Introducing Glass chess engine

[PK]

With two days to OpenWar it's high time to release Glass 0.1.0a - an UCI-compatibile chess engine, written by Edmund Moshammer and Pawel Koziol.

It is available at http://www.marittima.pl/glass and should play at the strength of 2250-2300 Elo (CCRL scale estimate), except for bullet games in which it is *much* worse.

Features:

* Bitboard infrastructure with a staged move generator
* Alpha-Beta search with null move, PVS, razoring and futility pruning
* Quiescence search with SEE and delta pruning
* Evaluation function centered around piece activity (as opposed to king attacks or pawn structure)
* Opening book (over 48.000 entries)
* Pondering

[gerold]

With two days to OpenWar it's high time to release Glass 0.1.0a - an UCI-compatibile chess engine, written by Edmund Moshammer and Pawel Koziol.

It is available at http://www.marittima.pl/glass and should play at the strength of 2250-2300 Elo (CCRL scale estimate), except for bullet games in which it is *much* worse.

Features:

* Bitboard infrastructure with a staged move generator
* Alpha-Beta search with null move, PVS, razoring and futility pruning
* Quiescence search with SEE and delta pruning
* Evaluation function centered around piece activity (as opposed to king attacks or pawn structure)
* Opening book (over 48.000 entries)
* Pondering

Thanks Pawel. Nice site and nice program.

Best to you,

Gerold.

[swami]

Thanks, Pawel. I hope you fix the time management bug atleast to make it work under 1/1 time conditions

[Graham Banks]

Thanks Pawel.

[Andres Valverde]

Congrats Edmund and Pawel !

[PK]

I had to update Glass Chess distribution packet because of a search bugfix.

sorry for that,

pawel koziol

[Graham Banks]

I had to update Glass Chess distribution packet because of a search bugfix.

sorry for that,

pawel koziol

Good that you picked it up early.

[trojanfoe]

Thanks! That's a nice engine and it successfully ripped through some test positions my own engine has been having problems with. I ran a set of mate tests though it through and found a move generator bug (below). Happy bug hunting!

Cheers,
Andy

Code: Select all

08:32:15 Processing ..\..\tests\mate.epd:9 'Mate in 2 Test 10'
+---------------+
|-|R|-|.|r|.|-|.| ply=0 (wtm)
|.|K|P|-|p|-|p|-| flags=
|p|P|-|P|b|.|P|p| ep=0
|p|B|.|P|B|-|p|n| hmclock=0
|n|N|-|.|p|q|r|.|
|P|-|.|-|.|p|N|k|
|P|.|-|.|P|.|-|.|
|.|-|b|-|Q|R|.|-|
+---------------+
bm Rh1
id Mate in 2 Test 10

08:32:15 glass> 'ucinewgame'
08:32:15 glass> 'isready'
08:32:15 <glass 'readyok'
08:32:15 glass> 'position fen 1R2r3/1KP1p1p1/pP1Pb1Pp/pB1PB1pn/nN2pqr1/P4pNk/P3P3/2b1QR2 w - - 0 1'
08:32:15 glass> 'go movetime 10000'
08:32:15 <glass 'info depth 1 currmove e5f4 currmovenumber 1 0'
08:32:24 <glass 'bestmove a4h4'
08:32:24 Failed to complete move 'Na4h4'
08:32:24 Failed to parse move 'a4h4'
08:32:24 glass> 'quit'

[Edmund]

Thanks! That's a nice engine and it successfully ripped through some test positions my own engine has been having problems with. I ran a set of mate tests though it through and found a move generator bug (below). Happy bug hunting!

Cheers,
Andy
[...]

Thanks Andy,
wasn't really a bug in the movegenerator, but the position you sent has so many possible captures and promotions in quiescence search that it took ages to search the first move. When the time ran out there was no bestmove present and thats why some random variable value was returned.

I sent an updated version to pawel to upload on the webpage. This one should fix the bug and solve the position in about 100sec.
That is still really slow, considering its just a mate in 2 problem, but please note that I am currently working on an updated version of quiescence (not ready for release yet), which takes about 0,2seconds to find the right move.

regards
Edmund

[PK]

bugfixed version 0.1.0c has been uploaded. thanks for running the test!