Awari Oracle

From Wikimanqala

Awari Oracle screenshot

The Awari Oracle is a web frontend of a remote database containing the eventual outcome for all the 889,063,398,406 positions^[1] that can occur during a game of awari, a computer science variant of oware.

1 The applet
2 The database
3 The rules
- 3.1 Rules examples
4 Incidents
5 References
6 Footnotes
7 External links

[edit]

The applet

This java applet was developed by John Romein at the Division of Mathematics and Computer Science, Faculty of Sciences, Vrije Universiteit Amsterdam, in 2002, as part of his research to solve awari.

The Oracle can be used in two modes, both needing a working connection to the database, game mode and edit mode^[2].

Users can set the mode to “game” and play a game against the applet. There, they can choose the playing level between fair, good, strong, expert, and perfect. Theres no information about the way the Oracle chooses its moves when in the first four levels, but in the fifth level (perfect) it plays according to the results stored in the database, ie perfectly.

In edit mode the users can set a given position on the board and get the eventual outcome of the game (the seeds difference) for every possible move if played perfectly from then on.

The Oracle can be used to test the artificial intelligence of oware playing programmes, as did the Oware Wizard author^[3]. Also, in 2002, it was used to check the movements of a game between two master players, Trevor Simons and Ibrahim Abubakar^[4]. It showed that more than 95% of all moves were perfect. However, the rules used by the players (abapa) were not the same as the ones used to build the database, so no conclusion can be drawn.

[edit]

The database

The Aware Oracle is nothing without access to the database created by John Romein and Henri Bal. In fact, it does not even work.

The database was created using parallel retrograde analysis to get the eventual score of all possible positions, if played perfectly by both players from that position on. A computer cluster was needed to compute it, requiring 51 hours. The grid (connected through Myrinet) was made of 72 dual 1GHZ PIII computers, each with 1GB main memory^[5].

The database requires 778GB for being stored^[6].

It demonstrated that, if both players play perfectly, the result is a draw.

[edit]

The rules

As previously said, the rules used to create the database were similar to these of oware, but not the same^[7]. The main differences are:

a position that is repeated for the third time (in fact, the Oracle stops at the second time) results in a even division of the remaining stones, including a possible odd stone

The even division of the remaining stones is not used in any oware variant played by humans. It is only used in computer sciences. This leads to odd results as 23.5 vs 24.5 seeds.

it is not allowed to do a move that leaves the opponent without countermove, unless all moves eradicate the opponent

This rule is used in some oware variants. However, in tournaments using international rules (abapa), a player can do such a move, but captures nothing. The Oracle has a weird behavior in such a case. It seems sometimes the Oracle will allow such a capture, if it is the only option, but afterwards if the move is done, there's no capture.

[edit]

Rules examples

South to move

This leads to a repetition, and in human oware one seed is for North and two are for South, ie. a draw. The Oracle splits the seeds, and the result is 24.5 North to 23.5 South.

South to move

In this position, in edit mode, the Oracle says South has a +2 (captures and gets the remaining seeds), but making any of the moves leads to a non capture, and then it is North who gets a +2 (if South played 5, otherwise, a +6)

[edit]

Incidents

Due to technical problems (the server crashed and was difficult to revive) the Oracle was offline for nearly a year, from September 2007 to August 2008. The database was regenerated (computed again) in 2007^[5], using a wide-area network, and installed on a different server by Kees Verstoep.

[edit]

References

Donkers, J.: (2002) 'Comments on the Awari Solution', in ICGA Journal; 25 (3).

Romein, J. W. & Bal, H. E.: (2002) 'Awari is solved', in ICGA Journal; 25 (3).

Romein, J. W. & Bal, H. E.: (2003) 'Solving the Game of Awari using Parallel Retrograde Analysis,', in IEEE Computer, volume 36, number 10, pp. 26-33, October.

Verstoep, K., J. Maassen, H. E. Bal & J. W. Romein: (2008) 'Experiences with Fine-grained Distributed Supercomputing on a 10G Testbed', in 8th IEEE International Symposium on Cluster Computing and the Grid, 2008. CCGRID '08, 19-22 May 2008, pp. 376-383.

[edit]