NUI Galway – UL Alliance First Annual ENGINEERING AND - ARAN ...
NUI Galway – UL Alliance First Annual ENGINEERING AND - ARAN ...
NUI Galway – UL Alliance First Annual ENGINEERING AND - ARAN ...
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
Evolution and Analysis of Strategies for Mancala Games<br />
Damien Jordan and Colm O’Riordan<br />
CIRG, Information Technology College of Engineering and Informatics<br />
damojordan@gmail.com, colm.oriordan@nuigalway.ie<br />
Abstract<br />
Mancala games are are a range of strategy games.<br />
This research attempts to better understand the variants<br />
of the game by investigating heuristics to play the game<br />
We then combine a number of heuristics together to<br />
form a strategy. A genetic algorithm is used to evolve<br />
the most successful strategy for this game.<br />
1. Mancala Games<br />
Mancala is a family of two-player board games that<br />
are popular all over the world. There are over 300<br />
documented variants of Mancala. The object of mancala<br />
games is usually to capture more seeds than one’s<br />
opponent.<br />
The game begins with the players placing an equal<br />
number of seeds, as per the variation in use, in each of<br />
the bowls on the game board. A turn typically consists<br />
of removing all seeds from a bowl, placing one seed in<br />
each of the following bowls in sequence, and capturing<br />
seeds based on the rules of the game. The exact rules<br />
for capturing vary considerably among the variants.<br />
For more than a century, board games and strategy<br />
games have been the topic of many scientific studies by<br />
psychologists and scientists. “Board games have long<br />
fascinated as mirrors of intelligence, skill, cunning and<br />
wisdom” [1]. Mancala games represent an interesting<br />
topic of study given the wide range of rule variations<br />
resulting in games of differing levels of difficulty<br />
2. Hypothesis<br />
Many interesting research questions exist in the<br />
domain of mancala games. These include: are there<br />
winning strategies? For which variants do these<br />
strategies exist? Can these strategies be represented as<br />
heuristics? Are heuristics developed for one game<br />
transferrable to another? Which changes to the rules<br />
change the difficulty?<br />
In this paper we focus our studies on one variant of<br />
the game, Bantumi. We hypothesise that a set of<br />
heuristics can be developed and empirically tested to<br />
measure their efficacy and secondly, that evolutionary<br />
computation can be used to learn a robust strategy<br />
3. Methodology<br />
The methodology employed in this study includes:<br />
design and development of a simulator, design and<br />
development of heuristics, empirical testing of these<br />
heuristics and the use of a genetic algorithm to evolve a<br />
suitable strategy.<br />
4. Current work/Results to date<br />
A simulation for the mancala game Bantumi (and<br />
variants) has been designed and implemented. Seven<br />
5<br />
heuristics have been designed (following analysis of the<br />
literature and game play) and implemented for Bantumi.<br />
These are:-<br />
H1-Pick a bowl that allows the player to have another go<br />
H2-Pick a bowl that allows the player to make a capture<br />
H3-If the opponent has seeds in bowls that allow him another<br />
go, disrupt it<br />
H4-If the opponent can capture some of the player’s seeds on<br />
the next go, move them<br />
H5-Always pick the closest bowl to the score bowl<br />
H6-Avoid picking a bowl that, after sowing, results in giving<br />
the opponent another go<br />
H7-Avoid picking a bowl that, after sowing, results in<br />
allowing the opponent to capture some of the player’s seeds<br />
All heuristics were tested against each other in a<br />
round robin tournament. The results of these<br />
experiments showed that H1 and H5 were the two<br />
strongest heuristics of the group, while H3, H6 and H7<br />
were the weakest. The results of this experiment are<br />
shown below:<br />
Win % after 100 games<br />
100<br />
80<br />
60<br />
40<br />
20<br />
0<br />
H7<br />
3 - F 3 - S 4 - F 4 - S 5 - F 5 - S 6 - F 6 - S<br />
Rand<br />
Digit = seeds/bowl, F = <strong>First</strong> move, S = Second<br />
move<br />
Combining heuristics H1, H2, H4 and H5 in a linear<br />
order to form a new heuristic was shown to win an<br />
average of 83% of games when played against all other<br />
heuristics. A genetic algorithm was designed and<br />
implemented in our simulator for Bantumi. After<br />
numerous generations, and millions of games played, a<br />
strategy has evolved when using 3 seeds per bowl that<br />
wins an average of 96% of games when played against<br />
all other heuristics.<br />
5. References<br />
[1] J. Retschitzki, A. J. de Voogt & F. Gobet, “Moves<br />
in Mind: The Psychology of Board Games.” Psychology<br />
Press Ltd, Hove UK, 2004.<br />
H1<br />
H2<br />
H3<br />
H4<br />
H5<br />
H6