Prediction markets may look fun and gambling, but they are actually human oracles.

By Stacy Muur, Substack

Compiled by: Shan Ouba, Golden Finance

Today’s focus is on the report released by Delphi (Markets Will Save the World). After reading this report, my view on prediction markets changed from “wow, this is just a new form of binary options” to “wow, this is really a big deal”. I can confidently say that this is some of the best work from the Delphi team and I can’t wait to share the highlights with you.

To better understand this article, you need to look at prediction markets from a macro perspective, not as an end user, or even as an investor.

What do artificial intelligence and prediction markets have in common?

ChatGPT and prediction markets have similar functionality: ChatGPT answers questions, while prediction markets predict events.

ChatGPT has access to all online knowledge, but cannot make unbiased judgments based on that knowledge. It knows the past and understands others’ views on future outcomes, but it remains passive.

The market can be viewed as a kind of general intelligence. While there are some key differences, the similarities between markets and AI models like GPT-4 are striking. Both rely on vast amounts of human-generated data, compressing vast troves of knowledge and reflecting our world back to us in implicit ways.

Yet we ask markets few questions that empower them to act. Instead, we often treat markets as unpredictable slot machines and judge them by the answers they provide.

Prediction markets actually offer something different from today’s AI tools. They are increasingly seen as a reliable source of information, acting as “truth machines”.

Collective Wisdom

Unlike AI models like ChatGPT that rely on developer expertise, Polymarket relies on the collective opinions of users. The key difference is where the trust comes from: ChatGPT requires trust in its developers, while Polymarket is based on consensus among participants, becoming a tool for evaluating collective knowledge.

There is growing interest in prediction markets because they reveal the truth through the predictions of their users. This is increasingly important today, especially as trust in institutions such as the media and government decreases, and prediction markets are gradually filling this gap.

A notable example of a prediction market is the 2024 presidential election.

In September 2023, Polymarket predicted that Joe Biden would drop out of the race at 22%, far ahead of traditional polls and expert opinions. By the summer of 2024, despite Biden's team's repeated public assurances that he would not drop out of the race, the prediction market continued to lower his odds. By July 2024, the prediction was confirmed - Biden officially announced his withdrawal from the race.

Another example is the changes in the prediction market during the Trump-Kamala Harris debate. Within 20 minutes of the start of the debate, market predictions began to shift in a direction that was unfavorable to Trump, affecting public opinion and social media discussions. This shows how prediction markets can correct bias and become an objective source of information.

Drivers and Challenges

78% of Polymarket’s volume in 2024 is election-related. Volume is expected to taper after November 5, but there could be a surge before the election, mirroring trends in 2020.

However, a significant challenge facing Polymarket and other prediction markets remains low liquidity.

Low liquidity leads to market inefficiencies, which in turn reduces market reliability.

For example, betting large amounts of money (like a $1 million bet on Kamala Harris to win) can cause prices to slide significantly. Lack of liquidity makes it difficult for large traders to enter or exit positions, increasing the likelihood of distortions such as biases. For example, many Polymarket users share similar political views, which can amplify related biases.

Another key issue is the operation of oracles such as UMA, which can delay dispute resolution even when the facts are clear. A well-known case involved a dispute involving Justin Bieber’s children, which was complicated by the platform’s technical rules.

Possible alternatives include moving to a centralized dispute resolution system or implementing AI oracles to enhance the user experience.

The zero-sum nature of prediction markets makes them less attractive to traders because one person’s gain means another person’s loss. This dynamic complicates profit potential.

User bias and liquidity issues also stand out. Notably, 95% of Polymarket’s volume comes from just 20 traders, making the market vulnerable to manipulation.

For example, a group of large traders tried to move the presidential election market in favor of Kamala Harris but lost $60,000 due to the quick response of other players.

Market Overview

With around 25 new platforms expected to launch in the next 6-12 months, competition will become even more intense.

Polymarket remains the market leader, while Azuro is trying to catch up by providing liquidity to other platforms. Interestingly, despite having its own token, Azuro still has difficulty competing with Polymarket, which does not have a token.

Drift is a Solana-based platform that launched in August 2024 and has attracted attention for its integration with perpetual DEX and lending platforms. It offers more than 30 types of staking collateral, helping to reduce opportunity costs through built-in yields.

The new platform emphasizes innovation, focusing on decentralization, enhanced user experience, and new gambling categories such as pop culture, games, and SocialFi. SocialFi has been featured on Polymarket through social gambling discussions. The use of artificial intelligence for dispute resolution also shows promise in improving decision accuracy.

Azuro and Drift offer unique approaches centered around decentralization and improved user interactions that could transform the market in the coming years.

Is the future Futarchy?

ChatGPT and prediction markets have similar functions: ChatGPT answers questions, while prediction markets predict events. However, they are both passive tools that cannot directly bring about real change. To have an impact, active decisions are required - AI requires AI agents, and prediction markets require Futarchy.

Futarchy is a concept proposed by economist Robin Hanson in 2000 that makes decisions through market mechanisms rather than voting. In 2023, developer Prophet implemented Futarchy in MetaDAO on the Solana blockchain.

Unlike Polymarket, where users bet on “what should happen,” MetaDAO is developing decision markets, where users bet on what they believe will happen.

Passive vs. Active Markets

In prediction markets, participants are passive. You sit on your couch, watch the game and bet on who will win. You have a "stake" in participating, but you cannot control the outcome of the game.

Decision markets are different. They are active. By speculating, you can actually influence the outcome. This dynamic allows participants to trade proposals and shape the outcome, thereby improving decision-making efficiency and promoting economic development.

Personal opinion

Markets aren’t just for speculation or chasing yield in a zero-sum game. They also provide valuable data.

Curious about what’s trending on platform X (formerly Twitter) tomorrow? Check out the charts of those top gainers — this pattern has been seen multiple times with projects like memecoin, Solana, and Sui. Price movement often precedes discussion, and discussion amplifies price movement.

Sometimes, you just have to look at the charts and let them do the talking.