Future Times

Inspiration

We want to quantitatively predict the future. Gathering the necessary data is usually a significant hurdle.

Polymarket has an API with a generous rate limit and free websockets and is much less regulated than other markets, so we have access to much more interesting behaviours visible in market data, such as insider trades. We therefore decided to take advantage of this.

What it does

It is a news app that provides news for future events. These events are predicted by assigning a probability to events on Polymarket, calculated by taking into account Polymarket sentiment and anomalous trades on its order book.

These events are displayed as LLM-generated news articles with the headline being the expected outcome, with RAG technologies providing links to similar news articles in other sites. Our predicted likelihood is also displayed.

Users can also view reports of which trades are flagged as anomalous, as well as the degree of their influence on the corresponding event probability.

How we built it

Built with next.js frontend and express.js backend. We maintain an in-memory order book of all events of interest, and get trade data via the public Polymarkets websockets. While it is running, it will constantly watch for anomalous trades.

Challenges we ran into

Initial ideas of detecting insider trading based on past trader performance were unfeasible as trade data is anonymised.

Accomplishments that we're proud of

This is our first dive into large-scala data processing to automate decision making and produce tangible results. Producing a working and stable project in this new area for us is a proud achievement.

What's next for Future Times

Potential new features: Backtester, trading decision recommendations, extension of the probability function to use more complex markers in trade data. The backend itself could be used as a standalone prediction market data pipeline, opening new possibilities.

Built With

• express.js
• javascript
• next.js
• polymarkets