top of page
Search
snitzoid

How the Trump Whale and Prediction Markets Beat the Pollsters in 2024

As I posted for several weeks before the election, the prediction markets appeared to be more accurate than polling.


What cinched it for me was watching Musk and Bezos (wouldn't allow the far left Washington Post to endorse Harris) each bet the ranch on Trump. The world's two richest guys wouldn't do that if there was any chance Voldemort might lose.


How the Trump Whale and Prediction Markets Beat the Pollsters in 2024

The success of Polymarket and other betting platforms in calling the election will bring an end to the era of political forecasting as we know it.


The Trump ‘whale,’ a trader known only as Théo, proved more accurate in his forecast than pollsters like Nate Silver. John Cuneo

By Niall Ferguson and Manny Rincon-Cruz, WSJ

Nov. 15, 2024 11:27 am ET


The 2024 election was a resounding victory not only for Donald J. Trump but also for prediction markets like the crypto-based Polymarket, which allow users to trade contracts that pay out based on the outcome of future events.


By the morning of the election, Polymarket showed $1.8 billion in trading volume on who would win the presidency (Trump at 62%) and an additional half billion on who would win the popular vote (Harris at 73%). The biggest bet on a Trump victory was placed by an enigmatic “whale” known only as Théo.


Trump’s victory was even more decisive than the prediction markets foresaw. Even on Polymarket, few shared Théo’s conviction that Trump would win the popular vote. But the prediction markets were still a lot closer than most opinion polls and political pundits, nearly all of which clustered around a neck-and-neck result.


The Wall Street Journal reported on Election Day that “Prediction Markets Point to Likely Trump Victory,” giving the former president a chance of success between 57% and 62%. But most polls showed the election as headed for a tie. Renowned election forecaster Nate Silver wrote on election morning: “We ran 80,000 simulations tonight. Harris won in 40,012,” thereby giving the sitting vice president a 50.015% chance of winning the election.


Prediction markets, like all markets, mobilize knowledge that otherwise “never exists in concentrated or integrated form, but solely as the dispersed bits of incomplete and frequently contradictory knowledge which all the separate individuals possess,” as Friedrich Hayek argued in 1945.


The 2024 election may go down in history as the last election when we paid more attention to self-anointed election experts than to prediction markets.


The news that the FBI raided the home of Polymarket CEO Shayne Coplan early Wednesday and seized his phone and other electronic devices has added another dramatic turn to the story. The basis of the investigation remains unclear. Calling an election result certainly isn’t an offense.


A “crypto-powered gambling website”


Incumbent pollsters and pundits spent the months before Nov. 5 heaping scorn on prediction markets. The better Trump’s odds became on Polymarket and Kalshi, the more negative were the reactions. The New York Times dismissed Polymarket as a “crypto-powered gambling website” and Trump’s lead as “an illusion.” The Times quoted Barnard Professor Rajiv Sethi’s suggestion that the platform was vulnerable to manipulation because traders would “take the losses in order to change public perceptions” and “possibly have an effect on things like donations and morale and volunteer support and turnout.”


Similarly, one “BlueAnon” account with over 500,000 followers alleged that Peter Thiel’s investment in Polymarket was the reason Trump led on the platform, as Thiel was “JD Vance’s mentor.” Crypto commentator Adam Cochran told The Wall Street Journal that “if Trump loses, his favorable odds in the betting markets could bolster arguments that the election was stolen from him.”


A different critique was that prediction market participants were not representative of the general population. Scott Kominers of Harvard Business School wrote that, although polls had failed to predict Trump’s victory in 2016, prediction markets didn’t do much better because “even the best-functioning markets don’t do a good job of pricing when key players aren’t represented.”


On that basis, the ban on U.S. citizens using Polymarket, which dates to a 2022 settlement with the Commodity Futures Trading Commission (CFTC), should have rendered it useless—after all, it was precisely U.S. citizens who got to vote in the election. But Americans don’t have a monopoly on political knowledge and judgment.


Trying to explain away Trump’s surge on Polymarket, Nate Silver offered a laundry list of explanations: “bored and angsty” traders, crypto bros who had “become more Trumpy lately,” retail investors letting “their imaginations run wild,” Musk’s tweets and a lack of “sharp money” in the markets.


What makes all of this so unconvincing is the chronic unreliability of traditional pundits’ preferred source of insight into election outcomes: polls.


A science of prognostication?


The opinion poll was born in 1936, when George Gallup and Elmo Roper predicted Franklin D. Roosevelt’s victory on the basis of sample surveys—that is, responses from groups of people deemed representative of the larger population. Since then, polls have saturated election coverage. The sheer quantity of polls in the recent past has given the illusion that something like a science of political prognostication is evolving. But quantity of data is not necessarily correlated with quality.


Silver got his big break in 2008 after correctly calling 49 of 50 states and Barack Obama’s victory. His core innovation was to use other polls as inputs into a model of his own, weighing and adjusting each poll to produce one aggregate success probability for each candidate.


Yet in 2016 almost all the pollsters were proved wrong when Donald Trump won the presidency. On the eve of the election, Silver gave Hillary Clinton a 66.9% chance of beating Trump and projected that she would win 294 Electoral College votes. It was not for a lack of polling data. It was just that the polls nearly all had Clinton ahead.


What had gone wrong? The first problem with polls is representation. In 2016, surveys overrepresented college graduates, creating a bias in favor of Hillary Clinton. Second is the “observer effect,” when merely being asked a question in a survey influences how people respond to the question. Third is the tendency for polls themselves to become the news. The fact that news organizations became the biggest customers for pollsters created a perverse incentive. A race that goes down to the wire is more likely to sell papers than a foregone conclusion.



Polymarket advertised its electoral betting market at a Brooklyn subway entrance in July, though it is not open to U.S. customers. Photo: Michael Nagle/Bloomberg News

These are similar to the critiques of prediction markets. From Kominers’s idea that prediction markets are unrepresentative of the general population, to Shen’s concern that monetary incentives distort how users bet on Polymarket, to Sethi’s claim that users are betting primarily as a means to create momentum for candidates, we are dealing here with projection. These are in reality the sins of polling, not of markets.


Outperforming pollsters and pundits


Surprisingly, the history of prediction markets goes back further than the history of polling. Paul W. Rhode and Koleman S. Strumpf of UNC Chapel Hill have recovered data for “large and often well-organized markets for betting on presidential elections that operated between 1868 and 1940.”


But these early political betting markets waned following a series of setbacks, beginning with New York City Mayor Fiorello La Guardia’s crackdown during World War II. Betting was also hindered by federal bans on the interstate transmission of wagers, laws designed to fight organized crime and racketeering.


Political prediction markets took years to re-emerge under the wary supervision of the CFTC. It was not until 1988 that economists at the University of Iowa set up the Iowa Political Stock Market, today known as the Iowa Electronic Markets (IEM), which operates as a strictly educational project, with wagers capped at $500 and participation limited to students. Each market is furthermore capped at 1,000 or 2,000 participants, depending on the market question.


Twenty-six years later, in 2014, Victoria University of Wellington, New Zealand, established the second-most-cited prediction market, PredictIt. The platform obtained similar CTFC exemptions to those granted to IEM by following similar rules.


As this suggests, prediction markets continue to be severely hobbled by regulation. Liquidity is shallow—a drop in the bucket compared with any financial market. And transaction fees are so high it’s hardly worth betting.


Nevertheless, even with these disadvantages, prediction markets consistently outperform pollsters and pundits. A retrospective analysis of the IEM from 1988 through 2008 found that “predictions from markets dominate those from polls about 75% of the time, whether the prediction is made on election eve or several months in advance of the election.”



A pollster asks for the opinion of a voter during a pre-election survey on the streets of Washington, D.C., March 1964. Photo: United States Information Agency/PhotoQuest/Getty Images

That result is astounding given that the total amount of bets made over all IEM markets for the last 25 years was just $4.1 million—less than the amount Théo, the Polymarket “whale,” wagered in a single day.


Enter the Trump whale


As the maestro of value investors, Benjamin Graham, once said, “’In the short run, the market is a voting machine, but in the long run, it is a weighing machine.” Graham argued that prices are volatile in the short term, reflecting market sentiment, but in the longer term they tend to reflect a company’s true value. But Graham ignored two crucial variables—liquidity and size.


A better heuristic for understanding the past and future of prediction markets is that a shallow market is a voting machine, a deep market is a weighing machine. And crypto has begun to make prediction markets deeper than anyone thought possible.


In 2015 Joey Krug, now a partner at Founders Fund, did a Kickstarter for what would be the first major application built on the Ethereum blockchain: Augur. Augur was designed to be a universal prediction market. As a decentralized protocol—that is, without intermediaries—the platform sought to get around the ban on prediction markets imposed by the CFTC. Three years later, the Augur team launched their app, but it suffered from the same issues that plagued IEM—namely, low numbers of users and volume.


Polymarket has succeeded where Augur failed. Shayne Coplan founded the platform in 2020 not as a gambling app but as “something that is a beacon of truth.” Coplan’s aspiration, much like Krug’s before him, was to build a platform that could create markets to settle a broad array of matters of fact.


Cumulative Polymarket volume has surged from $73 million in 2023 to over $5 billion so far in 2024. Had Americans been allowed to participate, as they now are in Kalshi, the surge would have been even bigger. And of course success attracts competition. Interactive Brokers is the most recent firm to add election betting markets, following a September ruling by the District Court for the District of Columbia to overrule the CFTC and allow the prediction-market startup Kalshi to offer election betting. Robinhood added election betting just a week before the election. Silver himself joined Polymarket as an adviser earlier this year.


As these new markets have broadened and deepened, so has their ability to attract and process high-quality information. Very few Americans find it worth their time to wager on the IEM or PredictIt, given that at most one can net a few thousand dollars for being right. But with several billions in volume, Polymarket has proven attractive to international traders more akin to Wall Street wolves than undergraduates.


Polymarket’s remarkable liquidity explains Théo, the Polymarket whale. Théo behaved as one would expect of a smart trader: Once he had conviction, he sized the trade aggressively. He built his positions discreetly, with a variety of wallets and with small-sized buys, to avoid causing a run-up in price. This is exactly the opposite of what one would expect if, as much of the media and BlueAnon claimed, he was throwing away dozens of millions to influence the election.


More important, Théo did his own research. Rather than rely on polls, he theorized that there was a strong “shy Trumper” effect. That is, he believed that Trump supporters were both less likely to talk to pollsters and less likely to tell pollsters that they would vote for the former president. As he told The Wall Street Journal in an email, he commissioned his own surveys in which respondents were asked who they thought their neighbors would vote for. The results “were mind blowing to the favor of Trump!”


Not only did Théo bet that Trump would win the Electoral College. He also bet that Trump would win the popular vote—an outcome that was the minority position even on Polymarket—and that Trump would sweep six of seven swing states, including the “blue wall” of Michigan, Pennsylvania and Wisconsin.


The Wall Street Journal reports that Théo realized an $85 million windfall from his bet. Where the whale led, others will surely follow.


Pollsters as entertainers


What does all of this signify? The rise of prediction markets is inseparable from the broader decline of experts across many fields. In the same way that Elon Musk has declared to the users of X that “You are the media now,” Polymarket tells its users that they know better than the pollsters.


Though prediction markets have their shortcomings, the public will likely adjust to them in the same way they have internalized the limits of polling. One such limit is that markets tend to attract few traders early in a contest. While polling begins at the outset of an election year, before the candidates have been nominated, skilled traders like Théo will act only when there is enough reliable information for them to make a profitable wager. On the other hand, markets incorporate new information more quickly than polls. The odds in prediction markets changed immediately after President Biden’s dismal debate performance.


Polymarket’s big breakthrough this year also signals a coming explosion in liquidity for prediction markets, barring any adverse turn in the regulatory environment. This promises an exponential increase in the accuracy and therefore importance of betting markets.


In this new world, forecasters like Nate Silver will come to be seen more for what they are—entertainers in the same line of business as CNBC “Mad Money” host Jim Cramer. It’s no accident that on Monday, Nov. 4, Cramer told viewers that rising stock prices for home builders and importers might reflect growing investor confidence…in a victory for Vice President Harris.


Niall Ferguson is a senior fellow at the Hoover Institution at Stanford University and founder of the advisory firm Greenmantle. Manny Rincon-Cruz is the founder of Buttonwood, a decentralized finance project, and co-founder with Ferguson of FourWinds Research.

3 views0 comments

Recent Posts

See All

Inside the Four-Day Workweek Experiment

Results in conflict with the Spritlzer Report controlled work experiment. Arl Hts, Il. Last July a group of 1,000 Spritzler employees...

After the Fall of Syria’s Assad

Let's not declare victory just yet. Iran's government and support for it's terrorist proxies are funded by massive oil revenues...ergo...

Comments


Post: Blog2_Post
bottom of page