Geoff Mann: What’s the truth?

Geoff Mann on Prediction Markets

In the days leading up to Donald Trump’s State of the Union address in February, the prediction market Kalshi set up a trade, a wager in all but name, on the Yes/No proposition that the disgraced Republican ex-congressman George Santos would attend. The larger the number of people who bet ‘Yes’ (or, in the language the prediction markets prefer, bought ‘Yes’ contracts), the more the value of the bet would rise; if they bet ‘No’, it would fall. Santos saw an opportunity. Privately, he bet ‘Yes’, then started posting on social media that he was ‘going to be there’ for the speech, speculating about what he might wear and so on. As the money piled in, the price of ‘Yes’ rose. When he thought the time was right, Santos sold his ‘Yes’ contracts, bought ‘No’s, and then didn’t show up to the speech. He made $17,500 on the trade, which he had to return after his actions came to light, in addition to paying a fine of the same amount and being banned from Kalshi for life.

Prediction markets like Kalshi are exchange platforms that make it possible for users to take a ‘Yes’ or ‘No’ position on the outcome of future events, anything from the score in a football match to whether it will rain today in Madrid to whether the US will officially confirm the existence of aliens before 2027. The field is currently dominated by Kalshi and its main competitor, Polymarket, and tens of billions of dollars are on the line. Participants deposit money in a portfolio and use it to buy ‘Yes’ or ‘No’ shares. Each share has a value between 1¢ and 99¢, which rises and falls with the distribution of ‘Yes’ and ‘No’ bets in the market. If, for example, almost everyone predicts that Barcelona Femení will win the Spanish league this season (usually a very safe bet), then the share price will move close to 99¢; if ‘No’ dominates a market, the price will move towards 1¢.

The key feature of prediction markets is that, unlike stock markets, there is a point at which the event being traded takes place or concludes, and the market then ‘resolves’ to ‘Yes’ or ‘No’. The winners get $1 for every contract, and the losers get nothing. In the run-up to this, the price tends to move quickly towards 1¢ or 99¢ as ‘Yes’ or ‘No’ becomes the more likely outcome. If you bought a ‘Yes’ contract on a Barça Femení victory for 95¢, but with three games to go they sit fourth in the table, the price of that ‘Yes’ will be racing towards 1¢. Just as in conventional financial markets, you can try to find a buyer, but if you are left holding your stake when the season ends, you lose it all.

In principle, at every moment until the market resolves, the price of any prediction is an up-to-the-minute measure of the probability of ‘Yes’, at least according to the aggregate judgments of the participants. Prediction market boosters never tire of pointing out that these ‘probabilities’ sometimes prove more accurate than professional pollsters’ forecasts. The best-known instance is Trump’s re-election in 2024, when in contrast to old-school polling, both Polymarket and Kalshi indicated a comfortable win. Both firms took this as proof of a general pattern. Beneath a list of current bets on whether there will be a coup in this or that country by a certain date, Polymarket answers the FAQ ‘Why use Polymarket for coup predictions?’ with the response: ‘It cuts through the noise. Unlike polls or punditry, Polymarket shows you real-time odds on coup predictions backed by financial conviction that are often faster and more accurate than experts or surveys. You get an unbiased view of what thousands of traders think will actually happen, often more accurate than polls.’

At present, though, even if that were a fair description of the prediction markets’ geopolitical foresight (it isn’t), coups, elections and the like account for only a small proportion of contracts. Most events are sports-related: final scores, in-game occurrences, celebrity sightings in the stadium and so on. The obvious questions – aren’t these contracts really bets, and aren’t prediction markets really just sports-betting businesses? – are the basis of several cases currently before US federal courts. The matter is certain to reach the Supreme Court; a great deal of money, and substantial political capital, will ride on its decision.

In the US, gambling is regulated at state level. This makes for a very uneven landscape. Nevada embraced the industry a long time ago, but Wisconsin has traditionally been wary of the social costs of gambling and has kept the industry on a comparatively short leash. Utah, where the divide between a very conservative church and an equally conservative state is more blurred than it is elsewhere, bans virtually all forms of gambling. Other states have their own interests, as do the many Indigenous American communities for which casinos on reservations are a crucial source of revenue and employment.

These statutory inconsistencies have created some complications for the gambling industry, which includes not just casinos and online gaming but sportsbooks such as FanDuel and DraftKings (the American equivalents of the businesses that now sponsor football clubs all over the UK and Europe). But in other respects it has proved an advantage: if people aren’t allowed to gamble where they are, then they will go out of their way to access your services. Gambling is a lucrative industry. Nevada has spent millions of dollars over the years lobbying against the legalisation of casinos in other states; Indigenous communities across the American West have opposed off-reservation gambling for similar reasons. And then there are the licensing fees and tax revenues that gambling operations contribute to state budgets.

Kalshi and Polymarket, however, say that they are not gambling businesses but markets for financial derivatives – specifically, ‘futures’ – which are federally regulated. Nearly all of the cases before the courts turn on the legitimacy of this claim. This is the reason you won’t find the word ‘bet’ on any prediction market website (only ‘trades’, ‘contracts’ and so on), and market participants are never called ‘bettors’ or ‘punters’ in event contracts but ‘traders’, ‘market makers’ or ‘dealers’. Even if the event predicted is the score at the top of the fifth inning in the Dodgers game or Erling Haaland bagging a hat-trick against Ipswich – around 70 per cent (Polymarket) or 80 per cent (Kalshi) of the trading volume on these platforms is in sports-related propositions – the prediction market operators insist that their business isn’t sports betting but trade in ‘binary’ (winner takes all) derivative financial instruments. Accordingly, US prediction markets are subject to the oversight of a long-standing federal regulator, the Commodity Futures Trading Commission (CFTC), and operate in what has become a highly accommodating regulatory environment.

Prediction markets, and many of their participants, have roots in cryptocurrency. Polymarket, which is headquartered in New York City, has been a crypto-based exchange since its launch in June 2020. Traders on Polymarket can only buy, sell and receive earnings in the form of a proprietary ‘token’ called Polymarket USD (pUSD), the value of which is backed one-to-one by USDC, a widely used ‘stablecoin’. (A stablecoin is a cryptocurrency pegged to a fiat currency – usually the US dollar.) To withdraw their money, traders must convert the pUSD in their portfolio to a stablecoin, then transfer the funds to a blockchain-based account called a ‘wallet’. Then, if they wish, they can sell the coins for euros or dollars or other currencies on a crypto exchange such as Coinbase.

In January 2022, the CFTC under Joe Biden charged Polymarket with operating an unregistered derivatives market in violation of the Commodity Exchange Act. Polymarket was given a $1.4 million fine and agreed to close non-compliant markets to Americans. It still has offices in New York, but is now legally based in Panama City, where the vast majority of its business operates essentially without regulation.

When Kalshi, which operates almost entirely in US dollars, opened in 2021, it did not make the same mistake. Its founders, Luana Lopes Lara and Tarek Mansour, were careful not to represent their business as a traditional gaming operation, not least because they didn’t want to spend their time wrangling the demands of fifty different state regulators. Instead, they seem to have told anyone who asked that Kalshi was a futures exchange, not a sportsbook – their goal, in Mansour’s words, was ‘to financialise everything and create a tradeable asset out of any difference in opinion’ – and that the company was registered with the CFTC. The CFTC appeared to be taking no notice, at least at first. It had scant experience of the sort of operation Kalshi was running, and didn’t have a framework of rules that would enable it to oversee what looked on the face of it like sports betting (underage sports betting, in fact, since the minimum age for gambling is 21 in many states, but 18 for derivatives trading). This was a regulatory no-man’s-land, a good place to build a business that encourages its clients to bet not just on baseball but also on what the unemployment rate will be in three months’ time.

Eventually, the CFTC caught up with what had been going on. In the run-up to the midterms in November 2022, Kalshi had hosted markets in ‘congressional control contracts’: bets on which party would win a majority in Congress. The CFTC saw this as a potential violation of the Commodity Exchange Act, which bans contracts involving activity ‘contrary to the public interest’. In September 2023, it released a report which described congressional control contracts as a significant misinformation risk: ‘the opaque and unregulated sources of price-forming information’ for the congressional control contracts ‘may increase the risk of manipulative activity relating to the trading and pricing’ of the contracts, while decreasing Kalshi’s and the commission’s ability to ‘detect such activity’. Even if the CFTC were to permit election markets, the report added, they would be treated as betting markets, subject to state gaming laws. (An organised election market – the Iowa Electronic Markets, operating out of the University of Iowa – has existed since 1988, but was exempted by the CFTC in 1992 because it existed ‘strictly for experimental and academic purposes’.)

Kalshi appealed against the CFTC’s ruling, arguing that election market contracts were not wagers, but financial instruments used to hedge political risk. Somehow, a DC district court agreed, giving the opinion that ‘Kalshi’s contracts do not involve unlawful activity or gaming. They involve elections, which are neither.’ This left the CFTC responsible for regulating activities it had recently deemed illegal and which were outside its competence.

But before the CFTC could come to terms with the regulatory challenges posed by prediction markets, Trump returned to the White House. Kalshi – along with Polymarket’s new US subsidiary and other smaller prediction markets – could not have been luckier. Users bet $4 billion on the outcome of the presidential election. Trump decided he loved prediction markets, mostly because they said he would win. Big players in his kleptocracy immediately got involved in the industry, both as major investors and as ‘advisers’. Donald Trump Jr is paid lavishly by, and has been given equity in, both Kalshi and Polymarket, and Peter Thiel raised $70 million in venture capital for Polymarket. The chair of the CFTC since December 2025, Michael Selig, is a libertarian nationalist who writes self-promoting op-eds for Breitbart; as a lawyer, he worked for many years at a firm that represents Polymarket.

Selig has repeatedly insisted that the industry is subject only to the CFTC’s oversight – ‘coddling’ would be another way of putting it – and has sued at least nine states for ‘regulatory overreach’. He has also continued to slash the staff of the commission, dropped extant investigations of crypto and prediction markets, and initiated virtually no new investigations. The ban on Americans’ access to Polymarket has effectively been lifted: 50 per cent of transactions are resolved on US crypto exchanges. Anyone questioning this approach, or the close ties between the industry and the Trump family, has been placed on leave, barred from the office or fired.

In May, the New York Times reported that the CFTC had been ‘steamrolled’ by the firms it is supposed to regulate, but that is to assume there is a meaningful separation between the regulator and the regulated. In fact, like many other institutions, Trump’s CFTC was not ‘captured’ by anybody: it was handed to them. Selig and his cronies are rich anti-regulators who have devoted themselves to defending the crypto industry and now the prediction markets, and to actively undermining any significant limits on their operation. The public interest has never mattered to them.

The first ‘futures’ were fixed-price contracts for the future delivery of an agreed amount of grain of a specific grade, say a hundred bushels of No. 2 wheat on 1 September. It is commonly thought they began as a way for farmers to hedge their risks by guaranteeing a sale at a particular price, but they were actually first devised in the mid-19th century as contracts between grain brokers at the Chicago Board of Trade and buyers on the Atlantic coast. At the time, transporting goods took so long that prices could change significantly during the journey, creating problems for both sellers and buyers. ‘To-arrive’ or ‘forward’ contracts, in combination with standardised grading, stabilised a historically volatile market: buyers knew what they were getting, and sellers knew what they were being paid.

These early futures were a way to make the world more predictable. Soon, however, they were converted into riskier financial instruments. All it took was the recognition that futures themselves could be traded. If, for example, I had a contract to take delivery of that No. 2 wheat on 1 September, but the price of wheat rose in New York in the meantime, I could sell my contract to another buyer for more than its face value and pocket the difference. The next step was easy: if neither party to the contract was interested in actually taking possession of the grain, the future could be treated as a wager and settled with an amount of cash based on the difference between the contracted price and the spot price on delivery day. Eventually futures became common in virtually all agricultural commodity and natural resource markets, and then in markets for almost anything, from fishmeal to scrap metal to stock indices. (There are a couple of exceptions: futures in onions have been illegal since 1958, and in box-office receipts since 2010.) They were an early form of ‘derivative’, a transferable property claim whose value ‘derives’ from an underlying asset.

For the rest of the 19th century, futures trading in the US was regulated by individual exchanges – the Chicago Butter and Egg Board, the New York Cotton Exchange – and a patchwork of legislation covering specific commodities, such as the Cotton Futures Act of 1914. In the early 1920s, the Grain Futures Act established a regulatory framework that required all exchanges to register as ‘designated contract markets’ (which is what Polymarket failed to do in 2020). The CFTC was set up in 1974. Like the Grain Futures Commission and the Commodity Exchange Authority before it, it is part of the US government’s agricultural bureaucracy and is overseen by the House and Senate Agricultural Committees, its putative role being to ‘serve our nation’s farmers and ranchers with the purpose of promoting resiliency in the US derivatives markets’.

Derivative markets have grown hugely over the last fifty years, partly as a result of innovations like ‘collateralised debt obligations’ – the straw that broke the back of the US subprime mortgage trade in 2007-8. As a result, the range of activities the CFTC regulates has increased too, especially since the passage in 2010 of the Dodd-Frank Act, which expanded the committee’s jurisdiction to include ‘swaps’ – customised contracts between two parties to exchange cash flows from different sources over a specified period of time. The large majority of swap trades take place among big institutional investors; along with short-term foreign exchange swaps, the most common are interest-rate swaps, where fixed-rate payments are exchanged for floating-rate to hedge against interest-rate risk. It is difficult to overstate the centrality of these derivatives to the operation of contemporary financial markets: swaps in the US have weekly volumes in the tens of trillions of dollars.

To protect its regulatory domain, the CFTC insists that prediction market contracts are swaps, which is a bit of a stretch. There are commodity swap markets – crude oil is the biggest – in which parties agree to accept or pay a fixed price for a commodity over a period of time (rather than a fixed price on some future delivery date, as they would with a future). Oil producers or large-scale consumers like utilities use swaps to hedge against the risk of oil price volatility. The only way to represent the trades in prediction markets as commodity swaps is if, first, you ignore the sports betting that comprises most of those trades (and which is hardly mentioned in the CFTC’s material on prediction markets); and, second, if you characterise non-sports trades in prediction markets as hedging or speculating on commodity prices in which, as the CFTC puts it, ‘the commodity is the outcome of the expected event.’ This conceptual contortion is the way the CFTC under Trump justifies its eager accommodation of prediction markets’ efforts to ‘financialise everything’.

It’s true that prediction markets were originally designed to enable betting on ‘political risk securities’. In the early 2000s, the Advanced Research Projects Agency (DARPA) of the US Department of Defence devised the Policy Analysis Market (PAM), on the model of the Iowa Electronic Markets. Participants would predict future geopolitical developments, such as the date on which the US would withdraw from this or that foreign country. The goal was to see whether the ‘market’ could provide more accurate predictions than other forecasting methods, but when news of the project emerged and critics accused the DoD of selling ‘terrorism futures’, the programme was cancelled. That didn’t stop some private sector actors getting into the business: in 2003, the Irish betting platform TradeSports.com opened a market in what became known as the ‘Saddam security’, a wager on the date by which the US would have removed Saddam Hussein from power. On 1 February, the ‘June Saddam’ was trading at about 80, meaning 80 per cent of the total bets that Saddam would fall before the end of June were on ‘Yes’. The Saddam security tracked the price of oil pretty closely, which some took as evidence of the ‘wisdom of crowds’, the idea being that the security was a single-number aggregation of market participants’ independent assessments, based on their many sources of information and individual judgments.

Prediction markets have not been widely adopted by the biggest players in political risk, such as the DoD. One of their key limitations as sources of intelligence is that they only work for anticipated events or threats, not for nasty surprises such as the 11 September attacks – the sort of thing you really want to see coming. Participants in prediction markets can’t predict things that haven’t occurred to them as possibilities. More significant, it isn’t at all clear that the results in prediction markets like the Iowa Electronic Markets or PAM reflect the ‘wisdom of crowds’ theory of information aggregation. That would be to assume, among other things, that there is no herd behaviour, and that predictions don’t influence one another. Maybe this was the case for the select group of policy insiders and experts who participated in those markets. But even then mutual influence can’t be ruled out, and in publicly accessible prediction markets it is impossible to avoid. That the Saddam security tracked the price of oil cannot be taken as proof that it was ‘accurate’: all it shows is that both prices were formed on the basis of similar information and sentiment. And just because the price of the Saddam security was close to 100 immediately before he was ousted doesn’t mean the market was forecasting the ‘truth’. It was just telling us what pretty much everyone already knew.

A substantial proportion of trading activity on prediction markets involves long-shot get-lucky bets, herding and intentional market manipulation, legal and illegal. During the World Cup final in July, there were thirty thousand unique combo bets on Kalshi alone (combos are the equivalent of higher-risk ‘parlays’ on sportsbooks, wagers on the outcome of two or more events; you only win if you get everything right), 3 per cent of which resolved to ‘Yes’. The price on ‘When will Taylor Swift announce she is pregnant?’ swings wildly with every wink or online rumour. For all but a few exceptional markets, prices during trading don’t have any necessary relationship to truth or accuracy. Are election forecasts one of those exceptional markets? If the price and the electoral outcome turn out to be very strongly correlated, that is something to think about. But it doesn’t prove anything about prediction markets and elections in general (which are supposed to be hard to manipulate because in principle the decision is distributed across many people). In fact, prediction markets have got lots of elections wrong, which would be less of a concern if the ways in which they were wrong weren’t so damaging.

One of the more recent examples is the Republican reaction to the failed Los Angeles mayoral campaign of Spencer Pratt, a right-wing reality TV star, in June. A report on Bloomberg describes what happened when it became clear that Pratt’s ‘probability’ of victory on Kalshi and Polymarket had been much higher than his share of the vote turned out to be. Social media lit up with accusations of fraud, many of them from MAGA influencers who are paid by the prediction markets to spread the gospel. Trump himself posted on Truth Social that Pratt’s loss was ‘not possible’ given ‘the big lead he had’. Then he set about California: ‘3rd World Nation. Rigged Elections!’ In fact, Pratt never had a big lead, or any lead at all. But the ‘predictions’ – bets, after all, many of which might have been made by a very small number of people – change the narrative. This is what happened earlier this year in the primaries for November’s California gubernatorial race, when there was a flurry of publicity for a peripheral candidate called Matt Mahan after one person staked $44,000 on his victory on Polymarket. Since hardly anyone else was wagering – the market was ‘thin’ – this single anonymous bettor temporarily inflated the ‘probability’ of Mahan’s victory to 96 per cent. In the event, he won under 4 per cent of the vote, just as the conventional polls suggested.

The Mahan discrepancy is exactly the sort of thing Biden’s CFTC was trying to avert by taking a firm line on political prediction markets. Instead, under Trump, the CFTC is creating the conditions for the proliferation of manipulation and misinformation. Given the Republican Party’s increasing antipathy to democracy, it seems plausible that this is deliberate. At the same time it has allowed prediction markets to become deeply intertwined with news media: Dow Jones, Google Finance, X, Yahoo and Substack now show Polymarket predictions onscreen during related news stories; Kalshi is working with CNBC, CNN and Fox News. The prediction market numbers are themselves sometimes the story. Both firms are spending millions on social media, hiring influencers and rewarding ‘random’ online mentions. The target audience is mostly, though not only, young people, and not just on the right. During the build-up to the New York Knicks’s NBA championship win in June, M.D. Hossain, a 23-year-old fan, became instantly famous when a video of him shouting ‘My mayor’s Muslim. My bagel’s Jewish. My Christian’s Dior. Knicks in four!’ went viral. The microphone he’s holding is Kalshi-branded.

Concerns about misinformation, and about the conflation of predictions and real-world probabilities, are not the only reasons the prediction market industry should be strictly regulated. Insider trading like Santos’s is rife. If you know where reality is heading, you can win a lot of money from people who don’t. If you know that a business is going to announce quarterly losses tomorrow, you can bet on a drop in its share price. If you know when the US is going to enter Venezuela and kidnap Nicolás Maduro, you can bet on it and ‘earn’ $400,000, just as a member of the US Special Forces did in January. Or, if you don’t have inside information about things in the real world, maybe you are in a position to change those things so they fall in line with your bets. This has always been a problem in sport, and fixes are harder than ever to detect now that you can bet on almost any aspect of a sporting event; it took nearly two years for Lucas Paquetá of West Ham United to be cleared of charges that he had intentionally tried to get booked in Premier League matches in order to win bets for associates.

Real-world manipulation isn’t confined to sports. On 15 April, hundreds of thousands of dollars were traded on Polymarket on the question of whether the temperature in Paris would reach 19°C. The authorities’ suspicions were drawn that evening (not a time of day when temperatures normally rise) when the temperature measured by an official sensor at Charles de Gaulle Airport suddenly jumped by five degrees to 22°C. No culprit has been identified, but there was speculation on French meteorology forums that a battery-powered hairdryer may have been involved.

Mischief is one thing, but the incident makes one consider more damaging behaviour. There has been betting on the extent of wildfires on Polymarket and at wyldfyre.io, a startup focused solely on fires in California, which attracted so much negative attention that in August several senators demanded the CFTC shut it down. It isn’t hard to start a wildfire, or to help an existing fire grow. You can’t bet on assassinations or other violent crimes on Polymarket or Kalshi, though Polymarket’s offshore status and crypto-based anonymity mean that it monopolises trades on such events as Maduro’s capture or the ‘removal’ of the Iranian leadership. But it is easy to trade in predictions related to pandemics, say, or the screwworm outbreak currently killing cattle across the US. None of this is immune to deliberate human influence.

One of the fastest-growing forms of market manipulation on these platforms is closer to old-school financial boondogglery. Traders in asset-price contracts bet on short-term changes in the price of a cryptocurrency, usually Bitcoin. The wager is simple: at some predetermined future moment, will the price be higher or lower than it is now? The time horizons are very short – five minutes, fifteen minutes, an hour, four hours – and if the price goes up or down before settlement time, it doesn’t matter: the price at the end is the only thing that counts. The contracts open and close every five minutes, all day every day. What’s more, the process is easy for traders to automate. These ‘up/down’ crypto markets became hugely popular almost immediately after Polymarket launched them in September 2025: by July this year, the monthly trading volume had reached $4 billion.

Normally, the price of Bitcoin doesn’t change very much over such short periods, maybe a few basis points (one-hundredths of a per cent). As a result, even though the market in Bitcoin is enormous, it isn’t all that hard for a well-resourced trader to influence the price in the last seconds before closing. A team led by David Dai, a PhD student at Stanford, developed a test for a ‘manipulation footprint’ in the five-minute up/down trade on Polymarket, and in June found it in around a tenth of a sample of more than sixteen thousand trades. ‘Manipulators capture $8.2 million in aggregate across manipulated cycles,’ the researchers found, ‘$5096 per manipulated cycle.’ The ‘retail traders’ who comprise the vast majority of market participants ‘fund nearly all of it: they lose $7.6 million, $4715 per manipulated cycle.’ It is a general feature of prediction markets, even in the absence of market manipulation, that there are few winners and many losers.

The lopsided distribution of earnings on prediction markets is often held up as proof that the markets reward those with the most accurate foresight. You want to win? You just need to be right. ‘I knew we were entering an era where ways to find truth would matter more than ever, and Polymarket could play a critical role in that,’ Shayne Coplan, the company’s founder, has said. ‘After all, nothing is more valuable than the truth.’ The truth is so valuable, in fact, that many prediction market operators have no problem with insider trading. Robin Hanson, an economist at George Mason University who is sometimes called the ‘father of prediction markets’, has long argued that insiders should be welcomed. If the priority is discovering what’s true, he says, ‘then you definitely want to allow insiders to trade, even if that discourages other people from betting, because that makes the prices more accurate.’ The proponents of prediction markets talk a lot about truth. ‘These markets will run the world,’ one of their billionaire backers boasts. ‘People tend to say what they want, but in these markets, they will bet the way they think the outcome will occur, not what they want the outcome to be. It takes the emotion out of these questions.’

But truth and true beliefs are not the same thing. They may not have anything to do with each other. Prediction markets can only be, at best, true-belief machines. The ‘truth’, to the extent that it is not known in advance by insiders or created by a trader to win their bet, is revealed at the close, and not by the markets but by the real world. Prediction markets don’t produce truth. They are high-stakes wait-and-see machines, that’s all.

For the World Cup combo bettors and other punters, participating in prediction markets is mostly about the thrill of gambling. But not for those concerned with truth-discovery. According to Selig or Coplan or Mansour, prediction markets are a ‘global truth machine’, which enable some people to know what some piece of the future looks like. That knowledge can make you a lot of money, even if the piece of the future at stake is minuscule, or meaningless. As Matt Levine, who along with his Bloomberg colleague Christopher Beam is one of the sharpest analysts of prediction markets, put it: ‘One “accomplishment” of modern prediction markets is that they can, in theory, allow for an arbitrarily large amount of betting on arbitrarily small quantities of reality.’

Sometimes it can be hard to tell what the truth is. Which measure of inflation is the one that’s ‘true’? If you have a contract on the price of a widely traded financial asset like Bitcoin, which market has the ‘true’ price at settlement time? In the case of the five-minute up/down Bitcoin contracts, Polymarket solves the problem by way of what is (not at all jokingly) called an ‘oracle’. The term, which comes from the crypto world, describes a service that bridges between the blockchain and the non-crypto or ‘off-chain’ world. Blockchains are secure, closed computing environments, so when real-world information – like an asset price – is required for an on-chain transaction, the oracle is the service that provides it (in the broader world of decentralised finance, ‘oracle hacks’ are a big problem). Polymarket uses an oracle called Chainlink, which collects data from many different crypto exchanges and calculates a single price based on trading volumes.

When the ‘truth’ is less computable, things can get messy. When an American special forces unit entered Iran in April to rescue a downed pilot, did that mean ‘ground forces had entered Iran’? In the many ‘mention markets’ on Polymarket and Kalshi, you can bet on whether a public figure will say a particular word or phrase at an event. If the word is ‘border’, does ‘bordered’ mean the market resolves to ‘Yes’? What about homophones like ‘way’ and ‘weigh’? Sometimes the platform, which has the final say, will resolve the market on the basis of a contentious technicality, which traders call a ‘rulescuck’.

This can provoke a lot of bad feeling, and there is the potential for countless court cases, so the platforms have developed internal resolution processes. Kalshi has a team of expert word-splitters. At Polymarket, disputes are resolved by market participants who own a special token called UMA, or ‘universal market access’. Anyone can buy UMA; the more UMA you own, the more votes you get when the ‘truth’ is being decided. The entirely predictable problem is that 95 per cent of UMA is held by a small group of ‘whales’ – a term for participants who trade in large enough volumes to move a market on their own – each of whom owns at least 1 per cent of the total.

The dubious achievement of prediction markets is to have found a new way to exacerbate and exploit the simultaneously frantic and passive experience of the world that is a defining aspect of our age. Like any kind of gambling, they are enervating. We put our money down and then (unless we are using a hairdryer on an airport thermometer) fidget anxiously till it’s over. The world happens to us. This passivity is characteristic of the politics of those who think we are stuck with the world we have. The best we can do is make money while events wash over us: ‘Don’t blame me. I didn’t start the war, I just bet on it.’ Such people may think they are ‘courageous’ or ‘gutsy’ because they take ‘risks’ (this is the way people in the financial sector have always talked), but the whole charade is a profitable supplication to, even a cowering before, the power of ‘the markets’. To put the case more gently, the appeal of ‘risk-taking’ just now, especially but not only to young men, may be an indicator of how little control many people feel they have over their own slice of the world, and of the feeling that they can’t shape their lives in a meaningful way. Marking time through a series of ‘settlements’ and ‘closes’, hoping for a windfall, may look like the best way to buffer oneself against an uncertain future.

In the last years before his death in 1662, Blaise Pascal, one of the originators of the mathematics of probability, wrote a series of reflections published eight years later as Pensées. Number 418, ‘the wager’, contains his distinctive defence of Christian faith. He admitted there was ‘no certainty, apart from faith’ in God’s existence, and acknowledged the logic of doubt, the lack of evidence, the seeming impossibility of miracles. But still, he argued, faith was a bet worth making: if the expected value of a wager is a function of its likelihood and the benefits that winning will bring – probability multiplied by expected utility, in modern terms – then the better the potential outcome, the slimmer the chance required to justify the risk. With faith, then, there is only one logical position. If you choose to believe, and God does not exist, you have lost nothing; but if he does, you have everything to gain: ‘That leaves no choice; wherever there is infinity, and where there are not infinite chances of losing against that of winning, there is no room for hesitation, you must give everything.’

Three and a half centuries later, there is an open trade on Polymarket on the question of whether Jesus Christ will return to Earth before the end of 2026. At the time of writing, you can buy ‘Yes’ trades for 1.4¢ each. The proposition isn’t quite the same as whether or not God exists, but the payoff is still huge: if Christ does indeed return, you will be rewarded not only with the Second Coming, but a 7000 per cent return on your investment. The prediction markets clearly do not see things in the same way as Pascal. For every bet ‘Yes’, there are 49 sceptical ‘No’s. When he heard about the trade, the Wikipedia founder Larry Sanger, a born-again Christian, wrote: ‘This is free money. Bet that Jesus won’t come back. If he doesn’t, you win. If he does, will your money really matter?’ There’s no place in prediction markets for faith, it seems, even among the faithful.

Polymarket’s ‘Will Jesus Christ return before 2027?’ contract currently has a trading volume of $66,000,000. The fine print says: ‘This market will resolve to “Yes” if the Second Coming of Jesus Christ occurs by 31 December 2026, 11.59 p.m. Eastern Standard Time. Otherwise, this market will resolve to “No”. The resolution source for this market will be a consensus of credible sources.’ But what if Christ comes and the UMA whales don’t believe it’s him? For Pascal’s sake, let’s at least hope he’s planning on New Year’s Eve in New York, not Los Angeles. Imagine missing the Messiah because of a technicality.

Send Letters To:

The Editor London Review of Books, 28 Little Russell Street London, WC1A 2HNletters@lrb.co.uk Please include name, address, and a telephone number.