What is a prediction market? Prediction markets are open platforms for making bets on the outcome of various events. These can include the outcome of an election, a sports match, an economic event or something else. Themes are theoretically unlimited. This type of market therefore offers the possibility of betting on predictions and being paid according to the outcome.
By aggregating these bets, we obtain a collective view of the probabilities of an event coming true. This can be used to estimate the state of opinion on various subjects at a given moment. Since the reward is financial and not linked to individual opinions, prediction markets can theoretically offer a more objective perspective on the chances of an event coming true.
In the context of an election, for example, prediction markets can be used to visualise public opinion on the chances of a candidate winning the election. This differs from polls that simply collect voting intentions - it is important to note that voting intentions are not equivalent to the probabilities of a candidate being elected.
The history of prediction markets To fully understand the evolution and importance of prediction markets, we need to delve into their rich past and their many applications, which go far beyond simple betting. As early as 1503, people were betting on the identity of the next pope in Renaissance Italy, a time when betting was not just a form of entertainment, but also a means of gauging public opinion in restricted spheres. This type of prediction can also be found in 17th-century England, where informal bets were held in coffee houses to speculate on subjects as varied as the outcome of battles or the political rise of notorious figures.
In the 19th century, the United States, and Wall Street in particular, marked a major turning point. Betting on presidential elections, in particular, became so popular that it was seen as an indicator of choice, often more reliable than polls. In 1916, for example, Wall Street bookmakers estimated Woodrow Wilson's chances of re-election at 85%, a rate that corresponded almost perfectly with the final result. The funds raised through these bets were sometimes ploughed back into campaigns, thus indirectly helping to influence the course of political events themselves.
The twentieth century then saw the rise of the concept of the "wisdom of crowds", theorised in particular by the statistician Francis Galton. The latter had observed that during a cattle weight contest, the average of the estimates made by spectators was surprisingly close to the actual weight, sometimes exceeding the predictions of experts. This observation gave rise to the idea that groups, when sufficiently diverse and independent, can make more accurate predictions than those of isolated individuals, even if they are specialised. This concept has largely contributed to establishing the credibility of prediction markets as potential instruments for reliable forecasting.
The ongoing innovation around prediction markets reached a milestone in 1988, with the launch of the Iowa Electronic Markets (IEM) by the University of Iowa, one of the first electronic prediction markets allowing financial bets. Initially intended to anticipate the results of US elections, this academic market demonstrated the effectiveness of this approach: the IEM forecasts were often more accurate than traditional polls, partly thanks to the financial incentive that encouraged participants to find out and bet wisely.
Since then, major companies have followed this example to improve their own forecasts. Google, for example, set up internal prediction markets, called Prediction Markets , where employees could place fictitious bets on the feasibility of projects, the release dates of new products, and even issues as strategic as the market adoption of products. These markets proved particularly useful for detecting potential problems upstream, by revealing the collective intuitions of those working directly on projects.
Other companies, such as Hewlett-Packard and Microsoft, have also tested internal prediction markets to anticipate outcomes, particularly in the areas of sales and technological development. Hewlett-Packard, for example, has used markets to estimate demand for its printers, finding that these forecasts based on employee knowledge were often more accurate than estimates based on historical data alone.
The use of prediction markets has also extended to the political and social arena on a national and international scale, with platforms such as PredictIt in the US or Betfair in the UK. These platforms offer a space where anyone can place bets on political, economic or social events, thereby generating a "collective intelligence" that is of interest to observers and decision-makers alike.
How prediction markets work Prediction markets work like exchanges specialising in betting on future events, where each bet represents a position on the likely course of a given event. The mechanism is simple but sophisticated, allowing prices to reflect in real time the collective perception of the probabilities of specific outcomes. Share prices generally vary between 0 and 1 dollar using a system based on 'binary contracts'. Each contract represents a binary outcome (yes/no) or a range of possible outcomes for more complex events.
Participants buy and sell shares based on their predictions, and each transaction affects prices to reflect the perceived probability of each outcome. Let's take the hypothetical example of the 2024 presidential election: if the question is "Will Trump win the election?", participants buy "Yes" or "No" shares based on their expectations. If the price of a "Yes" share rises, this means that a majority of participants believe that his chances of victory are increasing, and this rise in price translates into a higher estimate of the probability of success. For example, if the "Yes" share reaches $0.70, the market implicitly believes that Trump has a 70% chance of winning, although this is only a probabilistic interpretation.
The binary nature of these contracts gives rise to the "winner takes all" phenomenon: once the event is over and the official result is known, punters who chose the right option see their shares reach the maximum value, i.e. $1, while the others lose everything. This model differs from traditional financial options, where gains and losses are often modulated. Prediction markets are thus dynamic and responsive tools that translate individual predictions into fascinating collective intelligence.
Some prediction markets also offer multiple contracts or results by tranche. For example, for the question "What percentage of the vote will Trump get in 2024?", participants can buy shares in intervals such as "40-45%", "45-50%" and so on. This allows results to be captured with more nuance and more detailed probabilities to be derived. These types of multi-outcome markets are used for complex predictions, such as quarterly economic growth estimates or the results of multi-round sporting events.
The integration of cryptos and decentralised models The arrival of cryptocurrencies has truly revolutionised prediction markets by enabling their complete decentralisation via blockchain and smart contracts, creating a secure, transparent ecosystem open to all. Augur, launched in 2018 on the Ethereum blockchain, marked the beginning of this transformation by offering a decentralised prediction market without intermediaries. Thanks to smart contracts, participants can create their own markets, and transactions, which are automatic and immutable, eliminate any risk of fraud or manipulation.
Operation on Augur is based on the classic prediction market model: users buy shares representing the outcomes of an event, for example "Yes" or "No" for binary questions. For multiple outcome questions, participants can also bet on specific intervals or customised scenarios, such as "Will New York rainfall exceed 10cm in December?", offering a variety of betting configurations.
Augur also enables a conflict resolution system via verification mechanisms. For example, if the outcome of an event is disputed, a community validation process, using the REP token, steps in to determine the actual outcome, making the system resistant to manipulation. This model incentivises REP holders to make honest checks, as any attempt to misreport will result in financial penalties.
This project has been considered inactive since the end of 2021. The main reason for this failure can be explained by the poor execution of its team, as it was never able to offer a simple, user-friendly interface.
Polymarket (launched in 2020) has since established itself as the major platform. It uses the Polygon blockchain, renowned for its low transaction fees and speed. Since its inception, Polymarket has grown rapidly, accounting for around 14% of the total locked-in value (TVL) on Polygon, demonstrating the popularity of these new forms of prediction. On Polymarket, we find classic events such as elections or sports results, but also atypical bets on various topical subjects. One notable example is where users could bet on specific media statements, such as the number of times Donald Trump would utter the word "Russia" in an interview.
More complex predictions are also being explored, such as bets on climatic events. Users can bet on specific variables such as the average temperature of a season in a given city, tapping into external data sources through oracles, specialist services that connect the blockchain to the real world. These oracles verify the data autonomously and inject the results into the smart contract, automatically triggering payments based on the verified result.
How confident can we be in their probabilities? Take, for example, the prediction of Fed decisions: while Polymarket recently correctly anticipated the level of interest rate cuts, 92% of economists usually surveyed by Bloomberg were wrong. How can such a difference be explained? Prediction markets benefit from a diversified flow of information in real time: punters keep a close eye on speeches, economic indicators and even the subtleties of language used by Fed officials, which helps them to assess the likelihood of rate changes more finely. This plurality of perspectives allows the market to approximate a more accurate consensus than a prediction issued by a panel of experts.
But its prediction markets do not guarantee infallible accuracy. They capture the "wisdom of crowds" by tapping into the intuitions and individual information of thousands of participants, but their predictions are not always accurate, and they can be wrong in the face of uncertainty or hidden information.
A striking example came recently when HBO announced a documentary promising to reveal the identity of Satoshi Nakamoto, the creator of Bitcoin. On Polymarket, around 50% of punters thought Len Sassaman would be revealed as Satoshi. However, the documentary ended up naming Peter Todd, an unexpected and disappointing result for a majority of punters. This failure highlights the vulnerability of prediction markets to rumour and hype: when crucial information is missing, even a well-informed market can fail.
What then can we conclude? Prediction markets are not infallible oracles; they provide indications based on an aggregation of perspectives and intuitions, but not a certain "truth". This lack of certainty is often amplified by the behaviour of insiders. Paradoxically, insiders - those who sometimes know an outcome in advance - may deliberately hide the truth in their bets to avoid revealing their knowledge, or even influence other punters by steering the odds one way or the other. This effect is particularly visible in markets with high financial or political stakes, where insiders choose not to intervene to avoid suspicion or compromising their own interests.
Thus, prediction markets should be understood as anticipatory tools, not certainties. Experienced traders use them for their insights, without seeing them as definitive answers. They can provide valuable insights, particularly when they reflect a consensus of several hundred or thousands of participants; however, those who consult them must bear in mind the potential for bias and manipulation. Prediction markets therefore have real added value in interpreting uncertain events, but their effectiveness depends on the transparency and reliability of the information that punters use.
The difference between a prediction market and a poll Polls and prediction markets offer two very different views of the probabilities of future events, and this distinction is crucial to understanding their respective usefulness.
A poll asks participants to respond about their personal intentions or preferences: for example, "Who will you vote for?" or "Do you think the economy will improve?". These responses reflect a snapshot of individual opinions and intentions, but do not necessarily indicate the final outcome. At election time, polls often measure voting intention in representative samples of a given population, but they can be biased by factors such as the over-representation of certain groups or the reluctance of respondents to share their true intentions, especially on sensitive issues.
Prediction markets, on the other hand, operate like stock exchanges where participants invest in future events based on their own analysis, knowledge or even rumours at the time. Unlike polls, a punter on a prediction market does not declare his personal preferences but takes a position on the most likely outcome in order to maximise his financial return. For example, someone might bet on the victory of a candidate they don't like, simply because they believe the circumstances are in their favour. This "utilitarian" bias guides participants to detach themselves from their personal opinions and focus on objective analyses or privileged information, often by consulting various sources such as economic trends, debates or campaigns on social networks.
Another notable difference: prediction markets allow individuals to bet on international events without having any direct link to the country or culture concerned. At election time, a French punter can bet on an election in the United States or vice versa, which introduces external perspectives that are often more objective and less influenced by domestic biases. In the case of Polymarket's bets, Americans are theoretically not allowed to use the platform for regulatory reasons regarding cryptos.
This globality can sometimes offer more neutral and better-informed predictions than polls, which, in turn, are limited to a local sample. Indeed, during Brexit, many UK polls (as well as financial markets) showed a strong Remain bias, while some international prediction markets, taking into account subtle signals and economic data, anticipated more of a Leave.
These distinctions show that, while polls and prediction markets can both provide valuable information, they fulfil different roles: polls capture trends in opinion, while prediction markets seek to assess the most likely outcome, often by incorporating more varied signals and minimising personal biases. This is why, during major events, analysts often follow both to obtain a balanced overview.
Risks of manipulation? Crypto prediction markets present a real risk of manipulation, particularly by 'whales', investors with significant financial resources who are able to strongly influence the odds by placing large bets. This phenomenon can be seen in Polymarket, where a mysterious investor, "Fredi9999", nicknamed the "Trump whale", bet millions to tilt the odds in favour of Donald Trump in the 2024 presidential election, despite polls and probabilities on other prediction markets showing a close race with Kamala Harris. In October 2024, this investor placed more than $30 million on Trump winning, which boosted his perceived probability, creating a notable gap with traditional polls.
Whales can take advantage of the relatively limited liquidity of crypto prediction markets to influence odds for strategic purposes. For example, by influencing public perceptions of a candidate's chances, they can attempt to shape media coverage or plant doubts in the minds of punters and analysts. In the case of Polymarket, some analysts believe that these massive bets could be aimed at reinforcing the public perception of the viability of Trump's candidacy.
Others note that the "Fredi9999" account buys Trump shares on a fairly regular basis, mainly when his odds become more advantageous. This behaviour is more akin to that of a wealthy investor who believes Trump will win - or at least believes his rating is undervalued.
A similar phenomenon occurred during the 2012 US presidential election. An investor nicknamed the "Romney Whale" placed a series of bets totalling almost $7 million on the victory of Mitt Romney, the Republican candidate, via the (now defunct) betting platform InTrade. This "whale" lost everything when President Barack Obama won his second term.
However, this vulnerability to manipulation calls for caution on the part of users of crypto prediction markets. The actions of the "whales" highlight the challenges these markets face in preserving their integrity and reliability, particularly in contexts where the financial and political stakes are considerable.
What regulation for prediction markets? The regulatory issues for crypto prediction markets are varied and mainly concern surveillance, market manipulation and the political implications of event contracts. In the United States, the CFTC (Commodity Futures Trading Commission) had long banned prediction contracts on political events, considering them to be forms of gambling that did not meet the criteria for authorised financial products. However, a September 2024 court ruling has allowed some platforms, such as Kalshi, to offer election contracts. However, the regulation of these markets remains unclear, and the CFTC could seek to redefine or restrict their use. Polymarket is currently unavailable to US users, and the company has paid a $1.4 million fine for offering it in the past.
Decentralised prediction markets are particularly vulnerable to manipulation by large investors (or "whales"), capable of injecting large sums to influence the probabilities of an outcome. On Polymarket, for example, a major investor tilted the odds in Trump's favour, which could influence not only public perceptions but also voter behaviour. The CFTC could intervene to limit such manipulations, relying on laws similar to those governing traditional financial products, which could include sanctions or bans.
In Europe, the regulation of prediction markets varies considerably from country to country, and no common EU framework currently exists to oversee these platforms, whether crypto or not. The situation is complex, as prediction markets can be considered either financial instruments or games of chance, depending on their structure and use.
In many European countries, such as France, prediction markets can be equated with games of chance. This places them under the supervision of gambling regulators, such as the National Gaming Authority (ANJ), which imposes strict licensing, transparency and player protection requirements. In practice, few prediction markets have obtained the necessary authorisation to operate legally in these countries, and regulators are often reluctant to approve decentralised platforms that do not comply with traditional gambling regulations.
In other countries, notably the UK, prediction markets are sometimes tolerated as financial products, which places them under the supervision of the Financial Conduct Authority (FCA). The FCA applies strict anti-money laundering (AML) and know your customer (KYC) standards. Platforms such as Betfair operate under FCA regulation to offer prediction services on sporting and political events. That said, crypto prediction platforms, particularly those using decentralised contracts and tokens, are still in a grey area as they partially escape FCA control.
Decentralised crypto exchanges such as Polymarket, accessible in Europe, are not yet subject to strict regulation, although they may be influenced by the new European regulation on crypto-assets, MiCA (Markets in Crypto-Assets). This regulation, due to come into force in 2024-2025, will impose transparency and user protection standards for platforms operating with crypto-assets, possibly including prediction markets. However, MiCA does not yet specifically address event contracts, leaving the legality and regulation of these markets in the EU unclear.
Qualities and shortcomings of decentralised prediction markets Decentralised prediction markets have several advantages:
No middleman: Decentralised platforms operate without a central authority, offering more freedom and lower fees.Greater transparency: Thanks to the blockchain, every transaction is traceable, guaranteeing greater transparency in betting and results management. This also makes it easier to detect potential manipulation.Total freedom to create markets: Users can create markets on any topic, which is impossible on traditional centralised platforms.More accurate than traditional polls: The "wisdom of crowds" and financial incentives enable decentralised prediction markets to provide predictions that are often more reliable than traditional polls.However, they also have a number of drawbacks:
Lack of regulation: The lack of centralised control can lead to abuses, such as betting on ethically questionable subjects (e.g. the death of public figures).Manipulation by "whales": Large investors can influence the market, which is far less liquid than in traditional finance. This risks distorting predictions in relation to the reality of the situation.Risk of invalid bets: Without a central authority to check the validity of proposed events, some markets may lack clarity or it may be difficult to validate the results.Future prospects for decentralised prediction markets The future of decentralised prediction markets is promising, but their success depends on their ability to navigate suitable regulatory and governance environments, which is essential to prevent manipulation and enhance their credibility. With the rise of cryptocurrencies and platforms like Polymarket, expanding the user base could increase the accuracy of predictions by diversifying data sources and better harnessing the "wisdom of crowds".
This diversification is crucial: by incorporating global perspectives on subjects such as politics, science or finance, prediction markets could become strategic tools in decision-making.
However, their sustainability will depend on a balance between innovation and ethics, particularly to avoid abuse and respect user confidentiality. If this model succeeds, it could establish a standard for future platforms, attracting institutions and the general public. The outlook is therefore exciting, but relies on continued adaptation to legal and ethical standards to inspire confidence.
