Prediction markets strengthened their standing as election-forecasting tools on Tuesday, with platforms favoring Democrats to win control of both the House and Senate before the first polls closed. In several races, market projections also appeared before those of major news outlets and before substantial vote counts were available.

Markets had put Democrats as narrow favorites to win at least 53 Senate seats, with a 54th considered a toss-up. Democrats have so far won 54 seats, including Iowa, which markets had rated evenly, and South Carolina, where Republicans had been slight favorites. Results were still outstanding in Mississippi and Alaska; markets gave Democrat Scott Colom and Mary Peltola a 90% chance of winning, respectively.

Platforms shared projections on X during election night. Kalshi spokesperson Jack Such said the company was confident making a call when a candidate’s odds exceeded 99% and trading in favor of the opponent had dried up. Polymarket said it called 10 key races a median of 31 minutes before CNN, Fox News and the Associated Press. CEO Shayne Coplan said in a post that markets were increasingly giving users real-time information.

Market odds also shifted quickly as vote counts began. In Texas, where Democratic state Representative Gina Hinojosa defeated Republican Governor Greg Abbott, her Polymarket odds rose from 39% to 73% within five minutes of the first polls closing. In South Carolina, Democrat Annie Andrews briefly became Kalshi’s favored candidate shortly after 7:30 p.m., about 30 minutes after polls closed.

Flip Pidot, a board member of The Coalition for Political Forecasting, said financially motivated analysts around the world can provide signals quickly. But Michelle Kendler-Kretsch, a researcher at the Anti-Corruption Data Collective, argued that a successful night should not end calls for stronger regulation. She cited a Reuters/Ipsos survey finding that most Americans oppose using prediction markets to trade on elections, and said the markets’ risks had not changed because they were accurate this time.

The platforms have also missed major contests. Before the midterms, Kalshi gave Brazil’s Luiz Inácio Lula da Silva about a 20% chance of winning the country’s presidential race. Markets also sharply misjudged some high-profile Democratic primaries over the summer, including races in Wisconsin and Michigan.

Researchers noted that a candidate with low odds can still win without the forecast being mathematically wrong. Barnard College economics professor Rajiv Sethi compared this to the possibility of getting two heads when tossing two coins. Dartmouth economics professor Eric Zitzewitz said casual users may mistake a 55-to-45 probability for a decisive lead, though he expects public understanding of market odds to improve as their use grows.

Experts said prediction markets are likely to remain part of election coverage through the next presidential race. Rutgers statistics professor Harry Crane described this election as the first in which prediction markets played a fully established role.

Disclosure: CNBC and Kalshi have a commercial relationship that includes customer acquisition and a minority investment.

Sources: CNBC Markets.