Donald Trump, at the time of writing, has a 47pc chance of winning November’s US election. On July 16, three days after the Republican presidential candidate narrowly avoided death at a Pennsylvania rally, it peaked at 72pc. At the start of the year it was around 40pc.

These numbers are based on close to $1bn (£750m) wagered on Polymarket, a cryptocurrency-based predictions website, by wannabe super-forecasters who believe they can beat the polls.

Trump’s fortunes on the site have swung dramatically in the last few weeks, as he has faced a change in opponents and dealt with multiple assassination attempts, alongside wavering polls.

What has not changed is that at any second, a single number sums up his chances in black and white.

After his first assassination attempt, his odds shot up dramatically in real-time; conversely, they started crashing as soon as the former president began talking about immigrants eating pets as he debated Kamala Harris.

Despite some polls that give Harris leads of five points or more, the cryptocurrency gamblers say Trump’s chances are being underestimated, suggesting the race is a toss-up.

More than $160m has been bet on a Trump victory on Polymarket, with one high roller betting $2.5m. The site has consistently given the former president a better chance than opinion polls suggest.

Election forecasts have been around for decades, but the billions being wagered, the tight race and decreasing trust in opinion polls mean so-called prediction markets, of which Polymarket has become the biggest, are being watched more closely than ever.

The sites operate more like financial markets than gambling sites. Users buy and sell contracts, for example, one that pays out a dollar if Trump wins. As the chance of the contract paying out rises, its price goes up, providing an indicator of the Republican’s chances. If the Trump contract trades hands for 60 cents, for example, it implies a 60pc chance of Trump winning.

Polymarket’s odds of a Trump or Harris victory have become regularly cited in news reports and punditry.

However, it is just one world event that the superforecasters on the website are tracking. Polymarket currently suggests there is a 39pc chance that Israel will invade Lebanon by November; a 50pc chance that Robert Jenrick will be the next Conservative leader, and a 23pc chance that Taylor Swift will be engaged by the end of the year.

On an alternative prediction site, Manifold Markets, punters believe there is a 48pc chance that Covid-19 came from a laboratory, a 50pc chance that artificial intelligence will surpass humans by 2030, and an 18pc chance that a third world war will break out before the next instalment in the Grand Theft Auto video game series.

Gambling sites have always been seen in online circles as a better predictor of politics than the polls or the punditocracy, for the simple reason that they have more skin in the game.

People may have reasons to lie, but are unlikely to willingly back a losing horse. Those who are consistently wrong are unlikely to remain liquid, and the wisdom of crowds is typically seen as a better forecaster than a panel of experts.

Prediction sites differ in being able to respond in real-time, rather than having odds set by a central party. The sites are also linked to the cryptocurrency world and Silicon Valley, giving them a contrarian flavour. Polymarket requires users to bet in USDC, a “stablecoin” linked to the dollar’s value, and is based offshore owing to anti-gambling laws in many US states.

The site’s star has risen after users predicted a series of events ahead of much of the chattering classes. The site had the odds of Joe Biden dropping out of the presidential race regularly above 50pc, even when Biden himself insisted he had no plans to. It predicted that Trump would choose JD Vance as his running mate before betting websites.

“Polymarket is officially becoming the internet’s source of truth. It’s the new news,” Shayne Coplan, the site’s 26-year-old chief executive, tweeted after Vance was confirmed.

The company has secured $70m in funding from investors such as Peter Thiel, the Trump-supporting libertarian tech investor, and Vitalik Buterin, the cryptocurrency pioneer. It has also signed up Nate Silver, the election-forecasting guru, as an adviser.

The backers of prediction markets say they are not merely seeking to incentivise gambling. They believe that financially-backed predictions can serve as a form of truth about the future; a crystal ball of sorts.

If we had better forecasts about the future, the idea goes, we would all make better decisions: whether to quit our jobs if a recession was likely, or whether to buy a house near a new rail line if there is little chance it actually gets built.

Proponents have even proposed a form of government, Futarchy, in which financially motivated parties bet on which policies will successfully realise the will of the people, with the market in turn influencing what the government implements.

Leighton Vaughan Williams, an economist at Nottingham Business School who has studied prediction markets in depth, says they have their problems. Like all markets, they are open to manipulation: a campaign may bet heavily on their own candidate to provide the illusion of momentum, discouraging rivals and securing fundraising.

Earlier this month, millions of dollars were wagered in an attempt to drive Harris’s odds up on Polymarket, in an attempt to profit from a derivative market, in which people bet on the status of the first market. The move briefly saw the Democrat’s odds spike.

Vaughan Williams says these attempts do not last long, as professional investors sweep in. “If [manipulators] create a mispriced market then the big, professional money comes in and mops it up.”

However, the practice has led to concerns that prediction markets could interfere with the democratic process.

“[Prediction markets] could harm public perception of election integrity and undermine confidence in elections,” the US Commodities and Futures Trading Commission argued this year, in response to an attempt to legalise the contracts.

They are also vulnerable to those with inside information.

In 2003, on an earlier prediction market, the odds that Saddam Hussein would be captured before the end of the year rapidly spiked, from the mid-20s to almost 100pc, after a heavy bet that the Iraqi dictator would be caught. A few hours later, Iraqi media confirmed the news: experts believe an insider with knowledge of the top secret operation placed a bet that moved markets.

“The markets in that sense had tomorrow’s news today,” Vaughan Williams says.

Polymarket is looking to cash in on the idea that its base of users provides a window into the future, launching newsletters and licensing its data to news organisations and forecasters.

However, despite its newfound influence, the legal status of cryptocurrency betting markets, at least in America, remains something of a grey area.

Kalshi, a rival site, won a court victory earlier this month that allowed it to provide political prediction markets in the US. Polymarket, which stopped operating in America after a $1.2m settlement with US regulators, has said it will remain offshore.

That has done little to dent its rise, however. If Trump wins in November, and prediction markets have once again beaten the polls, it will be time to start paying more attention to the crypto super-forecasters.

Broaden your horizons with award-winning British journalism. Try The Telegraph free for 3 months with unlimited access to our award-winning website, exclusive app, money-saving offers and more.