Prediction markets notched their first high-profile miss of Tuesday's US midterm elections, as Democrat Annie Andrews ousted Republican Darline Graham to unexpectedly win a US Senate seat in conservative South Carolina.

Prediction markets Kalshi and Polymarket had implied a 70% chance of Graham, the sister of the late US Senator Lindsey Graham, winning that race as of late afternoon Tuesday.

Andrews, a pediatrician, is the first Democrat to win a Senate seat in South Carolina in 28 years. Republican President Donald Trump carried South Carolina by 18 percentage points just two years ago, and neither Democrats nor Republicans considered the race to be competitive until recent weeks.

The outcome of several other competitive Senate races that have drawn millions of dollars in election bets will be under scrutiny. Many of those races were trending as the markets implied, with Democrats claiming Senate seats in Texas and Ohio.

With more than $1.48 billion wagered on Tuesday's US midterm elections, prediction markets are facing their biggest test yet of how reliably they can forecast political races amid growing scrutiny of their businesses.

The platforms rose to prominence in 2024 when they bucked traditional opinion polls in signaling Trump would win the presidency, and have since amassed a global following among media outlets and investors.

Though the markets cite high accuracy rates in providing betting odds, some high-profile misses during the US primaries and in Brazil's presidential election have shown they are fallible, and analysts will be scrutinizing their midterm performance.

Spokespeople for Kalshi and Polymarket did not immediately respond to requests for comment late on Tuesday. The companies have previously said that events priced at relatively low probabilities frequently happen and political upsets are no different.