Crypto Betting Markets and Statisticians Diverge Ahead of the 2020 US Election

Humans have long been drawn to uncertain outcomes, seeking insight and entertainment from predicting the unknowable since long before recorded history. Technology has refined the craft of forecasting, but some degree of uncertainty about the future is unavoidable.

Living through COVID lockdowns made the influence of statistics unavoidably visible: daily life increasingly deferred to algorithms built by developers nobody elected. The broader forecasting industry runs on a constant appetite for insight into what happens next — gamblers, politicians, and professionals across fields all benefit from informed reads on future events, and a range of methods has emerged to try to supply that insight. Prediction markets are one such method, using money as an incentive to gauge the strength of public conviction.

Despite crypto offering conditions well suited to prediction markets, no decentralized platform had achieved genuine mainstream adoption, and every new entrant has struggled to build meaningful liquidity depth. Augur, the first decentralized prediction market, took three years to launch following its 2014 fundraising round; it saw some uptake around the 2018 US midterms but has never matched the volume or liquidity of centralized competitors.

By 2020, most political-outcome betting activity ran through centralized venues such as FTX and PredictIt. This dashboard tracks PredictIt markets in real time, while Coingecko ranks 2020 election-themed tokens by market capitalization.

With the outcome of a deeply contested election still unresolved, more than $3 million had been staked on Polymarket over the question of whether Trump would win a second term — a market where "No" had been the leading answer continuously since trading opened on October 10th. Both FTX (centralized) and Polymarket (decentralized) were pricing Trump's odds of victory at roughly 37.8%, notably higher than the 4-14% range assigned to him by conventional statisticians and traditional polling models.

Ethereum co-founder Vitalik Buterin offered two competing explanations for that gap on Twitter. Under what might be called a "pro-prediction market" view, he argued these markets more accurately "incorporate the possibility of heightened election meddling, voter suppression, etc. affecting the outcome," while conventional statistical models simply "assume the voting process is fair." Under an alternative "pro-statistics" explanation, he suggested prediction markets remain difficult for professional statisticians and political experts to access, are too small in scale for hedge funds to justify hiring such experts, and that the people — especially the wealthy ones — with the greatest access to these markets tend to be more optimistic about a Trump win. He dismissed as unlikely a third possibility: that professional forecasters are simply too stubborn to have learned from past failures to detect unexpected pro-Trump support.

That same day, statistician Nate Silver published FiveThirtyEight's final election forecast, giving Joe Biden the edge over Trump while still granting Trump roughly a 1-in-10 chance, projecting a 3-in-4 likelihood that Democrats would retake the Senate, and forecasting that the House would most likely remain under Democratic control, possibly with an expanded majority.

Is political polling genuinely less efficient than prediction markets? Markets are well suited to surfacing material non-public information — if a subset of participants know an outcome the public doesn't, and stakes grow large enough, someone eventually trades on that private knowledge and moves the price, as one commentator, @jdh, has pointed out. In this election, however, essentially nobody held genuinely decisive inside information, which limits how much edge market incentives alone could provide.

For now, it appears statisticians hold the advantage over prediction markets — though the same statisticians were proven wrong in 2016, and could be again. As these markets mature and blockchain adoption expands, crowd-sourced wagering may eventually out-forecast even the most skilled individual statisticians.

On the question of DeFi under either administration: some believe a Biden victory would be more likely to normalize Bitcoin and open the door to an ETF, framed in one outlook as follows: "A potential Joe Biden presidency should shine favor on further appreciation in the price of Bitcoin, in our view. New leadership may change the hands-off policy of the Trump administration — to the detriment of the broader crypto market — and nudge the firstborn benchmark toward the mainstream, improving chances for an ETF." Yet a separate Bloomberg crypto newsletter had already described the Trump administration's approach to crypto as "hands-off" — raising the question of whether continued regulatory distance might actually suit an industry still finding its footing, free from interference as this new financial ecosystem grows. Bitcoin and DeFi appear to be moving on separate tracks: large holders have been accumulating BTC at a record pace even as DeFi continues to be viewed as comparatively risky and unexplored territory.

If forecasting rests on accumulated data the way structures rest on their foundations, then whichever candidate wins, the next administration will be forced to confront growing demand for alternative currencies over the following four years. In a year short on conventional sports betting, crypto-based prediction markets provided an accessible outlet for political speculation — and as volume and adoption grow, these markets will generate ever more data to refine the forecasting models that follow.

Artwork credited to artpad.org and the 1936 film Things to Come.