What the Markets Got Right
On the eve of the 2024 election, Polymarket priced Trump at approximately 57% to win. Most major polling aggregates showed a toss-up, with some models giving Kamala Harris a slight edge. Trump won decisively, and the prediction market was closer to the actual outcome than the polls.
This was a significant validation for prediction markets, particularly because the election was one of the most heavily traded events in prediction market history, with over $3.3 billion wagered on the Trump vs Harris outcome on Polymarket alone. The high liquidity meant the price was not easily manipulated and reflected broad participation.
But the accuracy went beyond just the presidential race. Prediction markets correctly identified several key swing states that traditional polling missed. Pennsylvania showed Trump at 58% on Polymarket three days before the election, while most polls had it as a pure toss-up. Wisconsin and Michigan followed similar patterns, with prediction markets giving Trump slight edges that materialized on election night.
The speed of information incorporation was particularly striking. When early voting data from North Carolina suggested higher Republican turnout than expected, Polymarket prices moved within hours. Traditional polls, constrained by their methodology and publication schedules, couldn't react as quickly to emerging signals.
What Complicates the Story
A 57% probability still means 43% chance of the other outcome. If Harris had won, the prediction market would have been "wrong" in the sense of not predicting the winner, but correct in the sense of saying the outcome was uncertain. Evaluating prediction markets based on whether they picked the winner is the wrong framework. The right framework is calibration: are events they price at 57% actually happening about 57% of the time across many events?
The French whale who reportedly moved the Trump contract through massive directional bets raises questions about whether the price was reflecting genuine crowd wisdom or the conviction (and superior information, in this case private polling data) of a single well-capitalized trader. In this instance, the whale was right. But the mechanism of price discovery was closer to "one very well-informed trader" than "crowd wisdom" in the traditional sense.
This whale, identified through blockchain analysis, placed over $30 million in bets across multiple accounts, consistently buying Trump contracts even when prices rose. The trades were sophisticated, often executed during low-volume periods to minimize market impact. When news broke about the trader's private polling data showing Trump strength in key demographics, it became clear this wasn't just a gut feeling bet.
The incident highlights a fundamental tension in prediction markets. Large, well-informed traders can improve price accuracy by incorporating superior information. But they can also distort prices if their information is wrong or if they're trading on non-informational factors like risk preferences or hedging needs.
Information vs. Manipulation
Distinguishing between informed trading and manipulation becomes crucial for interpreting prediction market signals. Our Whale Finder tool tracks large position accumulations across major platforms, helping identify when price movements might be driven by individual actors rather than broad consensus.
In the 2024 election, several other large traders emerged beyond the famous French whale. A group of crypto investors collectively wagered over $50 million on Trump, citing concerns about Democratic cryptocurrency regulation. Their trades weren't based on electoral analysis but on policy preferences, creating a potential bias in the market pricing.
Platform Differences Were Revealing
Accuracy varied across platforms. PredictIt correctly predicted 93% of outcomes better than chance. Kalshi achieved 78%. Polymarket achieved 67%. These differences reflect structural factors: different user bases, different liquidity levels, and different incentive structures. No single platform was uniformly best across all types of prediction.
PredictIt's superior performance likely stems from its user base of political professionals and academics. The platform's $850 betting limit prevents whale manipulation but also reduces liquidity. Users tend to be more informed about political nuances but have less capital at risk, creating different incentive structures.
Kalshi's middle-ground performance reflects its position between PredictIt's informed but capital-constrained users and Polymarket's high-liquidity but potentially less-informed crypto traders. Kalshi's regulatory status as a CFTC-regulated exchange attracts institutional participants who bring sophisticated analysis but also face compliance constraints that can slow their reaction times.
Polymarket's lower accuracy rate, despite its massive volume, illustrates the complexity of market efficiency. High liquidity doesn't automatically translate to better predictions if the participant base lacks relevant expertise or if large traders can disproportionately influence prices.
Cross-Platform Arbitrage Opportunities
The platform differences created significant arbitrage opportunities throughout the election cycle. Trump contracts traded as much as 8 percentage points higher on Polymarket than on Kalshi during certain periods. These spreads persisted for hours, sometimes days, due to friction in moving capital between platforms.
Sophisticated traders exploited these differences, but the arbitrage wasn't perfect. Regulatory restrictions prevented many traders from accessing multiple platforms. Withdrawal delays and gas fees on Polymarket created additional friction. The result was persistent pricing disparities that revealed how fragmented the prediction market ecosystem remains.
Beyond the Presidential Race
Senate and gubernatorial races provided additional testing ground for prediction market accuracy. Markets correctly identified several upset victories, including unexpected Republican wins in states where traditional polling showed tight races favoring Democrats.
In Montana, prediction markets gave Republican Tim Sheehy a 65% chance of defeating incumbent Democrat Jon Tester, while most polls showed a dead heat. The markets proved correct, with Sheehy winning by a comfortable margin. Similar patterns emerged in Ohio and Florida, where prediction markets identified Republican strength that polling missed.
However, markets also missed some outcomes. Several House races in California and New York saw Democratic candidates win seats that prediction markets had favored Republicans to hold. The pattern suggests prediction markets perform better in higher-profile, heavily-traded contests where more information and attention flow to price discovery.
State-level ballot initiatives showed mixed results. Prediction markets accurately forecasted abortion rights measures in several states but missed the mark on some drug policy reforms. The accuracy seemed to correlate with the amount of trading volume each contract attracted.
Lessons for Future Elections
The 2024 election taught several practical lessons. Prediction markets are better than polls at incorporating information rapidly, but they can be influenced by large individual traders. Cross-platform comparison is more informative than any single platform's price. And the right metric for evaluation is calibration across many events, not whether the market "called" any single outcome correctly.
The whale phenomenon suggests prediction markets work best when large traders have genuine information advantages rather than just strong opinions or hedging needs. Platforms need better tools to distinguish between informed and uninformed large trades.
Regulatory clarity emerged as a crucial factor. Platforms operating in regulatory gray areas attracted different user bases and capital flows than those with clear legal status. This affected not just volume but the quality of price discovery.
For traders and analysts, the election highlighted the value of monitoring multiple platforms simultaneously. Price divergences often signaled important information or arbitrage opportunities. Our Prediction Markets Mispricing Engine tracks these spreads across major platforms, helping identify when consensus breaks down.
The speed advantage over traditional polling proved significant but not absolute. Prediction markets excel at incorporating breaking news and early indicators, but they can also overreact to noise or be manipulated by coordinated trading. The most reliable signals come from sustained price movements supported by growing volume across multiple platforms.
Looking ahead, the infrastructure for prediction market analysis needs to mature. Better tools for tracking large traders, identifying informed vs. uninformed flow, and measuring true calibration across diverse event types will help separate signal from noise in future electoral cycles.
Explore these tools on Blockcircle: Prediction Markets Mispricing Engine | Whale Finder