Our new AI model predicts what your users will bet on next—updated in real time from their on-chain history and interactions.
We believe that like every other successful content surface on the internet, Prediction Market discovery should be algorithmic and personalized in real-time. Here are at least two reasons why:
First, prediction markets are content—and most Polymarket traffic are lurkers, not traders. Yet PM UIs are still stuck in the Craigslist era. If prediction market platforms want to fulfill their potential as the modern day oracles or places where people go for entertainment, then users should get recommendations of markets they’re susceptible to be interested in, just like on YouTube. Here personalization algorithms can make prediction markets more relevant.
Second, better discovery is key to solving the lack of liquidity in a long tail of markets—since liquidity distribution depends on whether traders are able to discover markets that they’re interested in in the first place. Here discovery algorithms can broaden the amount of markets that users see.
Enter ❜embed’s prediction market recommendation API
In order to solve these issues, we’re rolling out a new API endpoint for prediction markets soon, reach out to get access to the demo. Under the hood: an AI model trained on millions of past Polymarket trades that predicts which markets a given user (or wallet) is likely to bet on next. Think LLM-style autocomplete, but for sequences of trades.
In our internal benchmarks, recommendations from the API deliver ~10× higher relevance than simply showing the most-bet (“popular”) markets in a period. Moreover we’ve observed broader coverage as well—our API has shown up to ~200× more distinct markets per user than static popular lists—helping spread liquidity beyond the usual top few.
Each returned market is scored using signals such as:
User’s last 10 trades and their volumes
Most-traded markets (by count)
Markets where the user has deployed the most $ (cumulative volume)
Scores are relative to other markets in the list of candidates presented to our AI model, which can be prefiltered according to parameters like.
Volume
Liquidity
Price change
Start and End Date
The output is a ranked list of markets that can be used to personalize the user’s experience.
For example, for @Domahhhh—a long-time, profitable and self proclaimed “political bettor”—our Alpha model surfaces a politics-heavy set with a dash of esports/sports discovery and some ETH price markets further down the list, making sure that the user doesn’t miss any significant bet in their core categories while adding some new, less conventional markets, to discover.
The API also returns hot movers—markets with significant short-term price/odds changes.
How to use the API
Provide a user/wallet ID
Choose how many recommendations to return
Optionally set min volume/liquidity, date range, price-change filters
Receive a ranked list of markets
DM us to play with it
You can use the API to build dynamic, relevant PM interfaces: feeds, carousels, and smart notifications that surface the right markets at the right time.
We’re actively looking to partner with builders in order to bring our new AI model to market. DM us on twitter @mbdtheworld and we’ll be in touch soon.