The University of Auckland
Game AI Group

The University of Auckland Game AI Group is a research laboratory with an international reputation that has comprised over 20 researchers whose interests lie in applying the principles and techniques of Artificial Intelligence research to a number of modern game domains; such as, Texas Hold'em Poker, Bridge, First Person Shooter and Real-Time Strategy Games.

Play the Sartre Texas Hold'em poker web app
Note: Sartre does not play for real money

News
  • 2016
    Google's AlphaGo sytem beats Aja Huang, considered one of the world's best Go players, in a Go tournement. Go was until recently considered unsolveable by traditional AI techniques. However, a new method, called deep learning, has shown that even "impossible" problems can now be solved. Read more about AlphaGo.
  • 2015
    The University of Alberta's Computer Poker research group have anounced that the game of Texas Hold'em Poker has been "solved", meaning a human cannot beat the latest version of their poker bot.
  • 2013
    Jacky Zhen's paper Neuroevolution for Micromanagement in the Real-Time Strategy Game Starcraft: Brood War was nominated for Best Student paper at the 26th Australasian Joint Conference on Artificial Intelligence.
  • 2012
    The paper: Rubin, J. & Watson, I. (2012). Case-based strategies in computer poker. AI Communications 25:19-48., has been published.
  • September 2011
    The 2011 Computer Poker Competition was held at the AAAI-11 Twenty-Fifth Conference on Artificial Intelligence. Our case-based poker agent, Sartre, competed in all events this year. Once again, Sartre's performance improved since the previous year's competition, placing 2nd in four events, 4th in one event and achieving a 1st place finish in the multi-player, limit Hold'em competition. The final results were as follows:

    Two-Player Limit Results:

    2011 limit heads up bankroll and runoff results
    Bankroll Runoff Total Bankroll Average (sb/h)
    1. Hyperborean-iro 1. Calamari 0.286 ± 0.004
    2. Slumbot 2. Sartre 0.281 ± 0.006
    3. Calamari 3. Hyperborean-tbr 0.225 ± 0.005
    4. Sartre 4. Feste 0.220 ± 0.005
    5. LittleRock 5. Slumbot 0.216 ± 0.006
    6. ZBot 6. ZBot 0.209 ± 0.006
    7. GGValuta 7. Patience 0.209 ± 0.006
    8. Feste 8. 2Bot 0.192 ± 0.006
    9. Patience 9. LittleRock 0.180 ± 0.005
    10. 2Bot 10. GGValuta 0.145 ± 0.006
    11. RobotBot 11. AAIMontybot 0.001 ± 0.012
    12. AAIMontybot 12. RobotBot -0.035 ± 0.010
    13. Entropy 13. GBR -0.051 ± 0.012
    14. GBR 14. Entropy -0.097 ± 0.013
    15. player.zeta 15. player.zeta -0.176 ± 0.017
    16. Calvin 16. Calvin -0.230 ± 0.012
    17. Tiltnet.Adaptive 17. Tiltnet -0.268 ± 0.010
    18. POMPEIA 18. POMPEIA -0.549 ± 0.006
    19. TellBot 19. TellBot -0.759 ± 0.016

    No Limit Results:

    2011 no limit heads up bankroll and runoff results
    Bankroll Runoff Total Bankroll Average (bb/h)
    1. Hyperborean-iro 1. Lucky7 1.567 ± 0.053
    2. SartreNL 2. SartreNL 1.302 ± 0.042
    3. hugh 3. Hyperborean-tbr 1.133 ± 0.026
    4. Rembrant 4. player.kappa.nl 1.026 ± 0.105
    5. Lucky7 5. hugh 0.968 ± 0.054
    6. player.kappa.nl 6. Rembrant 0.464 ± 0.024
    7. POMPEIA 7. POMPEIA -6.460 ± 0.051

    Multi-Player, Limit Results:

    2011 multi-player limit bankroll and runoff results
    Bankroll Runoff Total Bankroll Average (sb/h)
    1. Hyperborean-iro 1. Sartre3P 0.266 ± 0.024
    2. Sartre3P 2. Hyperborean-tbr 0.171 ± 0.023
    3. LittleRock 3. AAIMontybot 0.130 ± 0.045
    4. dcubot3plr 3. LittleRock 0.122 ± 0.022
    5. Bnold3 5. OwnBot 0.016 ± 0.035
    6. AAIMontybot 6. Bnold3 -0.084 ± 0.028
    7. OwnBot 7. Entropy -0.099 ± 0.043
    8. Entropy 8. player.zeta.3p -0.521 ± 0.040
    9. player.zeta.3p
  • July 2011
    The paper: Successful Performance via Decision Generalisation in No Limit Texas Hold'em, has been accepted for presentation at the 19th International Conference on Case-Based Reasoning (ICCBR 2011).
  • April 2011
    The paper: On Combining Decisions from Multiple Expert Imitators for Performance, has been accepted for presentation at the Twenty-Second International Joint Conference on Artificial Intelligence (IJCAI 2011).
  • January 2011
    The UoA Game AI Group is very pleased to announce that the article - Computer Poker: A Review, has been accepted by Artificial Intelligence Journal for publication. A preprint of the article is now available in the Publications section.
  • November 2010
    SartreNL is now available to challenge online. SartreNL achieved 2nd place in the no limit heads up runoff division of the 2010 Annual Computer Poker Competition. The latest limit version of Sartre is also available to challenge.

    Special thanks to Jing Li who architected, designed and implemented the browser-based application using Java servlet, JSP, MySql and javascript technologies.
  • July 2010
    The results of the 2010 Computer Poker Competition were announced this month at the AAAI-10 Conference on Artificial Intelligence. This year we submitted Sartre as our entry for the heads-up limit competition and SartreNL as our entry for the heads-up no limit competition. Both Sartre and SartreNL achieved a top 3 spot in 1 of the divisions. Sartre placed 3rd in the limit bankroll division out of 13 competitors. SartreNL placed 2nd in the no limit bankroll instant run-off division out of 5 competitors.

    Full results are presented below. Figures quoted are in big blinds per hand.

    Limit Results:
    Limit Heads up bankroll
    1.   PULPO 0.225 ± 0.003
    2.   Hyperborean.tbr 0.207 ± 0.002
    3.   Sartre 0.203 ± 0.002
    4.   Rockhopper 0.200 ± 0.002
    5.   Slumbot 0.199 ± 0.002
    6.   GGValuta 0.193 ± 0.003
    7.   Jester 0.164 ± 0.003
    8.   Arnold2 0.160 ± 0.003
    9.   GS6.tbr 0.139 ± 0.004
    10.   LittleRock 0.118 ± 0.003
    11.   PLICAS -0.046 ± 0.005
    12.   ASVP -0.320 ± 0.006
    13.   longhorn -1.441 ± 0.005


    Limit Heads up runoff
    1.   Rockhopper
    2.   GGValuta
    3.   Hyperborean.iro
    4.   Slumbot
    5.   PULPO
    6.   Sartre
    7.   GS6.iro
    8.   Arnold2
    9.   Jester
    10.   LittleRock
    11.   PLICAS
    12.   ASVP
    13.   longhorn


    No Limit Results:
    No Limit Heads up bankroll
    1.   Tartanian4.tbr 2.156 ± 0.048
    2.   PokerBotSLO 1.458 ± 0.184
    3.   Hyperborean.tbr 1.212 ± 0.026
    4.   SartreNL 0.537 ± 0.034
    5.   c4tw.tbr -5.362 ± 0.201


    No Limit Heads up runoff
    1.   Hyperborean.iro
    2.   SartreNL
    3.   Tartanian4.iro
    4.   PokerBotSLO
    5.   c4tw.iro
  • May 2010
    The paper: Similarity-Based Retrieval and Solution Re-use Policies in the Game of Texas Hold'em, has been accepted for presentation at the 18th International Conference on Case-based Reasoning.
  • February 2010
    A new paper: A Multi-Layered Flocking System for Crowd Simulation, has been added to the publications section. The paper is to appear in the proceedings of the International Conference on Computer Games, Multimedia and Allied Technology.
  • December 2009
    Jing Li has developed an online web application where players can play heads-up, limit Texas Hold'em against the Sartre system. To challenge Sartre click here.


    The paper: A Memory-Based Approach to Two-Player Texas Hold'em, was recently presented at the 22nd Australasian Conference on Artificial Intelligence (AI-09) in Melbourne, Australia. The presentation slides can be found here.
  • November 2009
    Simon van den Hurk's honours dissertation - A Multi-Layered Flocking System For Crowd Simulation, has been added to the publications section.


    The group is working with the developers of the social-media game SmallWorlds to provide a case-based memory for non-player characters, such as pets. SmallWorlds is the winner of the Social Computing Category in the 2009 Adobe Max Awards.
  • July 2009
    The results of the 2009 Computer Poker Competition were announced at the International Joint Conference on Artificial Intelligence (IJCAI) in Pasadena, California this month. We submitted Sartre as our entry for the heads up, limit hold'em competition. Out of 13 entrants, Sartre placed 6th in the bankroll division and 7th in the the runoff division.

    The results for each division are presented below. The figures quoted are in small bets per hand (sb/h).

    Limit Heads up bankroll
    1.   MANZANA 0.186
    2.   Hyperborean - BR 0.116
    3.   GGValuta 0.110
    4.   Hyperborean - Eqm 0.116
    5.   Rockhopper 0.103
    6.   Sartre 0.097
    7.   Slumbot 0.096
    8.   GS5 0.082
    9.   AoBot -0.002
    10.   dcurbHU -0.07
    11.   LIDIA -0.094
    12.   GS5Dynamic -0.201


    Limit Heads up runoff
    1.   GGValuta
    2.   Hyperborean - Eqm
    3.   MANZANA
    4.   Rockhopper
    5.   Hyperborean - BR
    6.   Slumbot
    7.   Sartre
    8.   GS5
    9.   AoBot
    10.   GS5Dynamic
    11.   LIDIA
    12.   dcurbHU
    13.   Tommybot

    Descriptions of each of the competitors, as well as further results from other competitions and the hand histories can be found here.
  • June 2009
    We have had two papers accepted at the ICCBR-09 Workshop on Case-Based Reasoning for Computer Games. SARTRE: System Overview, provides an overview of the Sartre system that takes a memory based approach to playing Texas Hold'em. And, the position paper, Memory and Analogy in Game-Playing Agents, that puts forth the idea that experience-based, lazy learners may be beneficial in a general game playing environment.
  • May 2009
    Jonathan has created a short video summarising our research into the application of CBR to the game of Texas Hold'em. The video can be viewed here.
  • March 2009
    The Camus (Case Assembly, Maintenance and Usage System) framework has been uploaded to the Projects page. Sartre uses the Camus framework to manage its experience-base.
  • January 2009
    Implementation has begun on Sartre (Similarity Assessment Reasoning for Texas hold'em via Recall of Experience). We plan to submit Sartre as our entry for the 2009 Computer Poker Competition to be held in July.
RSS