Updated August 23, 2026 · 12 min read · by Jake Hari
On June 16, Gentner Drummond finished first in a nine-way Republican primary for governor of Oklahoma. He beat Mike Mazzei by 1,141 votes out of 403,099 cast. Eighteen days earlier, Donald Trump had endorsed Mazzei.
Drummond won anyway. Then the market decided it did not matter.
The Quick Answer
Prediction market traders on Kalshi give Mike Mazzei roughly a 78% chance of winning Tuesday's Oklahoma Republican gubernatorial runoff, and Gentner Drummond roughly 22%. We asked eight AI models to price the same race without letting any of them see the market, and they landed almost 24 points away from it: their middle estimate gives Drummond a 46% chance. That gap ranks among the widest we have ever recorded between our panel and the market on an election contract. The rest of this piece is about which side has the better argument.
Free: The Weekly PM Market Brief — the 8-model panel's graded record, the week's biggest market-vs-model gaps, and what's spiking next. One email, Sundays.
New to prediction markets? Start with our guide to election markets on Kalshi, which walks through how a contract that pays $1 actually settles.
Where The Price Sits Right Now
Prices below were read from Kalshi's public API on August 23, 2026. On a contract like this, a price in cents is just the probability with a dollar sign in front of it: 78 cents means the market thinks there is roughly a 78% chance that outcome happens, because a winning contract pays out $1.
This page re-reads the board as the story moves, and the stamp above always tells you when the numbers were last pulled.
The Board
Kalshi listed seven contracts on this event, one for each candidate it chose to price. Nine names ultimately appeared on the June ballot, and two of the seven Kalshi listed never got there. Five of the seven were settled to No and now trade at zero. Two are live.
| Candidate | What he is | Market price | AI panel |
|---|---|---|---|
| Mike Mazzei | Former state senator, former budget secretary, Trump's pick | 79¢ | 54% |
| Gentner Drummond | Sitting attorney general, won round one | 23¢ | 46% |
| Chip Keating | Third in round one, settled to No | 0¢ | settled |
| Charles McCall | Fifth in round one, settled to No | 0¢ | settled |
| Jake Merrick | Fourth in round one, settled to No | 0¢ | settled |
| Matt Pinnell | Listed by Kalshi, not on the June ballot | 0¢ | settled |
| Ryan Walters | Listed by Kalshi, not on the June ballot | 0¢ | settled |
The two live contracts are quoted at 22 to 23 cents on Drummond and 78 to 79 cents on Mazzei; the table shows the last traded price.
The penny of overlap between the two sides is the bid-ask spread, the small gap between what buyers will pay and what sellers will take. If sportsbook odds are your native language, 22 cents on Drummond is about +355 and 78 cents on Mazzei is about -355. It is a deep board by election-market standards: about 524,000 contracts have changed hands on each side.
Every Seat's Number
Each model saw the same fetched briefing and no price at all. Numbers are each seat's probability that Drummond wins the nomination, after a round in which every model read the others' reasoning anonymously and could revise.
| Seat | First estimate | After revision | Move |
|---|---|---|---|
| Claude Sonnet | 45% | 54% | +9 |
| Claude Opus | 45% | 53% | +8 |
| DeepSeek | 55% | 48% | −7 |
| Claude Fable | 43% | 46% | +3 |
| Kimi | 54% | 46% | −8 |
| GPT | 43% | 45% | +2 |
| GLM | 49% | 45% | −4 |
| Gemini | 42% | 44% | +2 |
| Panel (seat median) | 45% | 46% | +1 |
| Market | — | 22 to 23¢ | — |
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 · 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.
Read that table again, because the shape of it is the finding. After a second round on an enriched briefing, all eight seats landed between 44% and 54%. That is not the panel shrugging. Every seat committed to a number and defended it, and they converged there because the two strongest forces in this race genuinely cancel: Mazzei owns the one bloc with a clear organizing cue and the turnout math that usually favors it, while Drummond owns the larger bloc, the better favorability, the cash for the closing week and the only poll with no disclosed campaign or PAC sponsor, which has the race tied. The panel's actual verdict is that Tuesday is close to a coin flip.
That is what makes the price interesting. A market at 78-22 is not pricing a close race.
Every number in this piece gets graded in public once the market settles. You can see the whole ledger on the full graded scoreboard.
More live boards from the same panel: California Governor Odds (Xavier Becerra 96¢, the most settled governor board on the exchange) · Alaska Governor Odds (Bernadette Wilson 38¢ in a field nobody can clear) · 2026 Governor Races: Odds for Every Battleground, where Oklahoma's November contract sits at 93¢ for the Republican. Prices fetched August 23, 2026.
Hottest Prediction Markets Right Now
- 2028 Democratic presidential nominee · $190M traded
- 2027 Pro Football Champion · $69M traded
- 2028 U.S. Presidential Election winner? · $61M 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 Market Card
| Event | Oklahoma Republican Governor nominee, 2026 (Kalshi event KXGOVOKNOMR-26) |
| Settles | When the Republican Party of Oklahoma certifies a nominee. Runoff election day is Tuesday, August 25, 2026 |
| Settlement source | Republican Party of Oklahoma |
| How it resolves | A candidate's contract pays $1 if that candidate wins the Republican nomination to contest the 2026 Oklahoma governorship |
| Structure | Mutually exclusive. Seven candidates were listed; five were finalized at zero after the June primary |
| Availability | Kalshi is a CFTC-regulated designated contract market. These are event contracts, not sportsbook wagers. 18+ |
The Record: A Smelter, A Lawsuit, And A Reversal
This race broke on one project.
Emirates Global Aluminium and Century Aluminum want to build a $4 billion aluminum smelter at Inola, about 30 miles east of Tulsa. It carries a $500 million U.S. Energy Department grant and is backed by sovereign wealth funds tied to the United Arab Emirates government. Drummond, as attorney general, sued to block it, citing electricity, water and farmland concerns. His line, per Reuters: "I'm not going to let it be built. We have to protect farmland."
Mazzei had opposed the smelter too. He reversed on May 29, the same day Trump endorsed him, and said he was "proud to stand with President Trump and support a project he's made clear is a national security imperative." Drummond's answer, from the same reporting: "Mazzei was all too happy to sell out Oklahoma."
On July 15, Trump released a one-minute campaign video, transcribed in full by News 9. The relevant part:
Mike is running against a total RINO named Gentner Drummond, nothing but trouble, who is just another grandstander always looking to sell out Oklahoma to the radical left.
The smelter is not universally popular where it would be built. At an Inola city council meeting, residents heckled Energy Department officials with "Build it in Mar-a-Lago!"
Here is what that record did to the price. On May 29, the day of the endorsement, Drummond's contract fell from the mid-40s to single digits in one session, and Mazzei's rose from 19 cents to 84. It was the largest single-day move the board has ever seen. Drummond has not traded above 30 cents since.
We are reporting what the record did to the number. We are not taking a side in it.
The Arithmetic The Panel Kept Coming Back To
Round one left 192,576 votes, 47.8% of the primary electorate, with candidates who are no longer on the ballot. That is the whole game, and the panel spent most of its reasoning there.
Certified first-round results, June 16, 2026:
| Candidate | Votes | Share |
|---|---|---|
| Gentner Drummond | 105,832 | 26.3% |
| Mike Mazzei | 104,691 | 26.0% |
| Chip Keating | 74,379 | 18.5% |
| Jake A. Merrick | 58,341 | 14.5% |
| Charles McCall | 47,519 | 11.8% |
| Four others combined | 12,337 | 3.1% |
Only one of those eliminated candidates has endorsed. Jake Merrick, whose 14.5% was the most ideologically committed bloc in the field, backed Mazzei. So did Governor Kevin Stitt, who is term-limited and had backed Chip Keating in round one, and John O'Connor, the attorney general Drummond unseated in 2022. Chip Keating and Charles McCall, whose voters are 30.3% of the primary electorate between them, have said nothing.
Now the polling, which is where the panel split.
The complete public polling record for this runoff:
| Pollster | Field dates | Sample | Margin | Drummond | Mazzei | Undecided |
|---|---|---|---|---|---|---|
| CHS & Associates (R) | Completed Aug 6, 2026 | 600 LV | ±4.0% | 44% | 44% | 12% |
| Pulse Decision Science (R) | Aug 2 to 4, 2026 | 600 LV | ±4.0% | 38% | 44% | 18% |
| Pulse Decision Science (R) | Jun 21 to 23, 2026 | 606 LV | ±4.0% | 38% | 44% | 18% |
| Cygnal (R) | Jun 18 to 19, 2026 | 600 LV | ±4.0% | 44% | 42% | 14% |
The CHS dates come from the pollster's own release; the other three are as the public aggregator lists them, with no primary release retrievable.
Two things about that table matter more than the numbers in it.
First, there is no nonpartisan public poll of this race. Not one. Every survey ever released on this runoff comes from a Republican firm. The two Pulse Decision Science polls were run by Club for Growth Action, a super PAC that has endorsed Mazzei, and its June and August releases report an identical 38-44-18 line two months apart. More than $13 million has been spent across all of Oklahoma's August 25 runoff contests since July 1, most of it aimed at this governor's race. A number that does not move across two months of that is a message, not a measurement.
Second, the newest poll in that table finished interviewing on August 6 and was not published until August 11. Fieldwork stopped nineteen days before the runoff and fourteen days before early voting opened on August 20. Everything in the closing stretch of this campaign is invisible to it.
The one poll that published its internals is the CHS Sooner Survey, completed August 6 and published August 11, and its crosstabs are what the models argued over. It found the likely electorate splits 38% self-identified "Trump Republicans" and 56% "traditional Republicans." Trump Republicans back Mazzei 62-22. Traditional Republicans back Drummond 60-30.
Most of the panel read that as a Mazzei edge, because his margin inside his group is wider. Claude Sonnet did the multiplication and reached the opposite conclusion:
The decisive test is arithmetic several other forecasters get wrong. Trump Republicans are only 38% of the likely electorate versus 56% traditional. Weighting by actual size: Mazzei gets 0.38x62+0.56x30=40.4, Drummond gets 0.38x22+0.56x60=42.0. Drummond +1.6, which is what the reported 44-44 topline already reflects. — Claude Sonnet
[Editor's note: the CHS topline is a 44-44 tie. The slight Drummond edge Sonnet describes shows up in the poll's without-leaners number, which is Drummond 42 and Mazzei 40.]
That is the disagreement in one paragraph. The bigger margin belongs to the smaller group.
Where The Panel Changed Its Mind
Between the two rounds, five seats moved toward Drummond and three moved away. Claude Sonnet's nine-point move, quoted above, was the largest in the panel. The next two largest went in opposite directions, and both had a reason.
Claude Opus went from 45% to 53% on a piece of the calendar the rest of the panel had underweighted:
Trump endorsed Mazzei 18 days BEFORE round one and Mazzei still trailed; that edge is already priced in, and transfer math plus the independent poll both say tie. — Claude Opus
Opus also broke with the majority on turnout, which is the load-bearing assumption on the Mazzei side:
The claim that a 21% dropoff mechanically enriches the electorate with "Trump Republicans" assumes low-propensity populist voters outvote habitual ones in a sleepy August runoff. The usual pattern is the reverse. CHS also already screened for a runoff universe, so re-applying the shift double-counts. — Claude Opus
Kimi moved the other way, from 54% to 46%, after reading the transfer argument:
Merrick's 14.5% was the field's most ideological, Trump-style bloc, he endorsed Mazzei, and the CHS crosstabs show that voter type breaking 62-22 for Mazzei. Net, I score the transfer math as a small Mazzei gain: he captures most of Merrick, splits Keating, and loses McCall, worth roughly two to three points off a dead-even first round. — Kimi
And Gemini held near its opening number, at 44%, because it does not trust the polling at all:
I place near-zero weight on the polling record. Every survey carries a partisan label, the Club for Growth internals are suspiciously duplicated across months, and the most credible snapshot is 19 days stale. — Gemini
What none of them could do is close the gap to 22 cents. The most Mazzei-friendly seat on the panel still gave Drummond a 44% chance. The market gives him half that.
So Why Is The Market At 78?
Three honest possibilities, and the piece is better for saying we cannot separate them.
The first is that traders are pricing the Trump endorsement as close to decisive in a Republican primary, and treating the rest as noise. That is a heuristic with a real track record behind it, and it is not one the panel's arithmetic can refute.
The second is that the market can see something the panel cannot. Campaigns and the people who follow them closely watch the early-vote file. Early voting ran August 20, 21 and 22. Over that stretch Mazzei's contract went from 72 cents to 79 and Drummond's fell from 28 to 22, on the heaviest volume the board has seen since the week of the June primary. That is not a stale market drifting. That is money arriving during early voting and moving one direction. When traders are pricing something more recent than the evidence the models were given, the models cannot know it.
The third is that the market is anchored to a May 29 repricing that was correct on the day and has not been re-examined since, and that a 44-44 poll released in August simply never got the attention the endorsement did.
Our own reading, stated plainly: the panel's arithmetic on the crosstabs is sound, and a 22-cent price on a candidate who led the first round after his opponent had already been endorsed looks low. But the panel is reasoning from a nineteen-day-old snapshot, and the market is not. That is a real asymmetry, and it is the reason the gap is a question rather than an answer.
What Would Change The Panel's Mind
- A Keating or McCall endorsement before Tuesday. Their voters are 30.3% of the primary electorate and neither man has spoken. All eight seats named this as the single event that would move them most, which is the only thing the panel agreed on unanimously. It moves toward whichever man it names, and if Tuesday arrives with both still silent, that silence is itself a mild signal for Drummond, whose lane those voters already sit in.
- Any poll fielded after August 20. There is currently no public survey that includes a single day of early voting. One would be the most valuable new information available on this race.
- County-level early-vote totals from the Oklahoma State Election Board. Drummond leads the two major metros by three points and Mazzei leads rural counties by three points, so the geographic mix of who has already voted maps directly onto the margin.
- A ruling or filing in the Inola smelter litigation. The lawsuit is Drummond's signature act and Mazzei's central attack line. Movement in either direction re-argues the closing week.
Settlement Timeline
| Date | What happens |
|---|---|
| Aug 20 to 22, 2026 | Early in-person voting (completed) |
| Tue, Aug 25, 2026 | Runoff election day |
| Sep 1, 2026 | Expected contract expiration on Kalshi |
| Nov 3, 2026 | General election against Democratic nominee Cyndi Munson |
This page is re-scored when the story moves. Prices as of August 23, 2026. Panel verdicts generated August 23, 2026.
The panel's calls are on the record. Our 8-model AI panel puts a graded, public verdict on the other governor's race the market thinks it has already solved, in the live model verdict on California.
The Bottom Line
Gentner Drummond has spent this campaign doing the thing that made him a 22-cent underdog. He sued to stop a $4 billion project the president wants built, he kept suing after the president endorsed his opponent, and he still finished first in June.
The market has decided that is a losing hand. Eight models that never saw the price looked at the same evidence and called it a coin flip that leans slightly against him. Both cannot be right, and Tuesday settles it in public, which is the entire reason we publish the number before the result instead of after.
If Drummond wins, a contract the market priced at 22 cents settles at a dollar, on a board with more than half a million contracts traded on each side. If Mazzei wins, the panel spent a lot of arithmetic on a question a single endorsement had already answered in May. We will grade it either way.
Prices on this page are Kalshi's book. If you also trade on Polymarket, code OS4 gets new users a $20 bonus on a $10 deposit — affiliate link; terms as stated by Polymarket; 18+, availability varies by state.
