On October 25, Polymarket’s forecast for the US presidential elections turned out to be exactly the same as those that ended up becoming official after all the votes were counted. Donald Trump’s landslide victory over Kamala Harris was anticipated with astonishing accuracy. At least that’s what it seemed like, but in reality there is much more behind this prediction platform. What is Polymarket, and how is it different from surveys and expert forecasts? Let’s see it.
Prediction markets. Polymarket is, along with Kalshi and PredictIt, the great reference for so-called prediction markets, a type of systems in which users—from anywhere in the world—bet real money on the results of elections (or any other event). People gamble by buying “stocks” that range between $0.01 and one dollar. The value of the shares reflects the opinion of the users: if Trump’s shares are trading at $0.70, for example, that indicates that there is a consensus that Trump has a 70% chance of winning. From there, users can “bet” on Trump or Harris, which changes the value of the shares and the general sentiment.
The base. Those who defend platforms like Polymarket argue that by having to pay for actions (that is, by betting) and by having so many users, the system is much more accurate than, for example, surveys in which people can directly lie. The theory is clear: if there is a probable outcome, investors will not let the stock lose value since they can make money from that prediction. There is another element: investors take advantage of information from surveys and analysts to incorporate it into their bets, thus reinforcing (at least, theoretically) the validity of their bets. Those who get it right get a dollar per share. Those who don’t, lose what they bet.
A crypto-based infrastructure. Polymarket, for example, is based on Polygon, a technology derived from the Ethereum proof-of-stake blockchain. All transactions are made in USDC, a stablecoin linked to the US dollar that allows volatility to be avoided and also imposes very low commissions on transactions.
Polls and electoral forecasts. Traditional methods for predicting electoral results have always focused on two large groups.
- Surveys. The best-known system is based on something simple: asking a group of people—the sample—about their voting intention. These results are extrapolated to a large model, but they do not assign probabilities of who will win: they simply represent the number of people who support a candidate at a given time. Surveys can be useful, but they can also fail miserably.
- Election forecasts from experts. In the US, FiveThirtyEight, The Hill or analysts like Nate Silver (who created FiveThirtyEight) are famous. These platforms use surveys and other data (economic indicators, historical trends) to predict voting intention and the probability that one candidate or another will win. The aforementioned Nate Silver was in fact one of the “winners” of the 2016 elections: he successfully predicted that Trump would win that electoral process.
Too much volatility. What that forecast stated on October 25 was almost anecdotal. Polymarket simply boasted that at that time its system had accurately predicted the election result, but in reality that prediction kept changing. It was different before, when the balance was slightly tipped for both Trump and Harris, for example. And it also changed later, although always in favor of Trump. In fact, the day before – and the days before that day – the elections, Polymarket’s forecast was very close to the final result. Volatility is notable in this type of systems, and there are also factors that can influence these predictions.
Freddi9999. That alias is that of a French investor who ended up betting millions of dollars that Trump would win the elections. The bet turned out very well, and according to Bloomberg its profits – with data from the crypto consulting firm Chainalysis – could have amounted to $85 million. However, this immense bet logically made the validity of the predictions conditional, and demonstrates how these “whales” of the prediction markets amplify the opinions of speculators. They can influence the final result, since they condition the users who bet later.
But Polymarket and Kalshi have their advantages. Even so, these decentralized systems can display valuable information, and above all, they are capable of reacting much faster to all types of events during the electoral race. As they point out in The New York Times, in June it happened with that debate in which Joe Biden was so erratic. The pollsters were slow to react to the question of whether he should abandon his candidacy for re-election, but in Polymarket the change in the indicator was clear: before the debate the shares that were betting that he would abandon were at 21 cents. The morning after the debate they rose to 43 cents, and a week later to 60. Biden would end up announcing his withdrawal a month later. It is also true that this rapid reaction to events that occur contributes to the volatility of predictions.
Regional and monetary limits. In the US, betting on electoral processes is prohibited, and Polymarket only allowed operations outside that country. It is not known if, for example, US citizens have used VPNs to participate. Kalshi obtained an exception to operate in these elections and citizens were able to bet there. Polymarket does not limit the amount of money users can bet, but for example PredictIt limits bets to $850.
But who gets it right more? The truth is that at least in these elections Polymarket has become one of the indirect winners. Their predictions, like those of Kalshi or PredictIt, were closer to the final result. Both polls and analysts pointed to much closer victories for Trump. And yet, it remains unclear whether predictive markets are better than other prediction systems.
Imagen | Polymarket
In WorldOfSoftware | Elon Musk runs five companies, has participated in an election and has had plenty of time to break records in Diablo IV
