Key Points
- 1.Predictive markets and betting platforms often have correlated odds, making it difficult to find expected value simply by picking winners.
- 2.Normalizing bookmaker odds by removing the margin allows comparison with prediction market probabilities.
- 3.Comparing bookmaker odds with platforms like Polymarket and Kalshi can reveal mispriced odds and positive expected value.
- 4.Building an autonomous system using AI tools like Codex helps automate the process of tracking live prices and hunting for mispriced odds.
- 5.Successful betting strategies focus on finding price discrepancies relative to fair probability rather than predicting game outcomes.
Summary
The discussion opens by addressing questions about setting up AI systems to find good prospects and markets on prediction and betting platforms. Many participants focus primarily on who will win a game, such as an upcoming match between Arsenal and Manchester City. However, achieving long term profitability requires shifting focus toward expected value. This principle applies not only to sports betting but also to automated trading with artificial intelligence, where expected value is essential to avoid simple gambling.
To illustrate how to find mispricing, the video examines a specific match between Arsenal and Manchester City. On Polymarket, the percentages are listed as 37 percent for Arsenal, 36 percent for Manchester City, and 28 percent for a draw. Checking Kalshi yields very similar numbers, showing that these prediction markets are highly correlated, which makes finding value between them difficult. Instead, the analysis turns to bookmakers like Bet365, which list decimal odds of 2.45 for Arsenal, 3.60 for a draw, and 2.55 for Manchester City, alongside American odds of plus 150, plus 235, and plus 160.
Calculating the inverse of the decimal odds results in a total of 108 percent, indicating an 8 percent bookmaker margin. Normalizing these numbers down to 100 percent by dividing each by the margin gives fair probabilities of 38 percent for Arsenal, 26 percent for a draw, and 36 percent for away. Comparing these normalized bookmaker odds with Polymarket reveals potential mispricing. For instance, if bookmakers and alternative sportsbooks like DraftKings and FanDuel assign a significantly higher percentage to a home win than prediction markets do, an edge emerges. The core insight is that the focus should remain entirely on price discrepancies relative to fair probability rather than trying to predict the exact winner of the game.
To scale this process, the creator demonstrates building an autonomous system using OpenAI Codex. The system uses free web sockets on Polymarket and Kalshi to get live prices for an event, while bookmaker odds are supplied via API or manual entry. The Codex agent is prompted to pull the live prices, recalculate and normalize the bookmaker odds into percentages, and check for outliers or mispriced markets that contain positive expected value. Although running the example for the Community Shield match shows no actionable expected value at the current asks, the process highlights how developers can build AI systems to autonomously monitor markets and find trading opportunities over time.
The Search for Expected Value
The video discusses the importance of shifting focus from picking winners to finding expected value in order to be profitable in the long run on prediction markets and sports betting platforms.
Comparing Odds and Normalizing Margins
An example using an Arsenal versus Manchester City match demonstrates how to compare Polymarket and Kalshi percentages with bookmaker odds from Bet365 by converting decimal odds and removing the margin.
Automating with AI
The creator showcases how to build a simple system using Codex and web sockets to autonomously pull live prices from Polymarket and Kalshi, calculate normalized probabilities, and look for mispriced odds.
Worth watching for
This video is for people interested in automated trading, prediction markets, and finding expected value using AI tools.
- polymarket
- kalshi
- ai
- betting
- trading