Polymarket has emerged as a rising star in consumer crypto, especially driven by the upcoming U.S. presidential election and the related events. Many crypto experts have praised Polymarket’s success in achieving product-market fit, notably without relying on tokens or external incentives. As of this writing, Polymarket has hit another all-time high in daily active users and transactions. With over half of October remaining, it has already surpassed September's total monthly trading volume.

Besides its success in most metrics, observers have praised its role as a more timely and nuanced source of information compared to traditional media, polls, and other public opinion channels. As Nassim Nicholas Taleb wrote in Skin in the Game: Hidden Asymmetries in Daily Life, “Don’t tell me what you “think”, just tell me what’s in your portfolio.”

In August, Bloomberg even started including Polymarket’s odds in its terminal service as an additional data source for analytics.

Let's explore the differences between prediction markets like Polymarket and traditional polls, along with other proxies of public opinion, such as social media and expert commentary. Each of these channels helps us gauge sentiment and make forecasts about future events.

This chart contrasts various channels for gauging public sentiment and forecasting events, positioning them across two axes: Grassroots vs. Top-down and Dilettantism vs. Expertise.

• Polymarket, like other permissionless prediction markets (e.g., Drift’s BET on Solana, or Wintermute’s OutcomeMarket, an open-source multichain US election prediction market), operates as a fully decentralized, blockchain-based platform. Since anyone can create new markets and place bets, these prediction markets are inherently grassroots. The financial skin-in-the-game, combined with the need to exploit information asymmetry to win, motivates participants to gather and act on the best possible information, making the platform expertise-dense.

• Expert commentary from outlets like The New York Times, The Economist, and The Wall Street Journal is highly expertise-dense, often with reputational stakes. However, these opinions represent a limited subset, strictly aligned with editorial guidelines, which makes them highly top-down.

• Traditional polls (e.g., YouGov, Ipsos, Pew Research) focus on gauging political convictions and beliefs rather than expectations. They typically rely on limited randomized sampling with mixed expertise, lacking the financial incentives that drive better-informed predictions in markets like Polymarket. Traditional polls are driven by deliberate research design choices which may also include biases, so it’s more towards the top-down approach.

• Social media allows everyone to express their opinions, making it grassroots by nature. However, centralized platforms are driven by highly-opinionated algorithms and selective censorship, making them far less grassroots compared to platforms like Polymarket. Moreover, fake news and disinformation campaigns, often driven by swarms of bots, create a high-noise, low-signal environment, making it difficult to filter out informed opinions and verified facts. This lowers the overall quality of discourse, leading to a tendency toward dilettantism.

Polls and prediction markets fundamentally ask different questions. In the context of a political election, traditional polls ask, "Who will you vote for?"—capturing individual intentions and preferences. Prediction markets, however, ask, "Who do you think will win?"—with financial incentives for accurate predictions, regardless of personal preferences. These two perspectives—intention versus expectation—often don’t align. The possibility of hedging against undesired outcomes further complicates the distinction (e.g., betting on Trump while supporting Biden). While both methods aim to forecast future events, this difference explains why poll results and prediction markets often diverge.

Another element to consider when thinking about prediction markets is the degree of influence that the user placing the bet has on the event. This variance makes multi-domain bets significantly different from sports-only bets.

In some cases, prediction markets reflect events where participants have direct influence over the outcomes, as is seen in elections. Here, the market dynamics can mirror the collective actions and decisions of voters, though it’s worth noting that U.S. citizens are not permitted to use platforms like Polymarket for this purpose.

In other scenarios, such as with inflation or economic policies, the influence is more indirect. Market expectations can shape behaviors that, in turn, influence outcomes. For instance, if households expect an interest rate cut, they may delay spending in anticipation of a more favorable economic environment. When this kind of behavior becomes widespread, it can help keep inflation low, potentially encouraging the central bank to actually lower interest rates. This is a classic case of a self-fulfilling prophecy, where the expectation of an event (the rate cut) leads to actions that bring about the expected outcome.

Finally, there are also events where participants have no influence at all, such as in sports competitions or natural phenomena. Here, prediction markets are purely speculative, as participants cannot affect the final outcome in any way.

Since prediction markets work 24/7 and reward accuracy, participants in Polymarket are far more reactive to real-time news than poll respondents. This volatility, which represent another major unique characteristic compared to traditional channels, can lead to more reactive and accurate measurements, but it can also skew market results, as it seemed the case in the disparity between Polymarket's odds and traditional polls during the 2024 Biden-Trump election campaign.

Throughout 2024, Polymarket consistently showed Trump with a lead of several percentage points over Biden, even when traditional polls indicated a much closer race. Furthermore, the gap fluctuated significantly following major events, such as the first public debate or the attempted assassination of Trump. In contrast, traditional polls remained more stable throughout. Many commentators expected market forces and arbitrage to correct this gap, but it persisted until Biden's official withdrawal. On the other hand, looking at traditional polls, we would have thought that Biden had similar chances of winning as Trump. This looked like the typical example of extreme volatility of prediction markets.

This chart clearly shows the significant impact the first debate had on public opinion, with bets in favor of Biden plummeting from nearly 40% to around 10% within just a few days. Interestingly, by the time the official withdrawal was announced, Polymarket had already priced in the outcome, with Biden's odds hovering close to 0%. In contrast, traditional polls continued reporting similar percentages as before.

This apparent distortion actually reveals a powerful dynamic: Polymarket participants likely anticipated Biden’s potential withdrawal (a scenario with its own dedicated market on Polymarket), and adjusted their bets accordingly, which led to consistently lower betting on Biden. In other words, Biden's chances of winning might have been lower because the market had already factored in the possibility of his withdrawal. If Biden’s candidacy had remained secure, the chances of him winning over Trump may have been higher, with numbers similar to what traditional polls kept showing until the official withdrawal.

Not coincidentally, with Biden officially out of the race and the looming possibility of withdrawal canceled from the equation, Polymarket’s election market (Harris vs. Trump) has stabilized and almost perfectly aligned with traditional media polls.

This chart illustrates how Polymarket’s bets in favor of Harris shifted after the first debate and, more notably, following Joe Biden's official withdrawal. Interestingly, odds for Trump also declined with Biden's announcement, eventually reverting to levels more in line with traditional polls.

In traditional polls, the probability that a specific candidate may win can be inferred by the amount of perspective votes the candidate receives compared to the total votes:

This reflects a binary view, where pollsters assess only the relative support between candidates, without accounting for complex external factors.

In prediction markets, on the other hand, participants often incorporate second-order considerations, such as the possibility of unexpected events (e.g., a candidate withdrawing). So, the equation for estimating the probability could be expanded to:

This additional factor—Biden’s potential withdrawal—demonstrates how prediction markets account for broader scenarios, leading to different results compared to traditional polls and commentary channels. By considering factors like withdrawals, scandals, or health issues, prediction markets capture not just head-to-head probabilities, but also second-order and tangentially-related events.

This distinction highlights the value of prediction markets as more dynamic and nuanced predictive tools for complex events.

Polymarket has emerged as a rising star in consumer crypto, especially driven by the upcoming U.S. presidential election and the related events. Many crypto experts have praised Polymarket’s success in achieving product-market fit, notably without relying on tokens or external incentives. As of this writing, Polymarket has hit another all-time high in daily active users and transactions. With over half of October remaining, it has already surpassed September's total monthly trading volume.

Besides its success in most metrics, observers have praised its role as a more timely and nuanced source of information compared to traditional media, polls, and other public opinion channels. As Nassim Nicholas Taleb wrote in Skin in the Game: Hidden Asymmetries in Daily Life, “Don’t tell me what you “think”, just tell me what’s in your portfolio.”

In August, Bloomberg even started including Polymarket’s odds in its terminal service as an additional data source for analytics.

Let's explore the differences between prediction markets like Polymarket and traditional polls, along with other proxies of public opinion, such as social media and expert commentary. Each of these channels helps us gauge sentiment and make forecasts about future events.

This chart contrasts various channels for gauging public sentiment and forecasting events, positioning them across two axes: Grassroots vs. Top-down and Dilettantism vs. Expertise.

• Polymarket, like other permissionless prediction markets (e.g., Drift’s BET on Solana, or Wintermute’s OutcomeMarket, an open-source multichain US election prediction market), operates as a fully decentralized, blockchain-based platform. Since anyone can create new markets and place bets, these prediction markets are inherently grassroots. The financial skin-in-the-game, combined with the need to exploit information asymmetry to win, motivates participants to gather and act on the best possible information, making the platform expertise-dense.

• Expert commentary from outlets like The New York Times, The Economist, and The Wall Street Journal is highly expertise-dense, often with reputational stakes. However, these opinions represent a limited subset, strictly aligned with editorial guidelines, which makes them highly top-down.

• Traditional polls (e.g., YouGov, Ipsos, Pew Research) focus on gauging political convictions and beliefs rather than expectations. They typically rely on limited randomized sampling with mixed expertise, lacking the financial incentives that drive better-informed predictions in markets like Polymarket. Traditional polls are driven by deliberate research design choices which may also include biases, so it’s more towards the top-down approach.

• Social media allows everyone to express their opinions, making it grassroots by nature. However, centralized platforms are driven by highly-opinionated algorithms and selective censorship, making them far less grassroots compared to platforms like Polymarket. Moreover, fake news and disinformation campaigns, often driven by swarms of bots, create a high-noise, low-signal environment, making it difficult to filter out informed opinions and verified facts. This lowers the overall quality of discourse, leading to a tendency toward dilettantism.

Polls and prediction markets fundamentally ask different questions. In the context of a political election, traditional polls ask, "Who will you vote for?"—capturing individual intentions and preferences. Prediction markets, however, ask, "Who do you think will win?"—with financial incentives for accurate predictions, regardless of personal preferences. These two perspectives—intention versus expectation—often don’t align. The possibility of hedging against undesired outcomes further complicates the distinction (e.g., betting on Trump while supporting Biden). While both methods aim to forecast future events, this difference explains why poll results and prediction markets often diverge.

Another element to consider when thinking about prediction markets is the degree of influence that the user placing the bet has on the event. This variance makes multi-domain bets significantly different from sports-only bets.

In some cases, prediction markets reflect events where participants have direct influence over the outcomes, as is seen in elections. Here, the market dynamics can mirror the collective actions and decisions of voters, though it’s worth noting that U.S. citizens are not permitted to use platforms like Polymarket for this purpose.

In other scenarios, such as with inflation or economic policies, the influence is more indirect. Market expectations can shape behaviors that, in turn, influence outcomes. For instance, if households expect an interest rate cut, they may delay spending in anticipation of a more favorable economic environment. When this kind of behavior becomes widespread, it can help keep inflation low, potentially encouraging the central bank to actually lower interest rates. This is a classic case of a self-fulfilling prophecy, where the expectation of an event (the rate cut) leads to actions that bring about the expected outcome.

Finally, there are also events where participants have no influence at all, such as in sports competitions or natural phenomena. Here, prediction markets are purely speculative, as participants cannot affect the final outcome in any way.

Since prediction markets work 24/7 and reward accuracy, participants in Polymarket are far more reactive to real-time news than poll respondents. This volatility, which represent another major unique characteristic compared to traditional channels, can lead to more reactive and accurate measurements, but it can also skew market results, as it seemed the case in the disparity between Polymarket's odds and traditional polls during the 2024 Biden-Trump election campaign.

Throughout 2024, Polymarket consistently showed Trump with a lead of several percentage points over Biden, even when traditional polls indicated a much closer race. Furthermore, the gap fluctuated significantly following major events, such as the first public debate or the attempted assassination of Trump. In contrast, traditional polls remained more stable throughout. Many commentators expected market forces and arbitrage to correct this gap, but it persisted until Biden's official withdrawal. On the other hand, looking at traditional polls, we would have thought that Biden had similar chances of winning as Trump. This looked like the typical example of extreme volatility of prediction markets.

This chart clearly shows the significant impact the first debate had on public opinion, with bets in favor of Biden plummeting from nearly 40% to around 10% within just a few days. Interestingly, by the time the official withdrawal was announced, Polymarket had already priced in the outcome, with Biden's odds hovering close to 0%. In contrast, traditional polls continued reporting similar percentages as before.

This apparent distortion actually reveals a powerful dynamic: Polymarket participants likely anticipated Biden’s potential withdrawal (a scenario with its own dedicated market on Polymarket), and adjusted their bets accordingly, which led to consistently lower betting on Biden. In other words, Biden's chances of winning might have been lower because the market had already factored in the possibility of his withdrawal. If Biden’s candidacy had remained secure, the chances of him winning over Trump may have been higher, with numbers similar to what traditional polls kept showing until the official withdrawal.

Not coincidentally, with Biden officially out of the race and the looming possibility of withdrawal canceled from the equation, Polymarket’s election market (Harris vs. Trump) has stabilized and almost perfectly aligned with traditional media polls.

This chart illustrates how Polymarket’s bets in favor of Harris shifted after the first debate and, more notably, following Joe Biden's official withdrawal. Interestingly, odds for Trump also declined with Biden's announcement, eventually reverting to levels more in line with traditional polls.

In traditional polls, the probability that a specific candidate may win can be inferred by the amount of perspective votes the candidate receives compared to the total votes:

This reflects a binary view, where pollsters assess only the relative support between candidates, without accounting for complex external factors.

In prediction markets, on the other hand, participants often incorporate second-order considerations, such as the possibility of unexpected events (e.g., a candidate withdrawing). So, the equation for estimating the probability could be expanded to:

This additional factor—Biden’s potential withdrawal—demonstrates how prediction markets account for broader scenarios, leading to different results compared to traditional polls and commentary channels. By considering factors like withdrawals, scandals, or health issues, prediction markets capture not just head-to-head probabilities, but also second-order and tangentially-related events.

This distinction highlights the value of prediction markets as more dynamic and nuanced predictive tools for complex events.