An investigation into 2048 AI strategies

Philip Rodgers; John Levine

doi:10.1109/CIG.2014.6932920

An investigation into 2048 AI strategies

Philip Rodgers, John Levine

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution book

11 Citations (Scopus)

Abstract

2048 is a recent stochastic single player game, originally written in JavaScript for playing in a web browser but now largely played on mobile devices. This paper discusses the applicability of Monte-Carlo Tree-Search (MCTS) to the problem, and also Averaged Depth Limited Search (ADLS). While MCTS plays reasonably well for a player with no domain knowledge, the ADLS player fares much better given an evaluation function that rewards board properties. Attempts to guide the roll-outs of MCTS using an evaluation function proved fruitless.

Original language	English
Title of host publication	2014 IEEE Conference on Computational Intelligence and Games
Place of Publication	Piscataway, NJ
Publisher	IEEE
Pages	1-2
Number of pages	2
ISBN (Print)	9781479935468
DOIs	https://doi.org/10.1109/CIG.2014.6932920
Publication status	Published - 26 Aug 2014
Event	2014 IEEE Conference on Computational Intelligence and Games - Park Inn Hotel, Dortmund, Germany Duration: 26 Aug 2014 → 29 Aug 2014 https://www.ieee.org/conferences_events/conferences/conferencedetails/index.html?Conf_ID=32016

Conference

Conference	2014 IEEE Conference on Computational Intelligence and Games
Abbreviated title	CIG 2014
Country/Territory	Germany
City	Dortmund
Period	26/08/14 → 29/08/14
Internet address	https://www.ieee.org/conferences_events/conferences/conferencedetails/index.html?Conf_ID=32016

Keywords

Monte Carlo methods
artificial intelligence
computer games
tree searching
mobile devices
JavaScript
Monte-Carlo Tree-Search
Averaged Depth Limited Search

Access to Document

10.1109/CIG.2014.6932920

Cite this

@inproceedings{062f6e4a2a204716b1066e03aa4e826c,

title = "An investigation into 2048 AI strategies",

abstract = "2048 is a recent stochastic single player game, originally written in JavaScript for playing in a web browser but now largely played on mobile devices. This paper discusses the applicability of Monte-Carlo Tree-Search (MCTS) to the problem, and also Averaged Depth Limited Search (ADLS). While MCTS plays reasonably well for a player with no domain knowledge, the ADLS player fares much better given an evaluation function that rewards board properties. Attempts to guide the roll-outs of MCTS using an evaluation function proved fruitless. ",

keywords = "Monte Carlo methods, artificial intelligence, computer games, tree searching, mobile devices, JavaScript, Monte-Carlo Tree-Search, Averaged Depth Limited Search",

author = "Philip Rodgers and John Levine",

year = "2014",

month = aug,

day = "26",

doi = "10.1109/CIG.2014.6932920",

language = "English",

isbn = "9781479935468",

pages = "1--2",

booktitle = "2014 IEEE Conference on Computational Intelligence and Games",

publisher = "IEEE",

note = "2014 IEEE Conference on Computational Intelligence and Games, CIG 2014 ; Conference date: 26-08-2014 Through 29-08-2014",

url = "https://www.ieee.org/conferences\_events/conferences/conferencedetails/index.html?Conf\_ID=32016",

}

TY - GEN

T1 - An investigation into 2048 AI strategies

AU - Rodgers, Philip

AU - Levine, John

PY - 2014/8/26

Y1 - 2014/8/26

N2 - 2048 is a recent stochastic single player game, originally written in JavaScript for playing in a web browser but now largely played on mobile devices. This paper discusses the applicability of Monte-Carlo Tree-Search (MCTS) to the problem, and also Averaged Depth Limited Search (ADLS). While MCTS plays reasonably well for a player with no domain knowledge, the ADLS player fares much better given an evaluation function that rewards board properties. Attempts to guide the roll-outs of MCTS using an evaluation function proved fruitless.

AB - 2048 is a recent stochastic single player game, originally written in JavaScript for playing in a web browser but now largely played on mobile devices. This paper discusses the applicability of Monte-Carlo Tree-Search (MCTS) to the problem, and also Averaged Depth Limited Search (ADLS). While MCTS plays reasonably well for a player with no domain knowledge, the ADLS player fares much better given an evaluation function that rewards board properties. Attempts to guide the roll-outs of MCTS using an evaluation function proved fruitless.

KW - Monte Carlo methods

KW - artificial intelligence

KW - computer games

KW - tree searching

KW - mobile devices

KW - JavaScript

KW - Monte-Carlo Tree-Search

KW - Averaged Depth Limited Search

U2 - 10.1109/CIG.2014.6932920

DO - 10.1109/CIG.2014.6932920

M3 - Conference contribution book

SN - 9781479935468

SP - 1

EP - 2

BT - 2014 IEEE Conference on Computational Intelligence and Games

PB - IEEE

CY - Piscataway, NJ

T2 - 2014 IEEE Conference on Computational Intelligence and Games

Y2 - 26 August 2014 through 29 August 2014

ER -

An investigation into 2048 AI strategies

Abstract

Conference

Keywords

Access to Document

Fingerprint

Cite this