The prediction accuracy of the Polymarket platform averages 94.1% four hours before the market closes, according to a dashboard created by data processing specialist Alex McCallow on Dune.

Even a month before an event is resolved, the accuracy level remains quite high, around 90%.

However, the expert noted that Polymarket markets clearly exhibit bias. He attributed this to the tendency to conform, herd instincts, and low liquidity.

In an interview for the platform’s blog, McCallow stated that investors generally overestimate the probability of an event by one or two points.

Long-term forecasts tend to be more accurate, he clarified. This is because they often include unlikely elements. For example, for the U.S. presidential election, the platform listed 17 candidates, including rapper Kanye West. However, it was clear that only Donald Trump and Kamala Harris had a real chance of winning.

This is why sports event markets, such as “one-on-one” matchups, show lower accuracy—around 80%. However, their accuracy steadily increases as the event approaches.

McCallow admitted that his research was hindered by the lack of significant historical data on Polymarket. In some segments he wished to analyze, the data amount “looked simply ridiculous.”

According to Polymarket Analytics, the volume of bets on the U.S. presidential election became a record on the platform, approaching $3.7 billion. Sports have become a rapidly growing sector, with the top events including NBA Finals (~~$1.6 billion), NFL (~~$1.2 billion), UEFA Champions League (~~$1 billion), and the English Premier League (~~$0.8 billion).