Towards a Characterisation of Monte-Carlo Tree Search Performance in Different Games

Dennis J.N.J. Soemers; Guillaume Bams; Max Persoon; Marco Rietjens; Dimitar Sladić; Stefan Stefanov; Kurt Driessens; Mark H.M. Winands

doi:10.1109/CoG60054.2024.10645675

Towards a Characterisation of Monte-Carlo Tree Search Performance in Different Games

Dennis J.N.J. Soemers, Guillaume Bams, Max Persoon, Marco Rietjens, Dimitar Sladić, Stefan Stefanov, Kurt Driessens, Mark H.M. Winands

Research output: Chapter in Book/Report/Conference proceeding › Conference article in proceeding › Academic › peer-review

Abstract

Many enhancements to Monte-Carlo Tree Search (MCTS) have been proposed over almost two decades of general game playing and other artificial intelligence research. However, our ability to characterise and understand which variants work well or poorly in which games is still lacking. This paper describes work on an initial dataset that we have built to make progress towards such an understanding: 268,386 plays among 61 different agents across 1494 distinct games. We describe a preliminary analysis and work on training predictive models on this dataset, as well as lessons learned and future plans for a new and improved version of the dataset.

Original language	English
Title of host publication	2024 IEEE Conference on Games (CoG)
Publisher	IEEE
Pages	1-4
ISBN (Electronic)	979-8-3503-5067-8
ISBN (Print)	979-8-3503-5068-5
DOIs	https://doi.org/10.1109/CoG60054.2024.10645675
Publication status	Published - 5 Aug 2024
Event	2024 IEEE Conference on Games - Milan, Italy Duration: 5 Aug 2024 → 8 Aug 2024 https://2024.ieee-cog.org/

Conference

Conference	2024 IEEE Conference on Games
Abbreviated title	IEEE CoG 2024
Country/Territory	Italy
City	Milan
Period	5/08/24 → 8/08/24
Internet address	https://2024.ieee-cog.org/

Access to Document

10.1109/CoG60054.2024.10645675

https://ieeexplore.ieee.org/document/10645675/

Embargoed Document

Full Text
Final published version, 203 KB
Licence: Taverne
Embargo ends: 5/02/25
Request copy

Cite this

@inproceedings{84902b287b554d64a07b5196fbb6ea88,

title = "Towards a Characterisation of Monte-Carlo Tree Search Performance in Different Games",

abstract = "Many enhancements to Monte-Carlo Tree Search (MCTS) have been proposed over almost two decades of general game playing and other artificial intelligence research. However, our ability to characterise and understand which variants work well or poorly in which games is still lacking. This paper describes work on an initial dataset that we have built to make progress towards such an understanding: 268,386 plays among 61 different agents across 1494 distinct games. We describe a preliminary analysis and work on training predictive models on this dataset, as well as lessons learned and future plans for a new and improved version of the dataset.",

author = "Soemers, {Dennis J.N.J.} and Guillaume Bams and Max Persoon and Marco Rietjens and Dimitar Sladi{\'c} and Stefan Stefanov and Kurt Driessens and Winands, {Mark H.M.}",

year = "2024",

month = aug,

day = "5",

doi = "10.1109/CoG60054.2024.10645675",

language = "English",

isbn = "979-8-3503-5068-5",

pages = "1--4",

booktitle = "2024 IEEE Conference on Games (CoG)",

publisher = "IEEE",

address = "United States",

note = "2024 IEEE Conference on Games, IEEE CoG 2024 ; Conference date: 05-08-2024 Through 08-08-2024",

url = "https://2024.ieee-cog.org/",

}

TY - GEN

T1 - Towards a Characterisation of Monte-Carlo Tree Search Performance in Different Games

AU - Soemers, Dennis J.N.J.

AU - Bams, Guillaume

AU - Persoon, Max

AU - Rietjens, Marco

AU - Sladić, Dimitar

AU - Stefanov, Stefan

AU - Driessens, Kurt

AU - Winands, Mark H.M.

PY - 2024/8/5

Y1 - 2024/8/5

N2 - Many enhancements to Monte-Carlo Tree Search (MCTS) have been proposed over almost two decades of general game playing and other artificial intelligence research. However, our ability to characterise and understand which variants work well or poorly in which games is still lacking. This paper describes work on an initial dataset that we have built to make progress towards such an understanding: 268,386 plays among 61 different agents across 1494 distinct games. We describe a preliminary analysis and work on training predictive models on this dataset, as well as lessons learned and future plans for a new and improved version of the dataset.

AB - Many enhancements to Monte-Carlo Tree Search (MCTS) have been proposed over almost two decades of general game playing and other artificial intelligence research. However, our ability to characterise and understand which variants work well or poorly in which games is still lacking. This paper describes work on an initial dataset that we have built to make progress towards such an understanding: 268,386 plays among 61 different agents across 1494 distinct games. We describe a preliminary analysis and work on training predictive models on this dataset, as well as lessons learned and future plans for a new and improved version of the dataset.

U2 - 10.1109/CoG60054.2024.10645675

DO - 10.1109/CoG60054.2024.10645675

M3 - Conference article in proceeding

SN - 979-8-3503-5068-5

SP - 1

EP - 4

BT - 2024 IEEE Conference on Games (CoG)

PB - IEEE

T2 - 2024 IEEE Conference on Games

Y2 - 5 August 2024 through 8 August 2024

ER -

Towards a Characterisation of Monte-Carlo Tree Search Performance in Different Games

Abstract

Conference

Access to Document

Embargoed Document

Fingerprint

Cite this