When is it better not to look ahead?

Nau, Dana S.; Luštrek, Mitja; Parker, Austin; Bratko, Ivan; Gams, Matja

doi:10.1016/j.artint.2010.08.002

Cited by 18 publications

(20 citation statements)

References 19 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We call these pathological situations local pathologies . As a consequence (and in accordance with the work of Nau et al), one can say that different games exhibit different degrees of local pathologies.…”

Section: Theoretical Analysissupporting

confidence: 83%

“…Finally, a large experimental study was performed to examine the relationship between the degree of pathology in a game tree and the 3 prominent causes of pathology, namely, branching factor, local node similarity, and evaluation function granularity . In that study, the authors defined the degree of pathology for a search of depth d as the fraction of correct decisions made by searching to depth d over the fraction of correct decisions made by searching one level deeper.…”

Section: Background and Related Workmentioning

confidence: 99%

“…Nevertheless, recent work has shown that pathological situations can occur in Chess (in 2 specific endgame positions), the African game of Kalah, and the single‐player 8‐puzzle. These new insights suggest that the widespread assumption that real games are not pathological should be revisited.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Avoiding game‐tree pathology in 2‐player adversarial search

Zuckerman

Wilson

Nau

2018

Computational Intelligence

Self Cite

View full text Add to dashboard Cite

Adversarial search, or game‐tree search, is a technique for analyzing an adversarial game to determine what moves a player should make in order to win a game. Until recently, lookahead pathology (in which deeper game‐tree search results in worse play) has been thought to be quite rare. We provide an analysis that shows that every game should have some sections that are locally pathological, assuming that both players can potentially win the game. We also modify the minimax algorithm to recognize local pathologies in arbitrary games and cut off search accordingly (shallower search is more effective than deeper search when local pathologies occur). We show experimentally that our modified search procedure avoids local pathologies and consequently provides improved performance, in terms of decision accuracy, when compared with the minimax algorithm. In addition, we provide an experimental evaluation on the African game of Kalah, which shows the improved performances of our suggested error‐minimizing minimax algorithm when there is a large degree of pathology.

show abstract

Section: Theoretical Analysissupporting

confidence: 83%

Section: Background and Related Workmentioning

confidence: 99%

See 1 more Smart Citation

Avoiding game‐tree pathology in 2‐player adversarial search

Zuckerman

Wilson

Nau

2018

Computational Intelligence

Self Cite

View full text Add to dashboard Cite

show abstract

“…For example, a well-known fact arising in some bipersonal games is that of lookahead pathology, that is, that the deeper is the exploration performed (and hence the better is suppossedly the heuristic minimax value assigned to the position), the worse is the decision taken (Nau, 1982). Recently, the same phenomenon was described in one-agent real-time search (Lustrek & Bulitko, 2008;Nau, Lustrek, Parker, Bratko, & Gams, 2010).…”

Section: Introductionmentioning

confidence: 80%

A Case of Pathology in Multiobjective Heuristic Search

Cruz¹,

Mandow²,

Machuca³

2013

jair

View full text Add to dashboard Cite

This article considers the performance of the MOA* multiobjective search algorithm with heuristic information. It is shown that in certain cases blind search can be more efficient than perfectly informed search, in terms of both node and label expansions.A class of simple graph search problems is defined for which the number of nodes grows linearly with problem size and the number of nondominated labels grows quadratically. It is proved that for these problems the number of node expansions performed by blind MOA* grows linearly with problem size, while the number of such expansions performed with a perfectly informed heuristic grows quadratically. It is also proved that the number of label expansions grows quadratically in the blind case and cubically in the informed case.

show abstract

“……" Many researchers have shown the importance of the lookahead depth for computer games, but none of them was related to checkers. Most of the findings are related to chess [14][15][16][17], where it was shown that increasing the depth level will produce superior chess players. However, Runarsson and Jonsson [18] showed that this was not the case for Othello, as they found that better playing strategies are found when TD learning when ε-greedy is applied with a lower look-ahead search depth and a deeper look-ahead search during game play.…”

Section: Introductionmentioning

confidence: 94%

The importance of look-ahead depth in evolutionary checkers

Al‐Khateeb

Kendall

2011

2011 IEEE Congress of Evolutionary Computation (CEC)

View full text Add to dashboard Cite

Intuitively it would seem to be the case that any learning algorithm would perform better if it was allowed to search deeper in the game tree. However, there has been some discussion as to whether the evaluation function or the depth of the search is the main contributory factor in the performance of the player. There has been some evidence suggesting that lookahead (i.e. depth of search) is particularly important. In this work we provide a rigorous set of experiments, which support this view. We believe this is the first time such an intensive study has been carried out for evolutionary checkers. Our experiments show that increasing the depth of a look-ahead has significant improvements to the performance of the checkers program and has a significant effect on its learning abilities.

show abstract

When is it better not to look ahead?

Abstract: D.S. Nau), mitja.lustrek@ijs.si (M. Luštrek), ajpark2@super.org (A. Parker), bratko@fri.uni-lj.si (I. Bratko), Matjaz.Gams@ijs.si (M. Gams).

Cited by 18 publications

References 19 publications

Avoiding game‐tree pathology in 2‐player adversarial search

Avoiding game‐tree pathology in 2‐player adversarial search

A Case of Pathology in Multiobjective Heuristic Search

The importance of look-ahead depth in evolutionary checkers

Contact Info

Product

Resources

About