The DeepMind Trio That Built Poker AI Now Makes Money for Quant Hedge Funds

From Poker Table to Trading Desk

Three former researchers at Google DeepMind, renowned for jointly developing an AI system that defeated top poker players, are now migrating the game theory (GTO) strategies from poker to financial markets, generating profits for a quantitative hedge fund.

The Birth of the Poker AI

During their time at DeepMind, the trio successfully developed a poker AI named "Pluribus." Pluribus defeated multiple professional players in six-player no-limit Texas Hold'em, becoming the first AI to beat humans in multi-player poker. Its core algorithm is based on self-play and approximate Nash equilibrium, enabling optimal decision-making under incomplete information.

Moving to Quantitative Finance

After leaving DeepMind, the three co-founded a quantitative hedge fund, applying the decision-making framework from the poker AI to trading assets such as stocks and futures. Leveraging similar game-theoretic models, they identify asymmetric opportunities in complex market environments, reportedly generating substantial returns for the fund.

Commonalities Between Poker and Finance

• Incomplete information: Both face missing information, requiring inference based on probabilities and opponent behavior.
• Game theory: The GTO strategy in poker is analogous to equilibrium pricing models in financial markets.
• Risk management: Bankroll management techniques from poker directly apply to portfolio leverage and drawdown control.

Impact and Outlook

This team's transformation further underscores AI's potential in the financial sector. The fruits of poker research are not limited to the game itself; their methodologies are advancing quantitative investing into a new phase. In the future, more poker AI techniques may be applied to high-frequency trading, market-making strategies, and beyond.