Prediction Markets

Prediction markets are contract-based markets that track the outcome of specific events.

Traders buy shares in a market (priced 0 < x < 100), and depending on the event’s outcome, those shares are either worth 0 or 100.

  1. A market is created to determine if the price of Ethereum is >= 3500 at the end of October.

  2. YES, shares are selling for 60c, implying a 60% probability that ETH >= 3500 on the settlement date.

  3. Trader X buys 100 YES shares for $60, whereas Trader Y buys 100 NO shares for $40.

  4. At the end of October, ETH is 3700. Trader X redeems his 100 shares for $100 (~1.66x), and Trader Y is zeroed out.

The only constraints on a prediction market’s existence are a willing external party to create the market and traders willing to purchase contracts for both sides.

There are three different types of prediction markets:

  • Binary: These markets are YES/NO, without a possibility for a third answer. The market above is binary.

  • Categorical: These markets include multiple outcomes. A simple example is a prediction market on the first crypto protocol to airdrop. The market will include a predetermined set of outcomes, and each outcome will have ever-changing, varying probabilities assigned.

  • Continuous: These markets handle events with many different possible settlements. Predicting the close of BTC on a given date would be a continuous market, as there are infinitely possible prices at which BTC could close. Due to this, continuous markets typically integrate predetermined constraints, such as >= 70,000, 60,000 < X < 70,000, and <= 60,000.

There are several different real-world practical applications for prediction markets:

  • Political: Political markets are arguably the reason prediction markets start seeing accelerated growth and volume. The majority of volume stems from presidential elections and senate/house races. The U.S. presidential election alone has 128.5M outstanding contracts, with more than five months left until the election.

  • Economic: Economic markets are normally continuous and consist of different financial indicators, such as the CPI rate, unemployment/housing figures, and GDP growth.

  • Corporate: Corporate markets are typically used to predict the sales of a certain product or merger. However, they can also be used in less sophisticated ways, such as “What is the probability Delta Airlines has a commercial during the Super Bowl?”

  • Entertainment: Entertainment markets are prevalent because sportsbooks under the hood are effectively prediction markets with a house edge. These markets can commonly be arbitraged, as discussed here. In a nuanced fashion, prediction markets are inefficient, so there is typically a disparity between sportsbook offerings and probability assigned to prediction markets.

  • Arbitrary: Arbitrage prediction markets are effectively any market not categorized under the above four.

Accurate Probability

Probability theory i

Leave your comment...

Nice - very detailed write-up!

Icymi - Omen on Gnosis Chain has integrated AI Agents deployed on the OLAS tech stack. That's been ongoing for about a year and there's been a new release that makes Prediction Trader Agents accessible to anyone (App is called Pearl and basically a UI that eliminates the difficulties of setting up the Agent). LLMs are being leveraged for Trader Agents to create new markets, eg some Agents browse the web for noteworthy news and then start making markets. Other Agents are then reviewing news, data, etc to derive their expected probabilities for these events. Agent "knowledge" can be customised to gain an edge.

I see that there's an open bounty on the OLAS website to hook up with Polymarket.

The AI scenario you are describing is playing out in front of our eyes and anyone can test it even with zero coding skills (thanks to Pearl).