Wisconsin Governor Primary Odds: The AI Model Verdict

Updated August 11, 2026 · 8 min read · By Jake Hari

Wisconsin votes Tuesday, and its Democratic primary for governor has been one of the most volatile races of the cycle: a retiring incumbent, a late endorsement, and two candidates exiting in the final weeks. Kalshi has turned the whole thing into a market, one yes/no contract per candidate, each resolving YES only for whoever wins the nomination. We asked six AI models to score the entire field the night before the vote without letting any of them see a single price. They came back with a clear front-runner, State Rep. Francesca Hong, but with more doubt than a near-95-cent market is showing, and with a bigger opening for Milwaukee County Executive David Crowley than the price implies.

The Quick Answer

Kalshi's market makes Francesca Hong the heavy favorite near 95 cents, treating the nomination as all but decided. Our price-blind, six-model panel leans the same way but not as hard: it blends Hong to about 82% and gives David Crowley roughly 14%, against a market price of just 5 cents. The gap to the market has narrowed since our first read on this race, because the polling firmed, but the models still price the late-race chaos and a large undecided bloc as real uncertainty a near-certain price does not.

• Market Front-Runner: Francesca Hong, ~95¢
• Model Front-Runner: Francesca Hong, 82.3% blend (still first, but short of the market's near-certainty)
• Where The Models Diverge Most From The Price: David Crowley, a ~14% blend against a 5¢ market price
• The Polling Picture, Firmed: every early-August poll, including Crowley's own internal survey, showed Hong leading by roughly 18 to 21 points
• The Panel's Logic In One Line: a corroborated ~20-point lead the day before a plurality primary is hard to erase, but a 34% undecided share and a single dominant pollster are why the models stop short of 95 cents.

The case for Hong, where the models part from the market on Crowley, the full model-by-model board, and a plain-English guide to reading these prices are all below.

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. Sign up in the box at the end of this article.

New to Kalshi event markets? Start with our plain-English guide to how Kalshi contracts actually work →

The Market At A Glance

Venue	Kalshi, a CFTC-regulated event-contract exchange (18+; availability varies by state, as of August 2026)
The Event	"Wisconsin Democratic Governor nominee?" (Kalshi event KXGOVWINOMD-26), one yes/no contract per candidate
The Contract	Each market resolves YES only for the candidate who wins the Democratic nomination for governor at the August 11, 2026 primary, per Kalshi's settlement rules
Settlement Quirk	Wisconsin primaries are decided by plurality, no runoff, so the nominee is whoever gets the most votes. Withdrawn candidates cannot win even if still printed on the ballot
Prices Below As Of	August 11, 2026 (latest snapshot from a thinly-traded board; prices on this page refresh through the vote)

These are model estimates, not predictions of fact and not financial or voting advice. Kalshi event contracts trade on a CFTC-regulated exchange; you must be 18 or older and in an eligible state to participate, and this article does not advocate for or against any candidate.

The Board

Here is the narrow, phone-friendly view: the four active candidates, the market price, and the panel's blended estimate. The full model-by-model table is just below it.

Candidate	Status	Market	AI Blend
Francesca Hong	Active (poll leader)	95¢	82.3%
David Crowley	Active (Evers-endorsed)	5¢	14.3%
Kelda Roys	Active	<1¢	1.2%
Joel Brennan	Active	<1¢	1.2%
Withdrawn / Not Running	Barnes, Rodriguez, Evers, Kaul, Wikler, Hughes, Hulsey	<1¢	~0%

These are model estimates, not predictions of fact and not financial or voting advice. The blend is the simple mean of the six model numbers after a revision round; the withdrawn and non-candidate names carry only leftover "field" mass and cannot win the nomination.

More live boards from the same panel: Florida Republican governor primary (Byron Donalds ~98¢, the same panel's other August gubernatorial-primary board) · Michigan governor (Jocelyn Benson ~84¢) · Michigan Senate (Abdul El-Sayed ~54¢ / Mike Rogers ~46¢). Prices fetched August 11, 2026.

Every Seat's Number

Candidate	Market	Fable	Opus	Sonnet	GLM	Kimi	DeepSeek	Blend	Gap (blend% − mkt¢)
Francesca Hong	95¢	81.0	86.0	82.0	78.0	90.0	77.0	82.3%	-13
David Crowley	5¢	14.0	12.0	14.0	17.0	9.0	20.0	14.3%	+9
Kelda Roys	<1¢	2.0	1.0	2.0	1.0	0.0	1.0	1.2%	~0
Joel Brennan	<1¢	2.0	1.0	1.0	2.0	0.0	1.0	1.2%	~0

These are model estimates, not predictions of fact and not financial or voting advice. The Blend column is the simple mean of the six model numbers in its row, taken after a revision round in which each model read the others' anonymized reasoning. The last column is a quick directional gap, the blend percentage minus the raw market price in cents; it compares a model probability to an un-adjusted price, so read it as direction and size, not a precise edge. On this board the two live contracts (Hong 95¢, Crowley 5¢) already sum to about 100, with the nine longshots pinned at or below the exchange's one-cent floor, so there is almost no overround to strip. The withdrawn and non-candidate names are omitted from the model table because they cannot win; they still carry listed contracts at the one-cent floor.

A note on the panel. Our standard Model Verdict lineup runs eight seats. On this run two were unavailable (the ChatGPT and Gemini seats were not installed on the machine that produced these verdicts), so this board runs on six models: Claude Fable, Claude Opus, Claude Sonnet, GLM, Kimi, and DeepSeek. All six scored the field on August 11, 2026, without seeing any prices, working from a fetched, dated summary of the candidates, the withdrawal timeline, and every public poll. Each model's number also lands in our back-tested Model Verdict ledger, where past price-blind calls are graded against settled results so the method is measured, not just asserted.

The one line that explains the whole board: the market treats Hong's ~20-point poll lead as nearly decisive; the models treat it as a strong, now-corroborated lead in a race where a third of the vote was still undecided in late July. Both make her the favorite, but only one of them calls it 95 cents.

What Actually Decides This Primary

Before any name, the shape of the race matters, and it is what kept the models from matching the market. Four fetched facts define the board:

• The Incumbent Is Gone, And He Picked A Side. Two-term Gov. Tony Evers announced in July 2025 that he would not seek a third term, opening the seat. A year later, on July 18, 2026, Evers endorsed David Crowley, who had suspended his campaign earlier that month and re-entered the same week.
• The Field Shrank In The Final Stretch. Former Lt. Gov. Mandela Barnes, who had been polling second, withdrew on Thursday, July 30, saying in a farewell video that "it's very clear who our nominee is going to be." Lt. Gov. Sara Rodriguez had ended her campaign earlier in July. That left four active Democrats: Hong, Crowley, state Sen. Kelda Roys, and former state administration secretary Joel Brennan.
• The Polls Firmed Toward Hong, And There Are Several Now. The Marquette University Law School poll fielded July 22-27 (n=407 Democratic primary voters, margin ±6.6) had Hong 38, Barnes 16, Crowley 7, Brennan 2, Roys 2, with 34% undecided, up from 26% for Hong two weeks earlier. After Barnes exited, a State Navigate poll released August 7 put Hong at 44% with Crowley about 20 points back, and 21 points back in a two-way. Even Crowley's own internal poll that week had him trailing by 18.
• It Is Decided By Plurality. Wisconsin's partisan primary has no runoff, so whoever leads the four-way field on Tuesday is the nominee, a structure that rewards a consolidated front-runner in a fragmented field.

The market is pricing Hong's lead as close to settled. The models are pricing the same lead, corroborated across several August polls, minus a discount for the roughly one-third of late-July voters who were still undecided and the fact that one firm, State Navigate, is carrying the topline.

Francesca Hong: The Front-Runner, Market And Models Agree

Every model made Hong the favorite, and no model put her below 77%. Her case is the consistency of the polling and the direction the field has moved. As Opus summarized it, every early-August poll "including Crowley's own internal, shows Hong leading by 18-21 points one day before a plurality-wins primary, with her lead firming." Kimi made the arithmetic plain: "with plurality-wins and one day left, that multi-poll margin makes her a heavy favorite."

Where the panel held back, in its own words: the models did not treat the lead as decided. The most cautious seats pointed at the same two things: the undecided bloc and the thin pollster base. DeepSeek, the low seat at 77%, weighed "a firming trend and no runoff" against the open questions in the data, and Fable defended keeping "a ~19% tail" by citing "Marquette's ±6.6 MOE, 34% undecided in its wave, and primary polls' history" of missing. That distance between a 82% blend and a 95-cent price is the undecided share the models refuse to hand entirely to the front-runner.

David Crowley: Where The Models Part From The Market

This is the sharpest split on the board. The market prices Crowley at 5 cents; the panel blends him to about 14%. The models are not predicting he wins; they are saying the crowd's near-zero price ignores a real, if secondary, path.

The reasoning was consistent across seats. Crowley re-entered in mid-July with Evers' endorsement, the most valuable in-state nod, and has since consolidated the moderate and establishment lane into clear second, and his State Navigate share climbed as Barnes exited. DeepSeek, the high seat on Crowley at 20%, read the freed Barnes vote and the undecided bloc as variance that flows to the consolidated runner-up, not the front-runner. The models still keep him second because none of that has closed a corroborated ~20-point gap in the measured data, and the clock is at zero.

Where The Panel Changed Its Mind

After round one, every model read the other seats' anonymized numbers and reasoning and could revise. Most held; the movement that did happen is instructive.

• Opus Moved, And Said Why. It cut Hong from 89% to 86%: "The cluster of peers pricing Hong at 77-82 correctly weighted the large undecided share and low-turnout polling error I under-counted at 89; that variance realistically flows to consolidated-second Crowley, so I lifted him 10 to 12." — Opus
• DeepSeek Gave The Longshots A Real Floor. It moved Roys and Brennan off zero: "I had previously given them 0%, which is overconfident given they are still on the ballot and could theoretically benefit from a black-swan event. I now assign them 1% each." — DeepSeek
• Sonnet Held, On Purpose. "No peer introduced a new data point, mechanism, or risk factor I hadn't weighed — the spread from 77% to 90% is just differing risk appetite on the same evidence." — Sonnet

The net effect was small (Hong's blend eased from 82.8% to 82.3%, Crowley firmed from 13.9% to 14.3%), which is itself the finding: the panel's disagreement here is about calibration on shared facts, not a missing piece of evidence.

What Would Change The Panel's Mind

This board settles within a day, so the triggers are near-term and concrete. Each would push the number a specific direction:

• A Youth-Turnout Surge. Hong drew 66% among voters 18-29 in the Marquette poll. If Tuesday's electorate skews younger than a typical August primary, that pushes toward Hong and toward the market's 95¢.
• The Undecided Bloc Breaking Late To The Moderate Lane. Roughly a third of Democrats were undecided in late July. If that vote consolidates behind the Evers-endorsed Crowley in the booth, it pushes toward Crowley and widens the model-vs-market gap.
• A Heavy Early/Absentee Margin For Either Candidate. Reported early-vote splits before polls close would move the number toward whoever leads them, since a banked lead is harder to erase in a same-day plurality race.
• A Single High-Quality Independent Poll Contradicting The ~20-Point Lead. The topline currently leans on State Navigate plus Marquette's late-July wave; one credible late poll showing a tighter race would push the panel toward Crowley, while another confirming ~20 points would push it toward the market.

Dark Horses The Panel Won't Dismiss

The two other active candidates are true longshots, but the panel gave each a real, if tiny, number rather than zero.

• Kelda Roys (Blend 1.2%). The Madison state senator has run statewide before and polled low single digits throughout; DeepSeek's revision note kept her on the board precisely because she is "still on the ballot and could theoretically benefit from a black-swan event." That is longshot mass, not a hidden edge.
• Joel Brennan (Blend 1.2%). The former state administration secretary registered 2% in both the Marquette and late-July State Navigate polls. No seat gave him a live path, but Fable and GLM each left a point or two on him as honest tail risk in a four-way plurality race.

Everyone else on the board, from Barnes and Rodriguez to Evers, Kaul, Wikler, Hughes, and perennial filer Brett Hulsey, has either withdrawn or is not a candidate, and cannot win regardless of the listed price.

The Traps: Rows That Look Like Signal But Are Not

Transparency matters more than a clean story, so here is where the table would mislead you:

• The Withdrawn And Non-Serious Names Cannot Win. Barnes, Rodriguez, Evers, Kaul, Wikler, Hughes, and perennial filer Brett Hulsey still have listed contracts, but none is an active, serious candidate for the nomination — a candidate who has withdrawn does not win even if printed on the ballot. Those contracts are expected to resolve NO.
• The Crowley Split Is The Real Disagreement. The seats range from 9% (Kimi) to 20% (DeepSeek) on Crowley, an 11-point spread. That is a real calibration difference about how much of the undecided and freed-Barnes vote flows to the consolidated runner-up, not a consensus edge.
• The Bottom Of The Table Is The One-Cent Floor, Not Information. Nine contracts sit at or below the exchange's minimum tick. Do not read a 1% blend or a sub-cent price as a distinct signal.

Settlement Timeline

Primary / Decision	Tuesday, August 11, 2026 (Wisconsin partisan primary; plurality, no runoff)
Kalshi Settlement	Contracts resolve YES for the nomination winner once the result is certified; Kalshi's expected expiration is August 31, 2026
Next Catalyst	The general election is November 3, 2026, against presumptive Republican nominee U.S. Rep. Tom Tiffany
Re-Score Policy	We re-score this board whenever the story moves, and post a settlement retrospective on this same URL once the result is final. Estimates and prices above are as of August 11, 2026.

How To Read These Prices

The models can tell you where they part from the crowd, but the number in front of you still has to be understood in context. Here is how to read a market like this one, plus the honest friction of a thin, event-driven board.

Work an example. Say Hong's contract shows a 95-cent YES price. Taken at face value, that implies about a 95% chance she wins the nomination. To sanity-check it you would normally strip the board's overround by adding up every candidate's YES price to see how far past 100 it sums, but here the two live contracts (Hong 95¢, Crowley 5¢) already sum to about 100, with the nine longshots pinned near the one-cent floor, so there is almost nothing to strip and Hong's 95¢ maps to roughly 94%. That is unusual: on a crowded field the favorite's number is always a little softer than it looks, but a two-horse board like this one is close to clean. The habit still matters: checking what a fair, spread-free version of a price looks like is the whole idea behind converting Kalshi's cents into real odds before comparing anything.

That same discipline of comparing a number to its fair value is the method behind our other boards. The Florida Republican governor primary verdict runs the identical playbook on a near-identical race, where the panel likewise trims a 97-cent favorite.

Every number in this piece gets graded in public once the market settles; you can see the method's full record on the graded scoreboard.

This board lives on Kalshi and our panel does not trade it. But if you came for the method rather than this one race, OddsShopper's free expert picks today apply the same shop-the-number, find-the-fair-value discipline to the live sports slate.

A few honest points before reading too much into any number in the table above:

• Eligibility. Kalshi is a CFTC-regulated event-contract exchange, not a sportsbook. You must be 18 or older and located in an eligible state, and availability varies by state and can change.
• Liquidity And Spread. This board trades thin, most contracts at a single cent with light volume, so the "price" is softer than one number suggests, and getting in or out at your number is not always possible.
• A Moving Story. This race has already seen two withdrawals and a major endorsement in recent weeks, and prices on this page refresh through the vote; these estimates reflect what was known on August 11, 2026.
• Not Advice. Everything here is analysis for information, not a recommendation to trade or to vote. These are model estimates, not predictions of fact and not financial or voting advice.

FAQ

Who is favored to win the Wisconsin Democratic governor primary? State Rep. Francesca Hong is the front-runner. She led every early-August poll by roughly 18 to 21 points, including David Crowley's own internal survey, and a State Navigate poll released August 7 had her at 44% with Crowley about 20 points back. Kalshi's thin market prices Hong near 95 cents; our six-model AI panel agrees she leads but blends her to about 82%.

When is the Wisconsin governor primary and how does the Kalshi market settle? The Wisconsin partisan primary is Tuesday, August 11, 2026. Wisconsin primaries are decided by plurality with no runoff, so the nominee is whoever gets the most votes. Kalshi lists a separate yes/no contract for each candidate, and each resolves YES only for the winner of the Democratic nomination. Candidates who have withdrawn, such as Mandela Barnes and Sara Rodriguez, cannot win even if their names remain printed on the ballot.

What do the AI models think of David Crowley's chances? The panel gives Crowley about a 14% blend against a 5-cent market price. The models note he re-entered the race in mid-July with retiring Gov. Tony Evers' endorsement, consolidated the moderate lane into clear second, and could gain if the roughly one-third of voters who were undecided in late July break his way. They still keep him second because every August poll, including his own internal, showed Hong leading by 18 to 21 points the day before a plurality primary.

Is Tony Evers running for Wisconsin governor in 2026? No. Two-term Gov. Tony Evers announced in July 2025 that he will not seek a third term, opening the seat, and in July 2026 he endorsed David Crowley in the Democratic primary. Evers is not a candidate, so his Kalshi contract cannot resolve YES even though it remains listed.

The Bottom Line

The market and the models are looking at the same race and reaching the same front-runner with different conviction. Kalshi prices Francesca Hong near 95 cents, close to settled; the panel blends her to about 82%, still first but a step short of certain, and it gives David Crowley roughly 14% where the market pays 5. The read to take is not that Hong is in trouble; her lead is corroborated across every August poll, including Crowley's own internal. It is that a 34% undecided share in late July and a thin pollster base leave a little more open than a near-95-cent price admits. If this method is your thing, our guide to Kalshi versus Polymarket is the natural next read.

Hottest Prediction Markets Right Now

• 2028 Democratic presidential nominee · $166M traded
• Rocket Classic Winner · $124M traded
• 2027 Pro Football Champion · $51M traded
• Pro Baseball Champion · $40M traded
• Bitcoin price at the end of 2026 · $30M traded
• 2027 Pro Basketball Champion · $19M 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.

Verdicts generated August 11, 2026. Prices as of August 11, 2026. Panel: Claude Fable, Claude Opus, Claude Sonnet, GLM, Kimi, and DeepSeek, price-blind, after a revision round. These are model estimates, not predictions of fact and not financial or voting advice.