Texas Holdem

Superhuman AI for heads-up no-limit poker: Libratus beats leading gurus

This speak offers a higher-level clarification of Libratus, the first AI to defeat leading humans in no-limit poker. A paper on the AI was posted in Science in 2017.

No-limit Texas hold’em is the most well-known kind of poker. Despite AI successes in best-information video games, the personal information and massive video game tree have built no-limit poker tricky to tackle. We existing Libratus, an AI that, in a one hundred twenty,000-hand competitiveness, defeated 4 leading human expert gurus in heads-up no-limit Texas hold’em, the primary benchmark and lengthy-standing obstacle trouble in imperfect-information video game resolving. Our video game-theoretic method attributes application-impartial procedures: an algorithm for computing a blueprint for the all round tactic, an algorithm that fleshes out the specifics of the tactic for subgames that are arrived at throughout engage in, and a self-improver algorithm that fixes probable weaknesses that opponents have identified in the blueprint tactic.



  1. Very cool

  2. A great video, you are explaining the approach in great detail but still having the overall architecture present.
    And I can only congrats you all to the great research community you managed to build up, well done!

  3. More people should do this! Start a trend! Incredible you took the time to make this

  4. absolutely fascinating!!!

  5. Good shit noam. Grats.


Have your say