Pour cold water on the prediction market

The prediction market is undoubtedly one of the most attention-grabbing sectors in the encryption industry. The leader project Polymarket has accumulated a volume of over 36 billion dollars and recently completed a strategic funding round with a valuation of 9 billion dollars. Meanwhile, platforms including Kalshi (valued at 11 billion dollars) have also received significant capital injections.

▲ Source: Dune

However, behind the continuous influx of capital and the impressive data growth, we find that the prediction market, as a trading product, still faces many issues.

In this article, the author attempts to set aside mainstream optimism and provide some observations from different perspectives.

01

Predictions are based on events - events are essentially discontinuous and non-reproducible. Compared to the prices of assets like stocks and forex that change over time, prediction markets rely on limited and discrete events in the real world. In relation to trading, it is low frequency.

In the real world, there are very few events that have widespread attention, clear outcomes, and settle within a reasonable period—presidential elections every four years, the World Cup every four years, the Oscars once a year, and so on.

Most social, political, economic, and technological events do not have a sustained trading demand. The number of such events is limited each year, and their frequency is too low to establish a stable trading ecosystem.

In other words, the low frequency of prediction markets is not something that can be easily changed by product design or incentive mechanisms. This underlying characteristic determines that, in the absence of significant events, the volume of prediction markets will inevitably not remain high.

02

Prediction markets do not have fundamentals like the stock market: the value of the stock market comes from the intrinsic value of companies, including their future cash flows, profitability, assets, etc. Prediction markets ultimately point to an outcome, relying on users' “interest in the outcome of the event itself.”

(Of course, here we discuss the original intention of the product, excluding objective arbitrage and speculation factors, etc.; even in the stock market, there are many speculators who may not care about the nature of the underlying assets.)

In this context, the amount of money people are willing to bet is significantly positively correlated with the importance of the event, the market's attention, and the time cycle: scarce and highly watched events such as the finals and presidential elections attract a large amount of capital and attention.

Naturally, an ordinary fan is more likely to care about the outcome of the annual finals and place heavy bets on it, rather than showing such behavior during the regular season.

On Polymarket, the 2024 presidential election event accounts for over 70% of the platform's total OI. Meanwhile, the vast majority of events have long been in a state of low liquidity and high bid-ask spreads. From this perspective, the scale of the prediction market is difficult to expand exponentially.

03

The prediction market itself has a gambling nature, but it is difficult to generate retention and expansion like gambling.

We all know that the true mechanism of gambling addiction lies in instant feedback—slot machines every few seconds, Texas Hold'em every few minutes, and contract and memecoin trading changing rapidly every second.

The feedback cycle for the prediction market is very long, with most events taking weeks to months to settle. If it is a fast-feedback event, it may not necessarily be interesting enough to warrant a heavy bet.

Immediate positive feedback significantly increases the frequency of dopamine release, reinforcing user habits. Delayed feedback fails to establish stable user retention.

04

In some types of events, there is a high degree of information asymmetry between participants.

For competitive sports events, in addition to the paper strength of the teams, a large extent also relies on the athletes' performance on the spot, thus there is a considerable degree of uncertainty.

However, for political events, it involves internal information, channels, connections, and other black-box processes. Insiders have a significant information advantage, making their betting much more certain.

It's like the vote counting process in an election, internal polling, and the organization of key areas; external participants find it very difficult to obtain this information. Currently, there has not been a clear definition from regulatory bodies regarding “insider trading” in the prediction market, which remains a gray area.

Overall, in such events, the party at an information disadvantage can easily become the exit liquidity.

05

Due to the ambiguity of language and definitions, it is also difficult for the events in the prediction market to be completely objective.

For example: “Will the Russia-Ukraine conflict ceasefire in 2025” depends on which statistical criteria are used; “Will the cryptocurrency ETF be approved at a certain time” involves complete approval, partial approval, or conditional approval, etc. This raises the issue of “social consensus” - in a situation where both sides are evenly matched, the defeated party will not admit defeat willingly.

Such ambiguity requires the platform to establish a dispute resolution mechanism. Once the prediction market touches on linguistic ambiguity and dispute resolution, it cannot fully rely on automation or objectivity, leaving room for human manipulation and corruption.

06

The main value proposition of prediction markets in the market is “collective intelligence,” which means that, in contrast to the low trust in media and mainstream discourse, prediction markets can aggregate the highest quality information globally, thus achieving collective consensus.

However, before prediction markets achieve massive adoption, this kind of “information sampling” is bound to be one-sided, and the sample is not diverse enough. The user base of the prediction market platform may be highly homogeneous.

For example, in the early stages of the prediction market, it is certainly a platform primarily composed of cryptocurrency users, whose views on political, social, and economic events may be highly convergent, thus forming an information cocoon.

In this case, the market reflects the collective bias of a specific circle, and there is still quite a distance from “collective intelligence.”

Conclusion

The core of this article is not to sing the blues about the prediction market, but rather to hope that we can remain calm in the face of heightened FOMO emotions, especially after experiencing the ups and downs of popular narratives like ZK and GameFi.

Over-reliance on special events such as elections, short-term sentiments on social media, and airdrop incentives often amplifies the superficial data, which is still insufficient to support judgments about long-term growth.

Nevertheless, from the perspective of user education and user acquisition, prediction markets still hold an important position in the next three to five years. Similar to on-chain yield savings products, they have an intuitive product form and a lower learning cost, making them more likely to attract users outside the community into the encryption ecosystem compared to on-chain trading protocols. Based on this, prediction markets are likely to further develop and to some extent become entry-level products in the encryption industry.

Future prediction markets may also occupy certain verticals, such as sports and politics. They will continue to exist and expand, but in the short term, they do not have the fundamental conditions for exponential growth. We should think about investing in prediction markets with a cautiously optimistic perspective.