Kalshi's betting markets gave Bolsonaro a big lead right up until Lula's surprise win, and the founder says that's not a failure — it's how probability actually works.

• Kalshi had Bolsonaro favored by a wide margin for weeks, but Lula pulled off the upset, winning by millions of votes.
• The founder says prediction markets aren't polls — they're constantly updating based on new information, like a stock price reacting to news.
• New information (he pointed to late vote-count disputes and legal moves around contested districts) shifted the odds fast once it came out.
• He argues a 25% chance of Lula winning isn't "wrong" just because Lula won — a one-in-four shot should hit about one in four times, and Kalshi's track record checks out across millions of past markets.
• He applies the same logic to U.S. midterm predictions: if a race is given 40% odds for one party, that party should win roughly 40% of similar races, not zero.

Outlook: Expect prediction-market odds to keep swinging sharply around contested elections as new information breaks, with their accuracy best judged over many races rather than any single result.