Back of the net
For bookmakers, the World Cup group stage has delivered plenty of goals, but rather less of the usual spoils that come from customers getting their football predictions wrong.
According to Bank of America estimates cited by the Financial Times, DraftKings may have lost as much as $50m during the opening phase of the tournament, as popular teams won and the game’s biggest names started treating the net like it had personally offended them.
Lionel Messi, Erling Haaland and Kylian Mbappé all scored twice last week, creating a particularly expensive problem for operators exposed to multi-leg bets.
Parlays, or accumulators to those on this side of the Atlantic, are generally a sportsbook’s friend. They offer the promise of a chunky payout but require several outcomes to land, which is why they are typically a strong-margin product.
That maths became less reassuring when bettors piled into bets involving all three stars to score. Those who backed the trio to score at least twice each (an outcome priced at roughly a 1% probability) enjoyed an especially healthy return.
And it was not just the forwards causing trouble. Bettors’ fondness for their home teams has made the form of England and the US particularly consequential.
Bank of America analyst Julie Hoover said the strong performance of the US had become “the biggest liability” for most American sportsbooks. Flutter, meanwhile, said its UK customers won £4.1m from England’s 4-2 victory over Croatia, although England’s goalless draw with Ghana subsequently “more than balanced the books”.
The World Cup may therefore not be an immediate revenue bonanza, even after operators spent heavily on promotions, in-play products and customer acquisition.
But executives appear prepared to tolerate a few painful results if the tournament brings in customers who can later be cross-sold into higher-margin products.
Flutter chief executive Peter Jackson recalled losing heavily on Argentina’s 2022 final win over France, but said he still viewed the competition as a success because it had provided “great entertainment”.
For now, the entertainment has been excellent. The trading results, perhaps less so.
Build or buy?
Elsewhere, NPR reported that Mark Zuckerberg considered buying Kalshi before deciding that Meta could have a crack at prediction markets itself.
Because when a fast-growing corner of the internet catches fire, the company’s first instinct is apparently to ask whether it can acquire it or, failing that, build something adjacent.
The report says Zuckerberg met Kalshi chief executive Tarek Mansour last year as the prediction-market operator’s popularity surged. The discussions did not progress, although accounts differ on why: one version has Mansour unwilling to sell, while another suggests Meta found Kalshi’s legal and ethical baggage too complicated.
Either way, Meta still wants in. Internal documents reviewed by NPR show the company is developing a standalone prediction-market app called Arena, where users will make forecasts on news events and online trends.
The key difference is that Arena will use play money rather than real cash, while Meta’s AI systems will generate questions and determine whether users have won or lost.
That may make Arena a slightly more family-friendly version of the current prediction-market boom, or, as Columbia law professor Tim Wu put it, a “casino app with fake money” that may not offer much of a thrill.
Wu also suggested it fitted a familiar pattern for Meta, which has previously retreated from projects including Libra and the metaverse while retaining the financial firepower to chase the next big thing.
The prize is obvious, with trading volumes and valuations alike ballooning for market leaders Kalshi and Polymarket over the past year.
Meta has already partnered with Kalshi to integrate markets into Threads, but the reported takeover interest is likely to revive arguments around its broader “buy or bury” reputation. The FTC has accused the company of acquiring emerging rivals or cloning their services, though a judge last year ruled that Meta’s purchases of Instagram and WhatsApp did not breach competition law.
Arena may not be preparing to take real-money action, but it is another sign that prediction markets have become far too tempting for Big Tech to ignore.
Easy listening
Finally, it is back to the Financial Times which reported that Spotify’s US chart briefly had an unlikely number one this week: “Earrings”, a 2024 indie-pop track by Malcolm Todd.
Then it had an investigation, more than 500,000 streams removed and a prediction-market payout that had already gone out the door.
The report said US streams of “Earrings” jumped almost 70% between Sunday and Monday, propelling the song to the top of Spotify’s daily US chart.
The timing caught attention because traders on Kalshi had spent the preceding week pricing Todd’s chances of landing a number-one Spotify USA song before the end of June at just 2.5%.
Spotify later concluded that a large number of the streams had been initiated by bots, which can repeatedly play tracks to manufacture popularity.
Once those plays were stripped out, “Earrings” would have fallen to fourth place for Monday. By that point, however, Kalshi had already settled the market, handing traders who backed the long-shot outcome returns of roughly 20 times their initial stakes.
There is no suggestion that Todd or his team were involved in the apparent attempt to inflate the track’s audience. The suspicious betting was first highlighted by Kalshi trader Caleb Davies, who posted details on X.
The episode is a neat, if awkward, illustration of a growing concern around prediction markets: where a market can be influenced, someone may have an incentive to influence it. Kalshi and Polymarket have already faced scrutiny over potential insider trading and manipulation, including cases involving trades connected to sensitive political and public-information events.
Stream manipulation itself is hardly new. Fraudsters have long used bots to boost obscure tracks and generate extra royalty payments. But the arrival of a market that pays out on chart outcomes creates another reason to manufacture a spike, and a potentially highly lucrative one.
Spotify said it had “best-in-class detection and mitigation practices” and does not pay royalties on manipulated streams.
Kalshi said it was investigating, while Spotify’s legal team asked the platform to remove its logo from its app and website. Kalshi has since added a disclaimer stating that its products are not endorsed by Spotify.
For one day, at least, the charts became less about who was listening, and more about who had seen the market coming.
