Beyond Oracle Machine: The Future Evolution and Application Scenarios of Prediction Market

Author: NotDegenAmy, Derek, yiwei Source: Ocular Translation: Shanooba, Golden Finance

Background

Now, when people think of intelligence, the first thing that may come to mind is the LLM model, such as GPT, Claude, Llama, etc.

But in fact, the market itself may be the best form of general intelligence, as it is essentially a synthesis of all actions. Artificial intelligence itself is trained based on a large amount of information generated by the public. However, artificial intelligence is passive and requires goals and instructions (at least before we enter the real agent world). In order to better utilize and express market intelligence, we need something that can capture the constantly changing thoughts of the crowd, something forward-looking.

Enter the prediction market. The prediction market is a platform where participants can buy and sell contracts based on their beliefs about potential outcomes of future events. These events can be political (such as election results) and economic (such as interest rate changes), as well as entertainment and sports (such as match results).

This concept is not new - people believed that an early form of predicting the market existed more than 500 years ago, mainly used to predict political outcomes.

In the early 21st century, prediction markets such as Intrade and Betfair began to emerge, especially during the US presidential elections. However, these were centralized platforms, often limited by geographical restrictions, regulatory constraints, and the need for trusted intermediaries to manage funds and settle bets. This affected their development and expansion capabilities.

• For example, Intrade was forced to close in 2013, and the U.S. Commodity Futures Trading Commission sued Intrade, seeking to prohibit Americans from using the website, claiming that it illegally sold futures contracts, leading to a sharp decline in its user base.

In the late 2010s, with the rise of blockchain technology, the prediction market resurfaced with greater strength than before. This time, platforms utilize blockchain to create decentralized and global platforms, which have several advantages compared to their centralized counterparts:

However, the prediction market has never become mainstream. Until this year. Due to the 2024 US presidential election, people have regained interest and attention in this new form of market intelligence. In this article, we will delve into the mechanism of web3 prediction markets, specifically covering 1) use cases and current landscape of prediction markets; 2) case study of Polymarket; and 3) future trends.

1) Use Cases and Current Situation

In addition to providing users with opportunities to profit from their views/predictions of the future, prediction markets have other use cases as described below:

DeFiLlama estimates that the total locked value (TVL) of the current web3 prediction market is approximately $140 million. This is lower than the pre-election high of $545 million.

In October 2024, the monthly total revenue of all prediction markets is estimated to be about $750,000, and the annual calculation is about $9,000,000.

The main players in this field are Polymarket, Azuro and Drift (BET). Among these three companies, the total betting amount in the third quarter of 2024 increased by more than 550%, reaching 3.1 billion US dollars, while it was only 463.3 million US dollars in the second quarter of 2024 (see the figure below).

• (Source: CoinGecko and ocularvc) *

In November 2024, there were approximately 290,000 monthly active traders on Polymarket, with over 300,000 accounts opened on the platform. According to Polymarket’s leaderboard, the most active trader on Polymarket generated a trading volume of $397 million, while the highest-earning trader made over $22 million.

2a) Polymarket - Case Study

So how does the prediction market work? This can be divided into three subcategories, namely its features, fee model, and dispute resolution process. As Polymarket currently dominates, we will refer to its operation mode.

Features

• Market - Typically associated with real-world events. Polymarket has a wide range of markets, from politics and cryptocurrencies to popular culture and weather outcomes, while others may choose to focus on specific areas such as sports betting.
• Result - can be:

• Multi-outcome markets are primarily characterized by binary outcomes. Some markets have multiple outcomes, but for each option (e.g. candidate), this will be a binary trade (see example below). Each side of the trade has a probability/price, and after the event, a portion of the correct outcome can be exchanged for $1, while a portion of the erroneous outcome is worth $0.
• Binary, for example, whether an event will occur ‘yes’ or ‘no’;
• Various results, such as predicting which candidate will win a multi-candidate election or which team will win the championship; or
• Continuous, that is, predicting a series of values ​​(such as stock prices or voting percentages).

Odds- There are mainly two ways to determine the price/odds of the market:

• One is through an order book system similar to the stock market. Participants submit buy and sell orders; prices are determined by matching these orders.
• The second is through a system based on Automated Market Maker (AMM). In this system, each buy and sell will be accepted. The price is automatically determined and adjusted based on an algorithm/mathematical formula that tracks the trading volume.

• Polymarket primarily utilizes an order book-based system.
• When the market was first established, there were no stocks and no predetermined prices or odds. Those interested in purchasing ‘yes’ or ‘no’ stocks could place limit orders at the price they were willing to pay.
• When the bids for both ‘yes’ and ‘no’ are 1.00 US dollars, the order will be ‘matched’, and the 1.00 US dollars will be converted into 1 share of ‘yes’ and 1 share of ‘no’, respectively owned by their respective buyers.
• For example, if you place a ‘yes’ limit order at $0.60, when someone places a ‘no’ order at $0.40, the order will be matched. This will become the initial market price.
• Subsequently, the price shown on Polymarket is the midpoint of the bid-ask spread in the order book—unless the spread exceeds $0.10, in which case the last traded price is used.
• As shown in the market below, 37% of the probability/price is the midpoint between a buying price of 34¢ and a selling price of 40¢. If the buying and selling price difference is greater than 10¢, the probability/price is displayed as the last trading price.

• Payment - Polymarket operates on the Polygon blockchain, and users place orders using USDC. Polymarket recently partnered with MoonPay, allowing users to purchase USDC with fiat currency.
• Order - Polymarket offers market, limit, and AMM orders. However, there are currently no leverage options. After placing an order, Polymarket allows users to trade the shares they own before the event described in the market actually occurs.

• Referring to the example above, let’s assume that we purchased Ethereum ‘yes’ shares at a 37% odds when the price of Ethereum reached $3,000. If the odds increase after we place the bet, we can decide to sell our shares at a higher price and lock in profits before the actual event occurs/the deadline arrives. Of course, if the odds decrease after we place the bet, we can also choose to sell the shares at a loss.

Fees

There are mainly two types of fees charged by decentralized prediction markets:

• Trading fees, which are the small amount charged by the platform for each transaction.
• Deposit/Withdrawal Fees, which are small fees charged each time fiat/crypto currency enters or leaves the platform.

Polymarket currently does not charge any transaction fees. However, they will deduct a fee of 2% from the net winnings of winning bets. Polymarket does not treat this fee as revenue, but instead uses it to reward liquidity providers (as part of their liquidity rewards program) and pay gas fees. Polymarket also does not charge any deposit/withdrawal fees.

• When asked about the pricing strategy, Polymarket founder Shayne Coplan said in July 2024: “We are currently focusing on expanding the market and providing the best user experience. We will focus on monetization later.”

Controversy

To solve market issues after the event, platforms usually rely on 1) oracle; and 2) community voting.

• In the case of Polymarket, they rely on both the Universal Market Access (UMA) optimistic oracle and the Data Verification Mechanism (DVM) to implement these two strategies. The simplified chart below illustrates the market resolution process:

2b) Risks and limitations

So far, there are different opinions on the effectiveness of Polymarket and the entire prediction market as a source of market intelligence. We outline these different views below.

Ocular believes that predicting the market as an intelligence source needs to meet three conditions:

• First, when the three factors of incentive, ability, and opportunity are matched, the prediction market can play its best role.

• Regarding incentives: Individuals you survey need to have vested interests in a certain issue. This could be because it affects their daily lives, their other investments, or it is a hot topic on social media that they want to be involved in. The key is that they need to have a sense of involvement/participation in this issue so that they can participate in the market.
• Ability: The public needs enough information to form their own opinions. It cannot be too niche or require deep technical knowledge, because in these cases, the public may not become smarter, and the results are also difficult to be convincing.
• About time: Although the market can be launched instantly, you need time to gather public opinions and allow the market to react to new information. Therefore, this is not suitable for time-sensitive decisions.

• Second, there needs to be sufficient liquidity. In the end, prediction markets can only work best when they can truly leverage collective wisdom. This means that the market needs to reach a certain scale, in terms of both the number of individuals participating and the amount of bets, in order to be meaningful and practical.
• Third, it should not be used in isolation. Predictive markets are typically driven by public information, such as news or mentions on social media. Others may have private data sources that may not be fully reflected in the trades, so seeking them out for a different perspective may be helpful.

3) Future Trends

We see a trend of expanding use cases for prediction markets.

The prediction market may be suitable for decision-making markets. Users vote to determine ‘what the result should be,’ rather than ‘what the result will be.’ Although prediction markets provide valuable insights, they are more passive. People vote and wait for the results to be announced, usually with little impact on the outcome. On the other hand, decision markets are more proactive and more suitable for governance.

Case Study: MetaDAO ($META)

MetaDAO is a project established in January 2023, invested by Colosseum and Paradigm. Its core product is Futarchy, which proposes governance proposals for voting and simultaneously initiates 2 conditional trading markets:

• When the market believes that the proposal will increase the value of the token beyond a certain threshold, they can raise the price of the ‘pass’ token. Conversely, they will raise the price of the ‘fail’ token.
• At the end of the voting, if the TWAP price of the Pass token is more than 3% higher than the TWAP price of the Fail token, the proposal will be passed and implemented; otherwise, the proposal fails and the market will return to its original state.

• For example, a proposal could be to hire a new CEO for the company. If the decision market indicates that hiring a CEO will significantly increase the value of the company’s stock, then the proposal will be approved, and the CEO will be hired.
• The general idea of the future system is as follows:

Futarchy was proposed by economist Robin Hanson in 2000. He proposed that Futarchy is a governance model that combines prediction markets with traditional voting systems. In the Futarchy governance model, decisions are based on dispersed market predictions. Participants do not directly vote on policies, but on measurable objectives, such as economic growth. Then, the prediction market will predict how these objectives will be affected by the proposed policy. The policy that is expected to achieve the best results as determined by the market will be implemented. This approach uses collective intelligence and financial incentives to guide decision-making.

In the following scenarios, Futarchy is superior to Polymarket’s binary betting model:

1. Decisions are goals, not just predictions.
2. Complex, long-term effects must be considered.
3. The incentive structure needs to be aligned with social or organizational goals.

Currently, in addition to its own DAO governance, MetaDAO is also collaborating with six projects for decision-making. MetaDAO has shown some early signs of success in decision-making, from preventing whales from buying $META at a steep discount to redirecting company resources away from new plans that owners believe will distract attention.

However, MetaDAO is still in its early stages. There are several key limitations to MetaDAO’s model, and its large-scale application remains to be seen:

• Oracle - Not every project has a token, and not every decision can be precisely measured with metrics. It may also be difficult to judge the impact on indicators of relatively small decisions. Liquidity- The concentration of users and wallets can lead to skewed results. The user experience can also be too technical to be accepted by more people, which can limit the number of voters.
• Applicability - Powerful individuals may not want to transfer decision-making power to the market. Participants must also be an informed group. How do we ensure that users vote according to the long-term interests of the company?

In addition to MetaDAO, prediction and decision markets are also constantly developing and innovating:

• New Market - This year, due to coinciding with the Olympics and the U.S. presidential election, the prediction market has attracted a lot of speculation and discussion. The challenge facing the industry is to maintain people’s interest in the prediction market even after these cyclical events end. Platforms can consider expanding their business beyond seasonal/one-time events and exploring different consumer groups and categories, such as popular culture and social media, targeting female and young participants.
• Advanced Oracle - To create a new market, it may be necessary to build a new oracle to fetch and obtain relevant data for pricing the new market. Overlay is such a project that is seeking to build oracles for unique markets, such as “Counter-Strike” skins and AI index.
• Efficient Arbitrage - Many platforms use order book-based systems (i.e., user-placed buy and sell orders) to price the market. Due to variations in user behavior and overall liquidity across markets and platforms, there are arbitrage opportunities both within and between platforms.

• Taking the odds on the US presidential election on October 28th on Polymarket and Kalshi as an example. Users can purchase a contract for Kamala’s victory on Polymarket (with odds of 33%) and a contract for Trump’s victory on Kalshi (with odds of 62%), with a total cost of 95 cents. Given that these two events are mutually exclusive, users will receive a payout of 1 dollar regardless of the outcome, creating a 5% arbitrage opportunity.

• Ensure that there are no duplicate or similarly worded listings on the platform;

• Incorporate the odds of other platforms into its pricing model; and/or
• Establish a trading robot to take advantage of such opportunities and minimize cross-platform propagation as much as possible.
• To effectively capitalize on these arbitrage opportunities, platforms may want to:

• Capital Efficiency - Currently, in order to execute transactions on most platforms, users need to have capital on hand, and once a transaction is executed, the capital will be locked on the platform. To improve capital efficiency, platforms can consider:

• Introducing leveraged products so that users can place leveraged bets;
• Allow trading with profitable stablecoins/tokens;
• Tokenizing user positions and allowing the token to be traded on other platforms; and/or
• Design a lending protocol that allows users to borrow and borrow based on their positions.

Anti-Manipulation - To address price manipulation, platforms may consider setting betting limits or limiting the number of accounts an individual can open. In small, illiquid markets, the use of cross-examination is also effective: for example, ask “what do you believe and what do you think others believe” and then compare the results.

• AI Participation - In order to make the resolution process (especially for simple markets) more efficient, the platform can consider using artificial intelligence/large language models to acquire and verify the information needed for resolving markets. AI agents can also be trained to research the truth more effectively and participate in future voting.

Conclusion

Ocular is closely monitoring the prediction market sector and its development.

Now, prediction markets are used to generate income, hedge positions, attract community participation, and measure market sentiment. Looking ahead, it can be used for collective decision-making/governance and more interesting areas:

• Replicability of scientific papers;
• Summarize private information (internal to the company);
• Predicting the success of drug trials;
• Obtain estimates on subjective issues (pre-release testing); and
• Reshaping news media - involving reporters and analysts.

Despite the inefficiencies/restrictions that need to be addressed in the industry (such as inconsistent liquidity, regulatory uncertainty, and oracle issues), we remain optimistic about the long-term prospects of the industry, especially in conjunction with artificial intelligence.