runoff's underdog on the board at 34 cents, and for a week he was the panel's favorite argument. On August 12 the six seats blended him at 44.7 percent while the market had him near 26 cents; the market then pushed him down to an 11-cent close at Graham's August 17 peak, and tripled him to the mid-30s in two sessions after the debate, and when the full eight-seat panel re-scored on August 23 it put him at 33.5 percent, half a point under his price. The panel did not abandon his case. It re-weighed it with the endorsement ledger filled in. He is a sitting U.S. Representative for South Carolina's 5th District and a member of the House Freedom Caucus, and he finished a clear second while Fry, his closest competitor for the hard-conservative lane, was eliminated. His endorsement column is real: former Governor Nikki Haley, Congresswoman Nancy Mace, former Senator Jim DeMint, Senators Rick Scott and Mike Lee, and eliminated candidate Mark Lynch have all lined up behind him, and on Friday, August 21, Mark Sanford joined them, a man who said he has "refrained from seeking or giving political endorsements" and broke that habit four days before the runoff, after Trump had personally called him asking for the opposite. (Haley and Mace had speculative contracts on Kalshi's original board but were never on the primary ballot; Lynch and Sanford are the two Norman backers who actually were.)

The seat that gives Norman the most credit explains exactly where his ceiling is.

"Fry's 19.6 percent bloc is the largest eliminated slice and already chose Trump-backed Graham once, capping Norman's realistic transfer share below the roughly 59 percent he needs." — Claude Sonnet, the panel's low seat on Graham at 61 percent

That is the shape of the whole board: the seats that like Norman most, Sonnet at 39 percent and the GPT seat at 37, get there by giving him two-thirds of Sanford's bloc and only a split of Fry's, and even that leaves him a few points short. The case for an upset is no longer that the market has the math wrong; it is that turnout breaks unevenly in a way no first-round share can show, which is the one scenario on this page that neither the crowd nor the panel can price.

The Record: What Is Documented On Both Candidates

Before pricing two named politicians, it is worth stating plainly what the public record actually documents on each, and what it does not.

Norman's file is the longer one, and it is old news the market has had years to digest. In April 2018 he took out his loaded handgun and set it on the table during a constituent meeting at a Rock Hill diner, telling attendees he was "not going to be a Gabby Giffords"; South Carolina's attorney general declined to prosecute, concluding the incident did not warrant charges. Separately, Rock Hill resident Travis Price sued the city and Norman in 2021 over statements about Price's arrest, including a Norman Facebook post; Norman was later dismissed from the suit as a defendant, per court records reported by The Post and Courier, and the city, not Norman, later settled Price's claims for $500,000. Whatever weight voters give either episode, there is no sign in the price history that old news is what is moving this market this week.

Graham's file contains no lawsuit or ethics action we could find. What is documented is a campaign attack file: before he was eliminated, Russell Fry's campaign attacked her over her donations to Katrina Shealy's 2024 reelection bid, her presidential-primary voting history, and her support for data-center development: opposition framing, sourced as such, not findings of misconduct. The sharpest documented moment of her runoff is her own: the debate answer covered in the next section, which is a résumé concession, not a scandal. That distinction matters for how you read the board: nothing in either record is new information to this market, which is why the repricing below traces to the debate stage and not to anyone's past.

The Week That Repriced The Board

Here is what nine days did to this market, in order. On the morning after the first round, Graham traded at 74 cents. Between August 12 and 14 the endorsements piled up on her side: Tim Scott, Pamela Evette, John Thune, Tom Cotton, Trump again. Then came the one the market cared about most. Fry, whose 19.6 percent of the first round was the biggest single bloc of eliminated voters, crossed to the front-runner on August 14. The crowd did the obvious math and marked her up day after day, to a 90-cent close.

Then came Tuesday, August 18, and the one event neither campaign could script: the only debate of the runoff, at the Newberry Opera House, moderated by Newsmax's Greta Van Susteren. Norman spent the night pressing Graham on her résumé and her lack of elected experience, and she handed him the soundbite of the race: asked about Taiwan and American security interests, she answered that, in her words, national security is "not my thing". Early voting opened the next morning, and the price history has moved one direction since. Graham dropped to 78 cents on August 19, then 67 by August 20, 66 on August 22, and 67 on August 23; Norman went from an 11-cent close to 22, then 33, then 34 in two sessions, with more than 1.1 million contracts changing hands across the pair on August 20 alone and about 586,000 in the latest 24 hours, roughly 308,000 on Norman's leg and 278,000 on hers. A two-sided book doing a million contracts a day while the price walks steadily one way is not a stray order; it is the market repricing the same runoff from "all but decided" to "a favorite with a real underdog," and the fact that the new level has held through a Trump rally, the close of early voting and a full weekend says the reprice stuck. Remember that 74-to-90-to-67 arc; it is the market grading its own two weeks, and the tracking table above is the panel grading its own: the market was early on the endorsements and late on the debate, the panel was the reverse.

The Runoff Math The Panel Keeps Returning To

The number the seats kept circling back to is a threshold. For Norman to overtake a first-round deficit of eight points on the certified count, he needs about 59% of the roughly 43% of voters whose first choice is now off the ballot. On the August 23 re-score every one of the eight seats opened its analysis with that same threshold, and the panel's disagreement reduced to a single assumption: how much of Fry's bloc actually follows Fry.

"Those voters chose Fry knowing Trump had already endorsed Graham, so they are demonstrably endorsement-resistant. Endorsements move maybe a third of a bloc, not all of it." — Claude Opus, first-round analysis, before revising Graham up from 61 to 68 on the size of the bar

That is the whole runoff in one line, and this week complicated it in both directions. The case for the upset was always that Fry's 19.6 percent came from the hard-conservative, Freedom Caucus lane that overlaps naturally with Norman. Fry himself then endorsed Graham, which cuts directly against the mechanism the panel leaned on when it scored, worth stating plainly. The honest counterweights: endorsements move candidates more reliably than they move coalitions, Mark Sanford's more independent-leaning 15 percent was always the harder bloc to place and Sanford himself endorsed Norman on Friday, and South Carolina runoffs typically draw fewer voters than the first round, which rewards whichever campaign holds the more committed base. Norman needing roughly six in ten of a pool whose largest single bloc, Fry's 19.6 percent, just watched its candidate cross to his opponent is a steep bar. Steep, but the market that priced it 90-cents-settled a week ago now prices it at 67, the second-largest bloc's candidate came his way, and a debate in which the front-runner conceded a whole policy domain is exactly the kind of event that makes eliminated-candidate voters persuadable again. The panel's answer, after weighing all of that, is that it gets Norman to about 52 to 55 percent of the transferred vote, meaningful catch-up and still four to seven points short.

What The Margin Boards Say About How Big A Win The Market Expects

The head-to-head is no longer the only way Kalshi prices this runoff; it now lists two margin-of-victory boards, and they answer a different question: not who wins, but by how much. The banded board came alive with the race: on August 20 it traded about 204,000 contracts in a single day, more than a quarter of its lifetime total, and it is still turning over tens of thousands a day on about 770,000 lifetime. The directional ladder is far thinner, and as of Sunday it is quoting numbers that contradict each other, so read it for what it says about its own book rather than the race.

The banded board (ticker KXPRIMARYMOV-SCRSENS262, opened August 12) sells a separate contract for each three-point margin band, one set per candidate. Its shape has changed twice in a week. On this page's August 19 refresh her 15-to-18-point band was the priciest contract on the board at 10 to 12 cents. By August 22 the top was shared between a Norman squeaker and Graham's six-to-nine band at 13 to 14 each. As of August 23 the board has moved again, and toward her: Graham by zero to three points is now the priciest contract at 17 bid, 19 offered, her three-to-six band is right behind at 15 to 19, and her six-to-nine band holds at 13 to 14. Norman's two narrow-win bands, zero to three and three to six, are each quoted 11 to 13. Every Graham band above nine points is offered at 8 cents or less, and her landslide of 24 points or more is down to 2 to 3 cents.

Margin Band	Graham Wins By	Norman Wins By
0 To 3 Points	17¢ to 19¢	11¢ to 13¢
3 To 6 Points	15¢ to 19¢	11¢ to 13¢
6 To 9 Points	13¢ to 14¢	5¢ to 6¢
9 To 12 Points	7¢ to 8¢	5¢ to 7¢
12 Points Or More	15¢ to 20¢ combined	4¢ to 10¢ combined

Add her side of the board up and the bids on her nine bands sum to 67 cents, the head-to-head price to the cent; Norman's sum to 36. The two boards agree on who wins. The row to read is the top one. A week ago the single most likely outcome on this board was a mid-teens Graham win; now it is a Graham win by under three points, with her three-to-six band nearly level. The crowd has not moved off her as the winner, but it has shrunk the expected size of the win to a single-digit margin and put two Norman squeaker bands within a few cents of her best one. That is the panel's roughly 53-to-47 two-way share, a four-to-seven-point race, showing up in the market's own derivatives.

The directional ladder (KXPRIMARYMOV-SCRSENS263, opened August 17) asks the cleaner cumulative question, will Graham win by at least X, and on Sunday it is not answering it coherently. Her win by three points or more is quoted 31 bid to 52 offered, a 21-cent spread, while her win by six or more is quoted 44 to 55, above the rung it is supposed to sit below, which is impossible for a cumulative ladder. The twelve-or-more rung at 12 to 17 and the landslide rungs, all bid 2 with asks tightening from 6 down to 4, are at least in order. On about 92,000 lifetime contracts against the banded board's 770,000, this is a book with no one minding it, and the rule from every thin market applies: when a deep board and a thin board disagree, trust the deep one and treat the loose quote as noise.

Every Seat's Number

The full per-model board for the runoff, showing each seat's estimate after the revision round. Percentages are rounded to the nearest point; the blend is the seat median.

Candidate	Fable	Opus	Sonnet	GPT	Gemini	GLM	Kimi	DeepSeek	Blend
Darline Graham	66	68	61	63	67	66	67	67	66.5%
Ralph Norman	34	32	39	37	33	34	33	33	33.5%

Models were scored price-blind on August 23, 2026, from a data card carrying the certified first-round result, the full endorsement ledger through Sanford, the August 18 debate, the 63,712 early-vote total with its daily curve, and both candidates' documented records; each then read the others' anonymized reasoning and could revise. No seat saw a market price at any point. All eight seats of the panel ran this board and all eight completed the revision round.

The row worth a second look is the spread, not the middle. Before the revision round the eight seats ran from 57.5 to 78 on Graham, a 20-point range; after it they sit between 61 and 68. Four seats came down and four went up, and the two that moved least, Kimi and GLM, were already at 66 and 64. The panel did not converge because it was told to; the prompt explicitly licenses holding. It converged because the seats at both ends were using different shares for the same Fry bloc, and the round made them say so.

Where The Panel Changed Its Mind

After the first round of estimates, each seat read the others' anonymized reasoning and could revise. This time the panel moved, in both directions, and the arcs are the most useful thing on the page.

The biggest drop was the seat that opened highest. Gemini had Graham at 78 percent, the most confident number on the board, on the view that Trump's in-person rally would move Fry's Trump-aligned voters far more efficiently than Fry's own endorsement. It read the other seats' transfer math and came down 11 points.

"Graham's eight-point first-round lead and Trump-reinforced coalition transfer from Russell Fry make Norman's required 59 percent capture of the eliminated vote mathematically prohibitive despite her debate gaffe." — Gemini, revising Graham from 78 to 67

The biggest climb was Opus, up seven points from 61 to 68, and the second biggest was the seat that opened lowest. The GPT seat had Graham at 57.5 percent, the lowest first-round number, weighting the debate answer as "the main reason this is not a safer Graham call." After the round it moved up five and a half points, not because it discounted the debate but because it nudged its Fry-bloc share toward Graham, from 55 to 60 percent to 57 to 62, while leaving its read on Sanford's voters untouched.

"Sanford's roughly 15 percent is Norman's clearest pickup: his voters are more independent and traditional-Republican, and Sanford's late endorsement against Trump's ask is a real signal. I would put that bloc around 65 to 70 percent Norman." — GPT seat, revising Graham from 57.5 to 63

And one seat went the other way from the pack on purpose. Sonnet opened at 64 and finished at 61, the panel's low number, on a point the others had stated but not priced.

"Those voters picked Fry over Graham despite Trump already having endorsed her, so they are somewhat endorsement-resistant, not a blank check for whoever Fry names. I would put only 55 to 60 percent of that bloc with Graham, not more, since 'follow Fry' and 'follow Trump directly' are not identical asks." — Claude Sonnet, revising Graham from 64 to 61

Opus made the same observation in its first round and drew the opposite conclusion, moving from 61 to 68 because, in its words, Sanford "defied a personal call from Trump to say so, which is a costly and therefore credible signal," and a credible signal on a 15-percent bloc still leaves Norman short of 59 percent of the pool. Same fact, two seats, a seven-point disagreement about what it is worth. That is the honest width of this race.

What Would Change The Panel's Mind

The panel re-scored on August 23 with most of its August 12 triggers already fired. Here is what it named this time, and what is left to fire before Tuesday night:

• A Runoff Poll. Still the missing piece, and still the event every seat named first. No public survey of the two-way race has surfaced; the panel is reasoning from first-round shares and an endorsement ledger, not a head-to-head number. Five of the eight seats named their own Fry-bloc split as the single fact they are least sure of, and a poll is the only thing that would measure it. One showing Norman near even would pull the blend back toward the coin flip it scored two weeks ago; one confirming a clear Graham lead would push it above the price.
• Fry's Voters Versus Fry's Endorsement. The modal assumption across the panel is that 55 to 60 percent of Fry's 19.6 percent follow him to Graham. Opus and Sonnet both flagged that those voters already declined Graham once with Trump's endorsement on the table. If Tuesday's count in Fry's Horry County base shows Norman running well above his 24.7 percent first-round share, the panel's central assumption was wrong and so was the price.
• Turnout Shape. The first hard number is in: 63,712 early ballots statewide across three days, with the daily count tapering from about 23,400 on Wednesday to about 19,900 on Friday rather than building. The seats split on what that means. Most read a falling curve as a smaller, more habitual electorate, which several argued favors the side with the Trump-organized machine; Opus gave Norman a modestly higher read from it, on the view that runoff voters skew more ideological; the GPT seat weighed the same fact and landed the other way after the revision round, calling low turnout a slight Graham positive because "she needs defense, not persuasion." The number nobody has is the Election Day share, and a runoff that finishes well above the first round's turnout would tell both the crowd and the panel that something neither modeled is happening.
• The Rally Lag. Trump rallied with Graham in Myrtle Beach on Friday, August 21, the final day of early voting, in a race the Washington Examiner frames as a proxy battle with Nikki Haley, who backs Norman. The price went from 67 before the rally to 66 after it and back to 67 by Sunday, so the event the panel once expected to move the board most has not moved it. Gemini's first-round view was that the rally is an "overwhelming turnout mechanism" that shows up in Election Day voting rather than in early ballots already cast; if Tuesday's in-person count runs heavy in the Upstate and the Grand Strand, that is the rally arriving late.

And one honest footnote on the list itself: the event that actually moved the market most, a debate, was not on the August 12 list either. The panel named endorsements, polls, Trump and turnout; the repricing came from a résumé answer on a Newberry stage. Markets and models alike get the direction argued right and the catalyst wrong.

How The Panel Called The Field Before The Vote

Editor's note: the section below is our pre-primary field verdict, generated before any votes were counted. It is preserved as it was published so the panel can be held to its record. Read against the result, the panel's clearest call was on Russell Fry: it ranked him a clear third at a blended 16.1 percent while the market priced him second, and Fry was indeed eliminated. Both of the panel's co-front-runners, Norman and Graham, advanced to the runoff.

Ahead of the August 11 vote, the six-model panel scored the full certified field price-blind and split from the market in two places. It put Ralph Norman a nose ahead of Darline Graham at the top, on the strength of the only public poll and likely-runoff math, where the market made Graham the clear favorite. And it kept Russell Fry a clear third where the market priced him second.

"Norman leads the only public poll (22% to Graham Nordone's 19%, within margin) and runs strongest among the under-60 majority likely to decide a runoff." — Claude Opus, pre-primary field verdict

"Fry has momentum and a wave of in-state endorsements but splits the hard-conservative vote with Norman and must first reach the top two." — Kimi, pre-primary field verdict

The result validated the caution on Fry and the read that this was a two-way race at the top headed for a runoff. It also showed where the panel underrated Graham: she led the first round outright at 32.7 percent, ahead of the poll-implied dead heat, which is why every runoff re-score since has started from her eight-point lead and worked backward to what Norman would need.

How To Read A Runoff Market

A two-candidate runoff board rewards understanding a few things about how it works, whoever you are:

• A First-Round Lead Is Not A Runoff Win, And This One Has A Number. Graham led with 32.7 percent, not a majority, so the race reduces to a single threshold: whether Norman can pull roughly 59 percent of the 43 percent who backed eliminated candidates. Fry endorsing her argues no; his Freedom Caucus base overlapping with Norman, Sanford's harder-to-place 15 percent, and a lower-turnout runoff argue it stays live. At 67 cents the market prices her lead as solid and, as of Sunday, so does the panel; what neither can price is whether Fry's voters actually follow Fry.
• Tuesday's County Returns Are The Tell. Norman's base is the 5th District, York, Lancaster and Chester counties around Rock Hill; the Graham name is strongest in the Upstate, around Seneca and Greenville, where Lindsey Graham built his career. Horry County, Fry's home base around Myrtle Beach, is the cleanest test of the panel's central assumption: if Norman is running well ahead of his 24.7 percent first-round share there, Fry's voters did not follow Fry. If he is running ahead of it outside his own district generally when the early totals drop, the eliminated-voter math above is breaking his way, and the 67-cent price and the 66.5 percent panel read were both too high. The contract itself pays on the nomination, per the exchange's written settlement rules, always the governing document.

Kalshi is a CFTC-regulated exchange open to users 18 and older, with availability that varies by state. Nothing here is advice about whether or how to trade, or about how to vote. This article is informational and does not constitute financial, investment, betting, or voting advice.

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.

More on this: 2026 Senate Races: Every Kalshi Board We Track · South Carolina Senate Runoff Odds: Kalshi Beat The AP Call · Why Ralph Norman Will Win The South Carolina Senate Primary · South Carolina Senate Runoff Turnout: The 375,000-Voter Question · Michigan Senate Election 2026 Candidates: El-Sayed Favored