As anxiety and anticipation build ahead of the Nov. 5 presidential election, political bettors are setting aside emotion and hunting for ways to profit off the mixed signals being sent by polls and prediction markets.

While traditional voter surveys show a toss-up race, prediction markets have swung hard toward Donald Trump in recent days, pushing his odds of winning to 60% or above and giving rise to what look like price anomalies across different websites.

That’s where bettors want to play. 

Bart Hanson in Boston recently made bids on both Kamala Harris and Trump, taking advantage of slight pricing differences across markets. Jacob Skaaning in Copenhagen is hedging his crypto exposure by putting money on the vice president. And Kubs Lalchandani in Miami sees markets like Polymarket and PredictIt moving too far away from the polling numbers — and is betting on Democrats retaining the Oval Office.

“The polls are showing a 50-50 race yet you can buy Harris at less than 40% odds,” said Lalchandani, a 47-year-old lawyer who has bet more than $1,000 on a victory by the vice president and plans to up that amount by around 10 times if odds continue to predict a Trump win. “Any sports bettor will tell you that’s a decent bet.”

Prediction markets have enjoyed heightened visibility in the run-up to the election, particularly after reports emerged that a trader had spent $45 million wagering on Trump on Polymarket. As of Monday afternoon in New York, the website was giving the former president a 66% chance to win.

Predictive wagers are made via so-called “event contracts.” On PredictIt, for instance, users can bet “for” or “against” a certain event occurring. Each share is priced between 1 and 99 cents, and if that event does happen, it pays out $1.

In an era when legal sports gambling has pushed betting into the mainstream, many observers have come to view prediction markets as more authoritative than traditional polling. As the argument goes, these sites are effective at predicting political outcomes because participants have financial skin in the game. The betting markets have also benefitted from some high-profile polling misses, including in 2016 when Trump’s win shocked the political world.

Critics contend betting markets suffer from sampling errors too. For example, Polymarket requires wagers be placed with cryptocurrency, which tends to skew male, as do voters for Trump.

“Not all prediction markets are created equal,” said Tom Miller, faculty director of the data science program at Northwestern University.

In his research, Miller uses PredictIt, which he said is likely to be more accurate because the platform has an $850 cap on individual bets, reducing the ability of any one person to move the market.

In many cases, bettors making political wagers are setting aside their personal political leanings in pursuit of profit. Hanson, a professional poker player who describes himself as a “Romney Republican,” is voting Libertarian in the race. Yet he backed Harris winning the state of Nevada on a platform called Kalshi and has Trump taking it on BetOnline.

The sites priced the odds slightly differently, so Hanson said he stands to gain no matter who wins the state. On BetOnline, he put $6,000 on Trump for the chance of winning $11,000. On Kalshi, he wagered $4,500 on Harris for a potential payout of $11,000.

“You can find arbitrage,” says Hanson, who plans to bet around $100,000 on this election. “Either way I win $11,000 but I’ve only put up $10,500.”

For Skaaning, bets on Harris are a hedge, even though he doesn’t think she’ll win. The 39-year-old crypto trader said a Trump win would send his Bitcoin positions upward, given the Republican’s embrace of digital assets. If Harris wins, he thinks the opposite could happen, so earnings from his bet on her victory could buffer crypto losses.

“The odds were so good,” he said. “I can bet against my own team. It’s all brain and no heart.”

This article was provided by Bloomberg News.