How accurate is Polymarket? Its favorite won 238 of 273 primaries in 2026

01 · The testHow often does Polymarket's favorite win?

Election news now quotes Polymarket's odds next to the polls. Polymarket is a betting site for real events, like elections and football games. So how often are its odds right? This year's primaries make a fair test. The site took bets on more than 270 of them for the Senate, the House and governor. All of them have now been decided.

Take Wisconsin. On the morning of the Democratic primary for governor in August, traders on Polymarket had Francesca Hong as a near-lock. David Crowley won by less than a point. He had quit the race in July and come back ten days later with the governor's backing.

On Polymarket you buy shares in an outcome. Each share pays $1 if it happens and nothing if it doesn't, so the price is the crowd's guess at the odds. Hong's shares cost 96 cents that morning, and Crowley's cost 4 cents. It is the same kind of bet the site runs on crypto prices every five minutes.

The site settles every bet on Polygon, a public blockchain, so every trade is on the record. Bitquery indexes Polygon, the same data we used to find how much of that chain Polymarket uses. For each race, we took every candidate's price on the morning of the vote, before any polls closed, and checked it against the result.

Across 273 races, the favorite, the candidate with the highest price that morning, won 87% of the time. When the site was very sure, it was nearly always right. When it was only fairly sure, it was right less often than its prices said, and that cost anyone who backed every favorite.

02 · The scorecardSure bets came in, close calls fell short

A good price comes true about as often as it says. A favorite at 80 cents should win about eight races in ten. So we grouped the races by the favorite's price that morning.

At the top, the record is close to perfect, as the chart shows. Lower down it slips. Favorites priced between 50 and 90 cents won 71% of their races, though they cost 77 cents on average. That gap is small, but it costs real money. Put $1 on every favorite on voting morning and you'd have lost 4 cents on the dollar.

The sure bets broke even, and the losses came from the favorites in between. The value sat with the underdogs nobody wrote off. A dollar on every candidate priced between 5 and 30 cents paid back at least $1.38. Candidates priced around 20 cents won about a third of their races. Those groups are small, so a few wins swing them. Real longshots were a bad bet. Almost every candidate under 5 cents lost.

03 · By raceHouse races were the hardest to call

Senate and governor races went to the favorite about nine times in ten. House races did worse.

A House primary is a small, local contest. Few traders follow it, and many of these markets traded less than $50,000 in all. You might expect those thin markets to be the weak spot. They weren't. Races with less money went to the favorite more often.

Money goes where the race is close. In the thin races, the favorite was usually a near-lock, and it usually won. The races in the middle band of money were closer calls, and that is where the prices were furthest off.

04 · The missesPolymarket's biggest misses in the 2026 primaries

Wisconsin was the year's costliest miss. Its market traded $1.6 million, more than any other race the favorite lost. In the last week before the vote, Hong's daily average price never fell below 89 cents.

Hong was one of 5 favorites priced at 90 cents or more that lost. Two of the others were tiny House markets. Half of the misses in the table below were decided by 5 points or less, and Iowa and Wisconsin by less than a point.

The oddest result came in New Jersey. On the morning of the Republican Senate primary, Justin Murphy's shares cost 0.4 cents. Alex Zdan was the favorite at 79 cents. Murphy won by 4 points, and Zdan came third.

05 · The last weekWhen the favorite changed in the final days

Prices kept moving right up to the vote, but the last week barely changed the record. The favorite a week out won only two fewer races than the favorite on voting morning. In 15 races the favorite changed hands in that last week. The new favorite won 7 of them and the old one won 5. In the other 3, someone else won.

The most traded House primary on the site shows how fast it can turn. In Kentucky's 4th district, Republican Thomas Massie faced Ed Gallrein, a challenger backed by Donald Trump. Massie was at 70 cents seven days before the vote and 40 cents the day before it. Gallrein won by about 10 points.

Minnesota's Republican race for governor flipped faster. Mike Lindell, the MyPillow founder, had Trump's backing and was at 84 cents five days before the vote. Two days before it, Lisa Demuth led at 55 cents. We found no news in those days that explains the move. Demuth won by 12 points.

Late moves are not always about polls or endorsements. Prices can also move on news that some traders get first, as with the bets the CFTC is looking into and the accounts that beat the odds on KPMG's clients. We did not test these primaries for that.

06 · The big racesTexas and Michigan: the market moved before the vote

The most traded primary of the year was the Republican Senate race in Texas. Ken Paxton faced Senator John Cornyn in a runoff, a second vote between the top two, on 26 May. Donald Trump endorsed Paxton a week before it. Paxton was at 78 cents that day and 94 cents the next. He won by 28 points.

In Michigan's Democratic Senate primary, Mallory McMorrow was the favorite in early April, in light trading. She quit the race in July, which left Abdul El-Sayed and Haley Stevens. El-Sayed went from 75 cents seven days before the vote to 96 cents three days later, as a poll put him 15 points ahead. He won by about a point, a far closer count than a price of 98 cents suggests.

07 · Who got paidWho held the upsets

Every trader shows up on the chain as a wallet, an account with its own address. Some wallets are people and some are bots, as we found when we measured how much of Polymarket is bots. Every upset has two sides: wallets that held the favorite into the vote, and wallets that held the winner. We looked at the 8 upsets with the most money.

In those races, more than twice as many wallets held the favorite into the vote as held the winner. The favorite's backers lost more than the winner's backers made. In Wisconsin, 80 wallets held Crowley into the vote, and their shares paid out more than ten times what they cost.

08 · The answerSo, how accurate is Polymarket?

Quite accurate, but too sure of its favorites. On the morning of the vote, Polymarket's favorite won 87% of this year's primaries. Its sure bets were nearly always right. Its close calls came true less often than the price said, and backing every favorite would have lost money.

That matters for November. Polymarket's traders give Republicans 35% to keep the Senate. In these primaries, candidates priced between 30 and 50 cents won about a third of the time, so a price in that range is a real chance. You can check any of these markets yourself through Bitquery's Polymarket API, or get the full trade history from the Bitquery Data Store.

09 · The recordAddresses behind this story

FAQ

How accurate is Polymarket?

In the 2026 primaries for the Senate, the House and governor, Polymarket's favorite on the morning of the vote won 87% of the time. It almost always won when priced at 90 cents or more, and won less often than its price said in closer races.

How often are Polymarket predictions correct?

The candidate Polymarket favored on voting morning won 238 of 273 primaries this year. The surer the price, the more often it came true: favorites at 90 cents or more won 177 of 182.

Are Polymarket odds reliable?

At the top end, yes. Favorites priced at 95 cents or more broke even for anyone who bought them. In closer races the price ran too high for the favorite, and in thin markets a price can rest on a handful of trades.

How accurate is Polymarket for elections?

In this year's primaries, Senate and governor races went to the favorite about nine times in ten, and House races 85% of the time. Its sure bets came in, but its close calls came true less often than the price said.

What was Polymarket's biggest miss in the 2026 primaries?

The Democratic primary for governor of Wisconsin. Francesca Hong was at 96 cents on voting morning. Crowley beat her by less than a point.

Did Polymarket call the Texas Senate primary?

Yes. Ken Paxton was the favorite before the 26 May runoff against John Cornyn, and his price jumped from 78 cents to 94 cents the day after Donald Trump endorsed him. It was the most traded primary market of the year.

Did Polymarket call the Michigan Senate primary?

Yes. Abdul El-Sayed was at 98 cents on voting morning and won by about a point. In the last week of July his price rose from 75 cents to 96 cents.

Did Polymarket call Thomas Massie's primary?

Only in the last few days. Massie was the favorite seven days before the vote, at 70 cents. By voting morning, Ed Gallrein led at 61 cents, and Gallrein won.

Can you win money betting on Polymarket favorites?

Not in these primaries. One dollar on every favorite on voting morning would have lost 4 cents on the dollar. These are past prices and say nothing certain about future bets.

Where can I get Polymarket data?

Bitquery records every Polymarket trade on Polygon. You can query it through the Bitquery Polymarket API or buy the full history as files from the Bitquery Data Store.