Prediction markets are unreliable truth machines: lawsuit

One of the things we discuss in the American Banker newsroom is how much coverage we should dedicate to online gambling – I'm sorry, prediction markets. Kalshi and Polymarket have marketed themselves into near-ubiquity, but they're not banks. On the other hand, both are marketing themselves as more akin to derivatives markets than, like, online gambling. The Commodities Futures Trading Commission, which has staked out the industry as within its remit, agrees. So there is a degree to which you (and therefore, we) need to be up on this burgeoning market.

Which brings me to a lawsuit that was filed against Polymarket by some users alleging they were wronged by the company. I think it's interesting because it offers valuable insights into how Polymarket's online gambl…prediction market works and how "the truth" can be very open to interpretation, and manipulation.

A handful of Polymarket users had bet on whether the tech/bitcoin company Strategy would sell any of its bitcoin holdings by a certain date, in this specific case, May 31. Strategy, if you don't recall, is a company that used to be called MicroStrategy and used to mainly sell software. It was being squeezed by larger competitors, so its founder Michael Saylor came up with a novel strategy (no pun intended) to save the company: Buy bitcoin. A lot of bitcoin. He started very publicly buying as much bitcoin as the company could afford. He issued debt to get cash to buy more bitcoin. He issued equity to get cash to buy more bitcoin. He started selling securities products to the public based on his bitcoin holdings, with the goal of, you guessed it, getting more cash to buy more bitcoin.

The entire Rube Goldberg machine was predicated on one core belief: that Strategy would never, ever, sell even a fraction of its bitcoin holdings. Well, and that bitcoin's price would never go down, but that's a different story. In crypto terms, Saylor had diamond hands and laser eyes. And he made a big, big deal out of this never-sell principle. But, as will happen in markets, some folks didn't believe him. So, these people went to Polymarket and bet on this question: "Strategy sells any bitcoin by May 31?"

This spring, those doubters were proven correct. In a June 1-dated SEC filing, Strategy revealed that it had, in fact, sold some bitcoin. Not a lot, mind you, but enough. And the sales happened between May 26 and May 31, according to the filing. The language in the Polymarket contract seems clear: if Strategy sells any bitcoin by 11:59 p.m. on May 31, the contract resolves to "yes." The people who'd bet on Strategy selling were ready to collect.

Then Polymarket updated the contract with "additional context."

Because Strategy did sell bitcoin before May 31, but did not announce that publicly until June 1, the contract on Polymarket became contested, and this shines a light on how Polymarket determines "the truth." The Truth in this case is apparently not decided by an SEC filing, but is determined using a decentralized platform called the Uma optimistic oracle. Anybody can participate in this process, taking a side within the oracle's platform and helping to sort out "the truth." People post a bond to be involved, submit evidence, and the winning side collects.

"Disputes are rare in practice because the incentives are always to be honest," Risk Labs' website claims. That is why it's an "optimistic" oracle. The Strategy-contract lawsuit is claiming, essentially, that the optimistic oracle was manipulated, that a bunch of people took the other side of the bet and brute-forced "the truth" into being "no," that the oracle was wrangled into saying that Strategy did not sell any bitcoin before May 31 although the SEC filing showed that it did sell bitcoin before that date, and that Polymarket is ultimately responsible for this. Other folks online have been complaining that what happened here is not a one-off, that the Polymarket truth machine is in fact broken.

Polymarket did not reply to a request for comment.

I bring it up here because I think it illustrates the fact that these platforms are less mature, and therefore less reliable, than might be assumed. The truth machines that power these platforms are not all-knowing oracles but fallible wizards operating behind a curtain. The more billions that get put on these platforms, the more people their ad campaigns attract, the bigger this problem becomes.