Beyond Polymarket: Exploring Opportunities in Prediction Market Design
Understanding Prediction Markets
Prediction markets are essentially open markets where participants trade to predict the outcomes of specific events. These markets operate similarly to free market economies, with market prices adjusted based on the collective wisdom of participants. Prediction markets allow users to trade the probabilities of certain events occurring, and the final market price reflects the expected likelihood of these events.
By definition, a prediction market is "a market created for trading the outcomes of events. Market prices can reflect the public's perception of the probabilities of events occurring." While this definition summarizes the basic concept, the depth and complexity of prediction markets go far beyond this, warranting further exploration.
The Key Role of Openness
The openness of prediction markets is one of their most important features. Unlike traditional betting, where odds are set by a bookmaker based on specific formulas, prediction markets start with the same odds. As participants trade based on their own knowledge and insights, the market naturally adjusts prices to reflect the most likely outcomes.
To illustrate how prediction markets operate, consider a hypothetical example of the FIFA World Cup final in December 2022, possibly between Argentina and England. Based on existing data, a centralized bookmaker might set the odds at a 67% probability of Argentina winning and 33% for England.
In contrast, prediction markets do not require a centralized bookmaker. Participants can create a market by posing a question like "Who will win the FIFA World Cup final?" and listing possible outcomes such as "Argentina" or "England." This setup is known as a binary prediction market.
In our example, there would be two outcome tokens available for trading:
ARGWIN (Argentina wins)
ENGWIN (England wins)
These tokens start trading at the same price, for example, 50/50. As participants buy tokens based on their expectations, prices fluctuate according to supply and demand. If more people buy "ARGWIN," its price will rise while "ENGWIN" will fall. Over time, the market will self-adjust, and the token prices will reflect the most likely outcomes, potentially aligning with the bookmaker's set odds of 67/33.
Thus, prediction markets can achieve accurate predictions without specialized forecasters or data analysts. Most participants only engage in predictions when they have some insight or information about the possible outcomes.
Prediction Markets as Derivative Markets
Prediction markets can also be viewed as derivative markets. Since these markets are essentially information processors, they can be designed within an information theory framework, making prediction markets particularly suited to this model.
Prediction markets, also known as betting markets, information markets, decision markets, creative futures, or event derivatives, allow participants to trade contracts based on the outcomes of future unknown events. The market prices formed by these contracts can be seen as the collective predictions of market participants. If these contracts are linked to the prices of certain assets, prediction markets effectively become derivative markets.
Advantages of Prediction Markets as Derivative Markets:
No underlying asset required: These markets can operate without the need for an underlying asset. Only an oracle to introduce underlying asset information and currency for collateral are needed to establish such markets.
Automated Market Makers (AMM): Implementing automated market makers in prediction markets is relatively straightforward. Research on prediction markets has played a key role in the development of AMM algorithms.
Versatility: By designing appropriate predictive events, prediction markets can offer universal products.
Isomorphism with European options: Prediction markets have an isomorphic relationship with European options, allowing option pricing models to be transferred to prediction markets.
Capital efficiency: Prediction markets are highly capital efficient, often more so than traditional betting markets.
No risk of short squeezes: In prediction markets, participants' liabilities are limited to their collateralized assets, eliminating the risk of short squeezes.
Disadvantages of Prediction Markets as Derivative Markets:
Risk for liquidity providers: Liquidity providers hold positions and face high risks, especially during black swan events. However, this may be acceptable for risk-neutral investors.
Novelty and learning curve: Prediction markets are a relatively new concept, and participants may need time to fully understand their mechanisms. However, novelty is a common feature in the blockchain space.
Unknown risks: Like any new design, there may be undiscovered drawbacks.
Mechanisms: CDA and LMSR
Prediction markets are specialized financial markets where participants trade contracts based on the outcomes of future events (such as political elections, sports results, or economic indicators). The prices of these contracts reflect the collective beliefs of market participants regarding the likelihood of these events occurring. The two main mechanisms supporting the operation of prediction markets are Continuous Double Auctions (CDA) and Logarithmic Market Scoring Rules (LMSR). Each mechanism has its unique advantages while facing specific challenges in liquidity and price accuracy. This article explores the complexities of these mechanisms, their applications in prediction markets, and their relationship with automated market makers (AMM).
Continuous Double Auctions (CDA)
Continuous Double Auctions (CDA) are one of the most commonly used mechanisms in financial markets, including prediction markets. In a CDA, traders interact by placing buy (bids) and sell (asks) orders directly into an order book. The order book is the core part of the CDA mechanism, listing all outstanding orders, with bids on one side and asks on the other. When a bid matches an ask, a trade occurs, executed at the matched price. The dynamics of the CDA mechanism can be described using S-shaped functions for bids and asks. The definition of the S-shaped function is:
Here, PPP represents the price level. The bid function gradually decreases as the price rises, while the ask function increases, forming a natural equilibrium point where the two curves intersect. This intersection point represents the price at which trades occur. The S-shaped function is used to simulate the gradual change in order quantities as prices deviate from the center value.
A key feature of CDA is its reliance on direct interaction between traders to facilitate price discovery. Traders can place orders at any time, and these orders remain in the order book until matched with opposing orders. The flexibility of CDA allows traders to set their desired prices, enabling effective price discovery in highly liquid markets. However, in markets with fewer participants, this reliance on direct interaction can become a limitation. In illiquid markets, CDA may encounter low liquidity issues due to insufficient traders to quickly match orders, leading to wider bid-ask spreads. This can reduce market efficiency and make accurate price predictions more difficult.
In the context of prediction markets, the CDA mechanism has been widely applied due to its simplicity and ability to facilitate direct trading. However, the low liquidity issues caused by a limited number of participants in prediction markets have prompted exploration of alternative mechanisms such as LMSR.
Logarithmic Market Scoring Rules (LMSR)
Logarithmic Market Scoring Rules (LMSR) are a specially designed automated market maker (AMM) mechanism aimed at addressing common liquidity issues in prediction markets. Unlike CDA, where trades are conducted directly between participants, LMSR involves a central automated market maker that acts as the counterparty to all trades. This market maker continuously provides buy and sell quotes and uses logarithmic scoring rules to calculate these quotes, adjusting prices based on the total amount of uncombined contracts.
The LMSR mechanism can be modeled using logarithmic functions for price adjustments and logistic functions for liquidity. The logarithmic function for price adjustments is represented as:
Where TTT represents the number of trades. This function reflects that as the number of trades increases, prices rise at a decreasing rate, preventing prices from becoming too extreme. Liquidity can be modeled using a logistic function:
This function shows how liquidity changes with the number of trades, peaking at a certain trading volume before gradually decreasing.
A significant advantage of LMSR is its ability to provide constant liquidity, ensuring that traders can execute trades at any time without waiting for matching orders from other participants. LMSR achieves this by automatically adjusting prices as more contracts are bought or sold. The price adjustments are logarithmic, meaning that as the number of contracts favoring a particular outcome increases, the rate at which that outcome's price rises gradually slows. This mechanism prevents prices from becoming too extreme, stabilizing the market even in the case of large one-sided trades.
LMSR is particularly well-suited for prediction markets because it mitigates the risks associated with low liquidity. In markets with fewer participants, LMSR ensures that trades can proceed smoothly, with prices reflecting the collective sentiment of the market, even when active traders are few. However, this also means that market makers may face potential losses, as they may need to subsidize trades to maintain liquidity. Nevertheless, the design of LMSR ensures that these losses are capped, making it a sustainable mechanism for market owners.
Ken Kittlitz, the Chief Technology Officer of Consensus Point, emphasizes the practical benefits of using LMSR in prediction markets. He notes that the presence of automated market makers "has a huge impact on the success of the market" because it provides stable liquidity and simplifies the trading process for participants. By ensuring that there are always buy and sell orders across various price ranges, LMSR makes the market more accessible and intuitive, potentially leading to higher participation and, consequently, more accurate predictions. Comparing CDA and LMSR in Prediction Markets
Although both Continuous Double Auctions (CDA) and Logarithmic Market Scoring Rules (LMSR) mechanisms are used in prediction markets, they serve different purposes and are best suited for different market conditions. CDA excels in high liquidity markets, where there are enough participants to ensure regular matching of buy and sell orders. In such environments, CDA can facilitate effective price discovery, allowing the market to reflect the true collective beliefs of participants. However, in lower liquidity markets, CDA's reliance on direct interactions among traders may lead to inefficiencies, such as wider bid-ask spreads and inaccurate price predictions.
On the other hand, LMSR performs exceptionally well in environments where liquidity becomes an issue. Its automated market-making functionality ensures that trades can occur at any time, regardless of the number of participants. This continuous provision of liquidity makes LMSR particularly valuable in prediction markets, especially when participation may be intermittent or limited. The ability of LMSR to dynamically adjust prices based on trading volume also helps stabilize the market, preventing extreme price fluctuations, which is crucial for ensuring the reliability of market predictions.
Automated Market Makers (AMM)
Automated Market Makers (AMM), such as LMSR, play a key role in maintaining liquidity, especially in markets that may suffer from liquidity issues due to low trading volumes. In prediction markets, the number of participants can fluctuate significantly, and the presence of AMM ensures that the market remains functional, with prices continuously reflecting the collective sentiment of traders.
AMM sets prices and automatically provides trading using algorithms. In the case of LMSR, this algorithm is based on logarithmic functions, adjusting prices according to changes in trading volume. This continuous adjustment helps prevent the market from becoming overly biased toward specific outcomes, ensuring that prices remain within reasonable ranges. By providing this stabilizing effect, AMMs like LMSR enable prediction markets to operate effectively even with fewer participants.
Classification of Prediction Markets
Prediction markets can take various forms, each suited for different scenarios:
Binary markets: Involve two possible outcomes, such as "yes" or "no." For example, the FIFA World Cup example is a typical binary market.
Categorization markets: Similar to binary markets but with more than two options. For instance, predicting the winner of a tournament with multiple teams competing.
Scalar (interval) markets: Predict outcomes within a specific range, such as forecasting the future price of an asset. Participants are rewarded based on how closely their predictions align with the actual results.
Composite markets: The most complex form, where users create multi-layered predictions by combining multiple prediction markets.
Categorization Markets and Scalar Markets
In a categorization market, suppose we want to predict the winner of the FIFA World Cup after the quarter-finals, with eight teams remaining, and each outcome token might start at a price of 0.125 ZTG. If you accurately predict the winner early before the market closes, you could achieve significant profits.
In a scalar market, suppose we predict the price of Polkadot tokens (DOT) at the end of Q3 2022. Participants can predict any price within a set range (e.g., $0 to $20), and their rewards will depend on how closely their predictions match the actual price.
Composite Markets
Composite prediction markets allow for more complex predictions by combining multiple prediction markets. For example, predicting the success of a new iPhone release may involve multiple variables, such as color options, included accessories, and pricing. By combining these factors, participants can generate more accurate predictions about the product's success.
Composite markets are particularly useful in scenarios like weather insurance, where multiple variables influence outcomes. A dedicated article on the complexities of composite prediction markets will further explore this topic.
Comparison of Prediction Markets and Traditional Polling
Prediction markets have unique advantages over traditional polling methods. Prediction markets encourage accurate predictions through financial incentives rather than relying on labor-intensive surveys. The natural dynamics of the market ensure that overpriced shares are corrected by participants purchasing undervalued shares, providing more reliable data.
Conclusion
Prediction markets are a powerful tool for forecasting a variety of outcomes, from sports events and asset prices to political decisions and weather events. Participants with valuable insights are incentivized to engage and correct market imbalances, while those with less information are naturally discouraged from taking significant risks.
The goal of any prediction market platform should be to create a user-friendly environment that attracts liquidity and provides quick responses, ensuring that the creation and participation in prediction markets become straightforward. Decentralization and permissionless participation further enhance the platform's potential, allowing users to discover valuable data about the world around us. Continuous Double Auctions (CDA) and Logarithmic Market Scoring Rules (LMSR) are two distinct mechanisms that serve different needs in prediction markets. CDA facilitates direct interactions among traders and excels in high liquidity markets, while LMSR, as an automated market maker, ensures continuous liquidity and price stability, making it highly suitable for markets with fewer participants. Understanding the advantages and limitations of each mechanism is crucial for designing effective prediction markets that can accurately aggregate information and generate reliable forecasts. As the field of prediction markets continues to evolve, automated market makers like LMSR may become increasingly important in ensuring the robustness and accuracy of market predictions.
