Kalshi says 479 of 487 forecasts were correct, but the record is not complete

Kalshi announced that 479 of 487 forecasts were correct after the midterm elections. The platform presented the fact that these forecasts held before voting took place as proof that prediction markets can see future outcomes in advance. The report also highlights the risk that third parties could misuse this information for their own gain.

However, relying on the surface success rate can be misleading. Polymarket had given Aleksandar Vučić, Serbia's ousted prime minister, a 70 percent chance of winning. Vučić lost to the Student List, which was formed after two years of mass protests, and he received only 39.2 percent of the vote. The race was not close at all.

Kalshi and Polymarket had given Francesca Hong a 95 percent chance of winning the Wisconsin Democratic gubernatorial primary. In the end, David Crowley won by less than one percentage point. The report notes that the platforms' claims about accuracy have repeatedly failed.

Markets can behave inconsistently because traders act on their own expectations. The people who trade also do not represent the entire electorate. A study of 2,500 political prediction markets found that these markets do not align with the principle of rational information aggregation and that there is a possible bias in trading. The platforms defend themselves with the argument of the wisdom of the crowd, but reviews show that only 1 percent of traders hold three quarters of the profits. Tracking large "whale" traders who place big bets also raises concerns about the use of insider information.

This picture shows that prediction markets may be a useful weather gauge, but they are not a crystal ball that reliably reveals the future.