Slashdot Mirror
Awari Solved
Gerard Jendras sent in a submission about applying computing power to an ancient game. The game of Awari has been solved: with perfect play, the game always results in a draw. There is a Java applet to test your skills against.
← Back to Stories (view on slashdot.org)
This is more commonly known as Mancala in the US.
An adaptation (simplified) of the game was used as a problem in last year's International Olympiad in Informatics: see the description of the problem here. For a description of how to solve it efficiently, see this booklet.
It appeared in Serria's Quest for Glory III: Trial by Fire which was set in a mythical Africa-like kingdom that included an Egyptian-type city, a savahhana and jungle. Awari was one of the minigames that needed to be completed in order to progress through the game. They don't make them like that anymore; awari or QfG3
I'm tired of bombing the universe
Some people once said that Awari was more complex (= offered more possibilities) than Chess...
You're thinking of Go.
I've heard the same thing. The next time you get what you consider to be an unsolveable game, fire up this bad boy and check it out:
freecell-solver
Do you have Linux and a DotPal? Click here now!
There are exactly 4 ways to draw in chess:
Accept a draw offer
Get stalemated (no legal move, but not in check)
Repeat the exact same position (same player to move, same en-passant square, etc.) 3 times (not necessarily in a row, or in check either)
Make 50 moves without moving a pawn or capturing a piece
There's a draw if a position is repeated three times (not necessarily consecutively).
There's no rule about consecutive checks. In fact some fun chess problems involve dozens of consecutive checks.
No, it's if the same position occurs 3 times, not necessarily on successive moves. But someone needs to claim the draw. If nobody claims, the game can go on indefinitely
See Laws of Chess for more details.Yes, there is about 5.26e+36 different ways of putting the pieces on a board. A harddisk that could hold this data (and subsequent best moves) would take a significant amount of our solar system and surrounding stars, even is we could encode a bit a single atom level.
IANAL, but imagine a beowulf cluster of in Soviet Russia all your belong are base to us welcoming the new SCO overlords.
In this instance, no matter what move you make the computer knows what every possible board combination from now to the end of the game can look like.. You can't throw him for a loop with imperfect play because he will have already 'predicted' (by way of a simple scoring algorithm) the possibility that one of your moves could result in weakining his position and he would have avoided making the move that allows you to make that move.
Your observation makes a bit more sense in chess, since chess isn't really "solved" yet (far more possible move combinations than in this game). But once a game is "solved" in this manner, you'll never beat the computer, ever, no matter how clever you are unless the game is flawed (one player has an inherent advantage) or the computer has a programmer error.
Another way to look at it is Tic-Tac-Toe. Tic-tac-toe is simple enough that it can be 'solved' in the human mind by any reasonably intelligent person. If you play against such a person, you'll never win no matter how 'tricky' your moves..
I think you're quite correct here. Brute force isn't a reasonable solution to most interesting realworld problems, and it's hard to see how this approach is instructive for future AI research.
Humans don't play games by checking every possible move and picking the best one and never will. The AI community really needs to stop looking for tricks that allow computers to solve problems in ways that humans never could
Ah, now I think you're overstating things a bit. It sounds like your objection is predicated (sic) upon the assumption that the sole purpose of AI research is to reproduce biological, and finally human, intelligence, and that the way biological brains do it is the only way it can be done. I think that's not necessarily the case.
In the first matter, there's plenty of higher-order problems that we might want solved, and there's no reason to suppose that a human thought process is the best way to go about things. The sole purpose of AI research isn't to pass a Turing test. A truly artificial intelligence, that solves hard problems in exotic ways would still be a very interesting and useful thing to possess, even if one couldn't talk sports with it.
In the second matter, it's still a contentious matter whether human-like intelligence could ever be built with hardware and software components like the ones we have today. In the event that it is, it's quite possible that the artificial solution would take a radically different path than the biological - a path worthy of exploration, interesting even if it ends in a blind alley.
That's not unlike saying that birds fly by flapping feathery wings, and thus so should passenger aircraft - they're solving different problems, and with different means.
Study of the biological solution is instructive - mere apeing of it isn't.
## W.Finlay McWalter ## http://www.mcwalter.org ##
There are hundreds, maybe thousands of variations on Mancala-type games (bowls and stones which are placed in them). These games are played all over the world.
I just returned from Kenya and Tanzania where I bought a Mancala borard for a local gamed called Bau (also spelled Bao). It is a lot more complicated than Awati and some say it is the most complex form of Mancala.