Combining Gameplay Data with Monte Carlo Tree Search to Emulate Human Play

Sam Devlin; Anastasija Anspoka; Nicholas John Sephton; Peter Ivan Cowling; Jeff Rollason

Combining Gameplay Data with Monte Carlo Tree Search to Emulate Human Play

Sam Devlin, Anastasija Anspoka, Nicholas John Sephton, Peter Ivan Cowling, Jeff Rollason

Computer Science

Research output: Contribution to conference › Paper › peer-review

Abstract

Monte Carlo Tree Search (MCTS) has become a popular solution for controlling non-player characters. Its use has repeatedly been shown to be capable of creating strong game playing opponents. However, the emergent playstyle of agents using MCTS is not necessarily human-like, believable or enjoyable. AI Factory Spades, currently the top rated Spades game in the Google Play store, uses a variant of MCTS to control non-player characters. In collaboration with the developers, we collected gameplay data from 27,592 games and showed in a previous study that the playstyle of human players significantly differed from that of the non-player characters. This paper presents a method of biasing MCTS using human gameplay data to create Spades playing agents that emulate human play whilst maintaining a strong, competitive performance. The methods of player modelling and biasing MCTS presented in this study are generally applicable to digital games with discrete actions.

Original language	English
Publication status	Published - 19 Sept 2016
Event	Twelfth Artificial Intelligence and Interactive Digital Entertainment Conference - Duration: 8 Oct 2016 → 12 Oct 2016 http://www.aiide.org/

Conference

Conference	Twelfth Artificial Intelligence and Interactive Digital Entertainment Conference
Abbreviated title	AIIDE
Period	8/10/16 → 12/10/16
Internet address	http://www.aiide.org/

Access to Document

http://www.aaai.org/ocs/index.php/AIIDE/AIIDE16/paper/view/14003

Peter I Cowling
- pic502york.acuk
- Computer Science - Anniversary Professor, Former employee
Person: Academic

The Digital Creativity Hub
Cowling, P. I. (Principal investigator), Austin, J. (Co-investigator), Cairns, P. A. (Co-investigator), Holliman, N. S. (Co-investigator), Hook, J. D. (Co-investigator), Marsden, E. (Co-investigator), Murphy, D. T. (Co-investigator), Petrie, H. (Co-investigator), Reed, D. J. (Co-investigator), Richards, J. D. (Co-investigator), Ursu, M. (Co-investigator), Wade, A. (Co-investigator), Baier, H. (Researcher), Beale, G. (Researcher), Block, F. O. (Researcher), Deterding, C. S. (Researcher), Devlin, S. (Researcher), Drachen, A. (Researcher), Kasprowicz, R. E. (Researcher), Smith, D. P. (Researcher) & Williams, D. A. H. (Researcher)
EPSRC
1/10/15 → 30/09/22
Project: Research project (funded) › Research
New Economic Models & Opportunities for Digital Games
Cowling, P. I., Cabras, I., Fernandes, K. J., Kudenko, D. & Devlin, S.
EPSRC
14/10/13 → 13/10/16
Project: Research project (funded) › Research

Cite this

@conference{a7b2d4d42c6341f399208781e8149206,

title = "Combining Gameplay Data with Monte Carlo Tree Search to Emulate Human Play",

abstract = "Monte Carlo Tree Search (MCTS) has become a popular solution for controlling non-player characters. Its use has repeatedly been shown to be capable of creating strong game playing opponents. However, the emergent playstyle of agents using MCTS is not necessarily human-like, believable or enjoyable. AI Factory Spades, currently the top rated Spades game in the Google Play store, uses a variant of MCTS to control non-player characters. In collaboration with the developers, we collected gameplay data from 27,592 games and showed in a previous study that the playstyle of human players significantly differed from that of the non-player characters. This paper presents a method of biasing MCTS using human gameplay data to create Spades playing agents that emulate human play whilst maintaining a strong, competitive performance. The methods of player modelling and biasing MCTS presented in this study are generally applicable to digital games with discrete actions.",

author = "Sam Devlin and Anastasija Anspoka and Sephton, {Nicholas John} and Cowling, {Peter Ivan} and Jeff Rollason",

year = "2016",

month = sep,

day = "19",

language = "English",

note = "Twelfth Artificial Intelligence and Interactive Digital Entertainment Conference, AIIDE ; Conference date: 08-10-2016 Through 12-10-2016",

url = "http://www.aiide.org/",

}

TY - CONF

T1 - Combining Gameplay Data with Monte Carlo Tree Search to Emulate Human Play

AU - Devlin, Sam

AU - Anspoka, Anastasija

AU - Sephton, Nicholas John

AU - Cowling, Peter Ivan

AU - Rollason, Jeff

PY - 2016/9/19

Y1 - 2016/9/19

N2 - Monte Carlo Tree Search (MCTS) has become a popular solution for controlling non-player characters. Its use has repeatedly been shown to be capable of creating strong game playing opponents. However, the emergent playstyle of agents using MCTS is not necessarily human-like, believable or enjoyable. AI Factory Spades, currently the top rated Spades game in the Google Play store, uses a variant of MCTS to control non-player characters. In collaboration with the developers, we collected gameplay data from 27,592 games and showed in a previous study that the playstyle of human players significantly differed from that of the non-player characters. This paper presents a method of biasing MCTS using human gameplay data to create Spades playing agents that emulate human play whilst maintaining a strong, competitive performance. The methods of player modelling and biasing MCTS presented in this study are generally applicable to digital games with discrete actions.

AB - Monte Carlo Tree Search (MCTS) has become a popular solution for controlling non-player characters. Its use has repeatedly been shown to be capable of creating strong game playing opponents. However, the emergent playstyle of agents using MCTS is not necessarily human-like, believable or enjoyable. AI Factory Spades, currently the top rated Spades game in the Google Play store, uses a variant of MCTS to control non-player characters. In collaboration with the developers, we collected gameplay data from 27,592 games and showed in a previous study that the playstyle of human players significantly differed from that of the non-player characters. This paper presents a method of biasing MCTS using human gameplay data to create Spades playing agents that emulate human play whilst maintaining a strong, competitive performance. The methods of player modelling and biasing MCTS presented in this study are generally applicable to digital games with discrete actions.

M3 - Paper

T2 - Twelfth Artificial Intelligence and Interactive Digital Entertainment Conference

Y2 - 8 October 2016 through 12 October 2016

ER -

Combining Gameplay Data with Monte Carlo Tree Search to Emulate Human Play

Abstract

Conference

Access to Document

Profiles

Peter I Cowling

Projects

The Digital Creativity Hub

New Economic Models & Opportunities for Digital Games

Cite this