An idea: PotChess

[sje]

1) PotChess is not intended to be playable by humans. However, it should be tractable to programming by humans.

2) A PotChess(10) board display at one pixel per square (think color coding) will fill slightly more than half of an HD display. Beyond that, some scrolling or abstraction is required for viewing.

3) For large values of N, PotChess(N) pawns are near worthless and knights are only good as king defenders.

4) PotChess can be made arbitrarily fast by making the move time allocation arbitrarily small.

[Rémi Coulom]

PotChess is the name of a set of chess-like games. The name comes from "power of two chess"

When I first read the name of the game, I imagined something completely different

Rémi

[bhlangonijr]

PotChess is the name of a set of chess-like games. The name comes from "power of two chess"

When I first read the name of the game, I imagined something completely different

Rémi

Same with me. I thought it was the name of an engine from Amsterdam...

[Daniel Shawul]

No silly..after you make the move, you hit the pot instead of the clock

[bhlangonijr]

No silly..after you make the move, you hit the pot instead of the clock

http://www.dangerousminds.net/images/up ... _thumb.jpg

Check this out: PotChess set

[sje]

PotChess(10) has 1,024 squares on a side and 1,048,576 squares total.

Count by man per color:

Pawns: 1,024
Knights, bishops, and rooks: 2,048 each
Queens: 2,047
Kings: 1

Total per color: 9,216
Total both colors: 18,432

Maximum moves per man:

Pawn: 3
Knight: 8
Bishop: 2,045
Rook: 2,046
Queen: 4,091
King: 8

The total number of moves in an average middlegame position would be in the tens of millions.

A single bitboard would require 128 KB storage. A full attack map bitboard array would need 128 MB storage.

Man indexed vectors would require much less storage, but very long scan times.

The overall move selection process might be tractable with a few thousand computers working in a cluster. Each machine would be assigned a single man and would be responsible for move generation and other tasks specific to that man. Similarly, each machine could be assigned a small region of the board and be responsible for positional evaluation in that region.

[mike_bike_kite]

The idea is do re-invigorate computer chess research by advancing the computational complexity of the game beyond what can be handled using the old and tired techniques on current hardware.

There was recently an interesting graph on xkcd that plots the relative complexity of games for computers - obviously Chess, Go and Arimaa are all on there. I think my personal favourite was the penultimate game.