On Friday, Polymarket’s “Presidential Election Winner 2024” market showed bettors giving Trump a whopping 60.1% chance of winning, while Harris’ chances were only 39.8%. That’s a massive difference from the start of October when the candidates were in dead heat.

To be sure, polls have also tightened, and due to the Electoral College small shifts in a few states can lead to a decisive outcome. Still, some experts praise prediction markets for their accuracy under the premise that since real money is on the line, it serves as a better indicator of voter sentiment. But there are signs prediction markets—Polymarket in particular—have been manipulated by pro-Trump bettors. 

On the “Presidential Election Winner 2024” market, four users have flooded the market with massive bets: Fredi9999, Theo4, PrincessCaro, and Michie, the Wall Street Journal first pointed out. In all, their bets account for more than 41 million shares that Trump will win, amounting to nearly $26 million in current value. These four accounts have been pumping money into the presidential prediction markets during the past few weeks, even as recently as yesterday. 

“I believe there has been a coordinated effort to change the perception of this race,” Tom Bonier, senior adviser at research firm TargetSmart and veteran political strategist, told Fortune. “A central argument has emerged in the closing weeks of this campaign: strength versus weakness. Donald Trump’s persona, and therefore his support from voters, relies on being seen as strong. But if the public perception is that he will lose, that all falls apart.” 

To that point, Trump supporters, including Tesla CEO Elon Musk, have drawn attention to the fact the former president is leading the prediction markets. 

“More accurate than polls, as actual money is on the line,” Musk posted on X earlier this month.

Prediction markets, which have grown in popularity, allow traders to buy and sell shares on the outcome of a given event, like whether Trump will get impeached or who will win the next Super Bowl.

Traders buy shares depending on which outcome they think is more likely. Share prices or “odds” rise and fall depending on demand, so if the event happens the way the trader predicted, the contract or “bet” will rise to $1 and pay out—or fall to $0 if it does not.

Polymarket did not respond to Fortune’s request for comment about whether they’re investigating potential influence or interference. Sources told the Journal that Polymarket is investigating the recent election bets with help from outside experts.

Are prediction markets accurate?

The argument for prediction markets is pretty much what Musk said: since people are putting real money on an outcome, they’re likely to be more accurate than the polls. 

“Political betting sites are the best at predicting the wisdom of the crowd,” Thomas Miller, a data scientist at Northwestern University, told Fortune’s Shawn Tully. Miller is renowned for his accurate predictions during the 2020 election, and much of his methodology comes from examining prediction markets.

With financial motivations, bettors are more motivated to analyze polls, current events, and other factors to make well-informed bets, Ryan Waite, vice president of public affairs at public affairs consultancy Think Big, told Fortune. 

But this isn’t what appears to be happening in the election prediction markets. 

“Unlike polls, prediction markets can be influenced by the interests of a small number of large players, as appears to be happening [on Polymarket],” Waite added. “The limitation here is that prediction markets often attract a specific kind of participant: people interested in betting who might not reflect the broader population and they can be influenced by deep pockets, overconfident individuals or herd mentality.”

The argument that prediction markets would be more accurate than polls also “doesn’t make sense,” Bonier said.

“For that to be true, we would have to believe that those who are betting have some level of expertise or inside information that is not otherwise available or apparent in polling data,” he said. “Yet, we know almost nothing about who is betting on Polymarket, nor why.”

Nicholas Creel, an associate professor of business law at Georgia College & State University and political scientist, told Fortune that for prediction markets to be accurate, they need to be fairly large and heterogeneous.

In other words, prediction markets need a large number of people of different backgrounds and demographics betting money on who they think will win. That way, the “wisdom of crowds” theory kicks in, like Miller argues. 

But there are a few red flags when it comes to Trump’s Polymarket predictions, Creel said. 

“In Polymarket in particular, you don’t have a particularly large or diverse crowd of people,” Creel said. “A large driver of this is that the exchange requires users to bet with cryptocurrency, which the vast majority of the population has little to no familiarity with. Cryptocurrency tends to be something mostly loved by those on the far right, particularly your libertarian types.”

Even individual markets show varying outcomes

While the “Presidential Election Winner 2024” market shows Trump with a great advantage over Harris, another market shows a different story. For the “Popular Vote Winner 2024” market, bettors have a 66% chance of Harris winning over Trump’s 34%. 

And that likely has to do with the number of shares large bettors have invested in these two separate markets. As is shown in the screen grabs from the two markets, the large bettors—Fredi9999, Theo4, PrincessCaro, Michie—invested heavily in the “Presidential Election Winner 2024,” but didn’t buy nearly as many shares in the “Popular Vote Winner 2024.”

This discrepancy correlates with the idea that bettors are attempting to influence the public perception of the presidential race, Bonier said. 

“On one hand, it could just reflect the potential divergence between the two outcomes,” he explained. “But on the other hand, if someone were to invest in manipulating the public perception of this race, there is no real benefit to spending money on the question of who will win the popular vote, when the only thing that matters is the electoral college.”