Quantum computer chess

STILL EXPERIMENTING. STILL UNDER CONSTRUCTION.

I've been experimenting with a simulation of a 256 qubit quantum computer playing chess.

At present the sim is set up to reflect board positions and present an evaluation.

Using probability amplification board positions relevant to the current state of play are measured/actualized and a classical back-end selects the best move to play.

It's just a randomized search with some possibly-tricky statistics to speed it up (order the search items in some kind of reasonable way to find good items first) and some data-mining to tune the eval procedure being used.

But there are a lot of tricks.

Initially I was playing against Crafty, but now "Quantum" has played against gnuchess and GroKo as well from black and white. It manages to win a few matches allowing 10 sec of cpu time per move for the sim.

This is much improved on 1 wk ago when a single QC sim for 1 move could take 1 hr of cpu or more.

I'm planning to tinker with the current setup for a few more days before moving on to very strange areas where the QC doesn't manipulate the positions/game boards, but sequences of moves from a given position.

This will allow some human-like behaviour to emerge, hopefully, because superpositions of programs (i.e. sequences of moves for white, sequence of moves for black) allow plans to be formed which have parts that are not decided yet. E.g. the opponent's plans are largely unknown at the start, but as moves are received parts of the QC wave-function can be "popped" to indicate those things have "actualized". OTOH your own plans can be nebulous to allow for unexpected things developing in the future; as moves are required from the program the next-move can be actualized but leave the rest of the plan flexible, to allow for any subsequent moves from the opponent.

Archives games between "Quantum" and other chess programs, in approx order played over past 2 wks

Games are in PGN format and should "play" themselves if your machine is configured with a chess GUI like xboard (linux/UNIX).

Some games are very early and are generally bloodbaths for "Quantum" (called in early days "myserver" or QCS).

Later games show some promise, but some end when the Quantum program got a segfault or generated a bad move.

The latest v1.10 version of Quantum generally understands enough chess to not generate bad moves, but segfaulting is still on the "todo" list.

The list includes a few wins against GreKo, gnuchess and Crafty. As I discover their source codes, the list will grow -- one way or another.

root-GNU_Chess_5.07.pgn
myserver-Crafty-23.4.pgn
myserver-Crafty-23.4x10.pgn
QCS_v0.5-Crafty-23.4.pgn
QCS_v0.6-Crafty-23.4.pgn
QCS_v0.7-Crafty-23.4.pgn
2.QCS_v0.7-Crafty-23.4.pgn
3.QCS_v0.7-Crafty-23.4.pgn
QCS_v1.8-Crafty-23.4.pgn
QCS_v1.8-GNU_Chess_5.07.pgn
QCS_v1.9-GreKo_12.0.pgn
FirstMate.QCS_v1.9-GNU_Chess_5.07.pgn
FirstMate.QCS_v1.9-Crafty-23.4.pgn
FirstMate.QCS_v1.9-GreKo_12.0.pgn
Mate2.QCS_v1.10-Crafty-23.4.pgn
FirstAsBlack.Crafty-23.4-QCS_v1.10.pgn
EPCaptureScrewedMe.GreKo_12.0-QCS_v1.10.pgn
QCSGotSegfault.GreKo_12.0-QCS_v1.10.pgn
AnotherSegfault.GNU_Chess_5.07-QCS_v1.10.pgn
MateAsBlack-GNU_Chess_5.07-QCS_v1.10.pgn
PlayingWithMyFood.Crafty-23.4-QCS_v1.11.pgn
MateAsBlack.GreKo_12.0-QCS_v1.11.pgn
BlackWins.QCS_v1.11-QCS_v1.11.pgn
GoodTussle.Crafty-23.4-QCS_v1.11.pgn
Mate46.GreKo_12.0-QCS_v1.12.pgn
Mate83.QCS_v1.12-Crafty-23.4.pgn
QCS_v1.12-phalanx.pgn
phalanx-QCS_v1.12.pgn


Kym Horsell /
Kym@KymHorsell.COM

ADVISORY: Email to these sites is filtered. Unsolicited or spoof email may be automajically re-directed to the relevant postmaster. SPOOFS AND CREDIT FRAUD ARE REPORTED IMMEDIATELY TO THE RELEVANT AUTHORITIES.