Polymarket Settles Bet Against its Own Rules
Unknown crypto token-holders that vote to decide Polymarket’s wagers aren’t bound by the website’s fine print
This post is about the recent failure of a prediction market for Venezuela’s presidential election on Polymarket, the blockchain-based betting website. It shows how the decentralized token-based voting system to adjudicate bets decided against Polymarket’s own stated rules, and outlines broad implications for crypto-based prediction markets, particularly the market for the U.S. presidential election on November 4th, currently worth over $500 million.
Polymarket’s Venezuela Fiasco
Suppose you are Venezuelan and wanted to hedge against Nicolas Maduro winning Venezuela’s presidential election. On Polymarket, there was a prediction market for just that. You could put 80 cents on Maduro to make 100 if and when he won the election or Venezuela’s institutions falsely awarded him victory, for a return of 25%.
The rules1 for the wager had a single sentence on how the bet got decided: “The primary source of resolution is official information from Venezuela, however a credible consensus of credible reporting will also suffice.”
The first clause – “The primary source of resolution is official information from Venezuela” – was key. In Venezuela, official information means statements from the national electoral authority, which is politically captured and loyal to Maduro and Chavismo. The second clause – “however, a credible consensus of credible reporting will also suffice” – was not relevant for the hedge. If the primary source of resolution (official information) and the secondary source (a consensus of credible media) disagreed, only one of two things could reasonably happen:
Either the primary resolution source overruled the secondary source
Or, the bet failed to settle and Polymarket returned everyone’s money
Per the rules, the prediction market was not about votes cast but about which candidate the electoral authority would declare the winner. Even though Maduro was down by a staggering 35 to 60 percentage points in the polls, he was still the 80:20 favorite to “win” this market, which priced in a high likelihood of fraud.
On election night, Venezuelan authorities declared Maduro the winner, as predicted. The official results were inconsistent with the opposition’s internal vote count, which had them ahead by over 20% based on vote tallies printed at polling stations across the country. Plus the vote shares, 51.20000% for Maduro and 42.20000% for Gonzalez, have four zeros from the second to the fifth decimal – an extremely unlikely result2. It was almost certainly fabricated, as the NYT and others have since reported.
Yet, despite the official announcement, the bet didn’t settle. Polymarket’s forums exploded with messages where Venezuelans learned that it’s not an expert committee at Polymarket’s New York headquarters or an independent third party that resolves the bet based on public information and the rules – that would be too inconsistent with web3 governance principles, too centralized.
Instead, it’s some “decentralized” blockchain-based mechanism that settles bets. It’s called UMA. Per their website, they’re a “decentralized truth machine” that can “record any verifiable truth or data onto a blockchain.”
The way it works is that anyone with an internet connection can buy UMA tokens and use them to vote to settle bets resolved by the associated protocol. UMA uses complex crypto incentives including staking, slashing, lockups, “up to 30.1% yields”, voting rewards etc in an attempt to incentivize truthful votes. It’s market capitalization is ~$160 million.
Five days later, Venezuela’s electoral authority made a second announcement: with 97% of votes counted, they again declared Maduro the winner 51.9500% to 43.1800%3. The numbers again didn’t match the opposition’s public and verifiable vote tallies and there were the same extremely unlikely zeros from the third decimal onwards. Despite this second official announcement, the bet again failed to settle.
As the days passed, the flaws of UMA’s resolution system dawned on bettors. UMA token-holders are unknown crypto-savvy persons that could be anywhere in the world. They aren’t accountable to Polymarket or its customers or liable in any legal system. They can place bets on the very same prediction markets that they vote to resolve and can be influenced and lobbied. In contentious cases, they can coordinate to settle a bet for a particular outcome – and also place bets to profit on that outcome.
As the evidence that Venezuela’s opposition got more votes became overwhelming, a vigorous lobbying effort ensued. People betting that opposition candidate Edmundo Gonzalez would win lobbied for Maduro to be declared loser on Polymarket even though the primary resolution source, Venezuela’s electoral authority, declared him the winner. Polymarket and UMA’s forums and Discord channels lit up with the same basic appeal: ignore the primary resolution source (official information from Venezuela) and defer to the secondary resolution mechanism (a consensus of credible reporting).
In the end, UMA token-holders were persuaded and coordinated to override Polymarket’s rules and resolve that Edmundo Gonzalez won Venezuela’s presidential election, even though Maduro was officially declared winner by Polymarket’s primary resolution source and will remain in power for another six year term.
Like Venezuela’s captured institutions, UMA’s decision simply ignored the rules, and Polymarket users who correctly predicted the future and hedged against Maduro winning or falsely claiming victory had their money stolen.
Could This Happen in America?
The U.S. presidential election will be held on November 4th and there’s already half a billion dollars in the pot on Polymarket, a number that could grow 5x in size if the betting patterns from 2020 are a guide. It isn’t hard to imagine how Polymarket might declare the loser the winner considering the UMA system and the market’s rules:
“The resolution source for this market is the Associated Press, Fox News, and NBC. This market will resolve once all three sources call the race for the same candidate. If all three sources haven’t called the race for the same candidate by the inauguration date (January 20, 2025) this market will resolve based on who is inaugurated.”
Suppose Trump loses the deciding state in the electoral college by a few hundred votes. During a drawn-out recount, Fox News and Twitter are flooded with fake news about mail-in voter fraud and illegal immigrants voting, along with a handful of plausible allegations worth investigating. When the recount finally concludes and narrowly awards Harris victory, The Associated Press and NBC call the election for the Democrats, but Fox News drags its feet citing Trump's allegations and his pending cases at the supreme court.
The bet can’t settle without Fox calling the election, but Trump’s shares drop to 1-2 cents anyways, pricing in a bet resolution in favor of Harris no later than January 20. In December, the Supreme Court rules 5-4 in favor of the Democrats winning the presidency, but the decision is technically reversible pending further cases and evidence that Trump claims to have. Republicans are radicalized by the Supreme Court’s dissenting opinion and go up in arms when Fox calls the election for Harris with half of its prime-time anchors resigning in protest.
Immediately, someone posts the $100,000 bond to call for a vote on the UMA protocol to settle the bet in favor of Harris but it fails. UMA token-holders want to wait for the Supreme Court to hear Trump’s additional cases, so the bet is thrown into limbo. Meanwhile, more (false) evidence of voter fraud and illegal immigrants voting surfaces on social media, bolstering support for Trump.
Weeks before inauguration, Trump calls on his supporters to seize Congress again. Many rioters are armed, and police officers are forced to open fire. People die, and as gory images go viral on Twitter, someone pays a new $100,000 bond to call for another UMA vote, this time for Trump to be declared winner. In the heat of the moment, radicalized UMA token-holders talk themselves into supporting the vote. (They and also buy Trump shares for cents on the dollar in case the vote succeeds).
Before 24 hours are up, there are enough conflicts of interest and enough hyper-polarized UMA token holders for the vote to pass and billions of dollars are awarded to the wrong side, with traders that correctly bet on Harris losing everything, and people that bet for Trump taking the pot. (Conflicted UMA token holders also make a 50-100x profit betting on the outcome they voted for).
Users sue Polymarket, but the automatic payouts can’t be reversed on the polygon blockchain. The government takes too long to freeze the relevant stablecoins and by the time they manage to, most users already traded their earnings into uncensorable ethereum.
And so the multi-billion dollar bet is stolen, irreversibly.
Now… I’m not saying that this is likely. I’m just saying that the precedents are there. Trump has falsely claimed election fraud before and has attempted a violent, radicalizing self-coup to stay in power. And Polymarket bets have been decided against their stated rules on the politics and conflicts of interest of unknown, unaccountable UMA token-holders, most recently in Venezuela.
It could happen for America’s presidential election too.
Note: For a wonderful, insightful overview of the economics of prediction markets, see “Why prediction markets aren’t popular” by Nick Whitaker & J. Zachary Mazlish for Works in Progress.
“The 2024 Venezuela presidential election is scheduled to take place on July 28, 2024.
This market will resolve to "Yes" if Nicolás Maduro wins. Otherwise, this market will resolve to "No."
This market includes any potential second round. If the result of this election isn't known by December 31, 2024, 11:59 PM ET, the market will resolve to "No."
In the case of a two-round election, if this candidate is eliminated before the second round this market may immediately resolve to "No".
The primary resolution source for this market will be official information from Venezuela, however a consensus of credible reporting will also suffice.”
These vote shares are based on Maduro’s reported 5,150,92 votes and Edmundo Gonzalez’s 4,445,978 votes
Based on updated reported vote totals of 6,408,844 for Maduro and 5,326,104 for Gonzalez