Scratch Chess Engine - Game of Kings

AZURUS41

birdracerthree wrote:
S_P_A_R_T_Test wrote:
AZURUS41 wrote:
https://scratch-mit-edu.ezproxy.canberra.edu.au/projects/1073860232
Glad to announce LA timed is here !
This version WON to ECE 6+8 this morning in a back-and-forth game, with some blunder from both sides, but it did it !

Except is an unknown bug is found in the following days, this version will replace the main one on both the studio and the current olympic for which France is the next team to play btw… so can you all test it out ?

Congrats! Beating ECE 6+8 is no small achievement
Haha, you would know…

Jokes aside, LA’s improvement has been incredible over less than 2 months. Congrats! Is Element falling off?

no

for sure no lmao, that one bushop just got horizon effected for 20+ moves and LA got lucky I guess

birdracerthree

ArnoHu wrote:
The Scratch 3 Runtime Chess Engine Olympics (team event) currently takes place, you can watch it here:

https://lichess.org/broadcast/scratch-chess-engine-olympics-october-2024/sce-olympics-round-1/S2lEhC3G#teams

Participants:
Austria
Germany 1
Germany 2
France
Canada
United States
Asia
Australia

More games are occurring live!

Hello, I am @birdracerthree , the creator of the fourth strongest chess engine on Scratch, Element
Element’s approximate rating is 1800 FIDE or 2050 chesscom.

AZURUS41

Olympics final standings :
1 - Austria
2 - Germany
3 - French Canada
4 - France
5 - USA
6 - Germany 2
7 - Australia
8 - Asia

birdracerthree

Just for fun, I wonder which engines can checkmate with three knights

Last edited by birdracerthree (Oct. 13, 2024 17:46:36)

Hello, I am @birdracerthree , the creator of the fourth strongest chess engine on Scratch, Element
Element’s approximate rating is 1800 FIDE or 2050 chesscom.

AZURUS41

v1.54 LA is out ( the 1.53 stood for 2 minutes lmao ) !

There is now a partial 2 fold detection, preventing LA to play 2 times the same move in the same position if he thinks it has an higher than 50 centipawns advantage. There is also a huge improvement in time management, helping playing better in quick-paced games like blitz ones.
Example game where LA wins against a 1500 FIDE friend of mine in 4+2 time control : https://lichess.org/PyqeI0yj

As a quick side note, it now converts the endgame who ended in a 3-fold against ECE 6+8 yesterday

Last edited by AZURUS41 (Oct. 15, 2024 09:07:58)

ababoin07

AZURUS41 wrote:
v1.54 LA is out ( the 1.53 stood for 2 minutes lmao ) !

There is now a partial 2 fold detection, preventing LA to play 2 times the same move in the same position if he thinks it has an higher than 50 centipawns advantage. There is also a huge improvement in time management, helping playing better in quick-paced games like blitz ones.
Example game where LA wins against a 1500 FIDE friend of mine in 4+2 time control : https://lichess.org/PyqeI0yj

As a quick side note, it now converts the endgame who ended in a 3-fold against ECE 6+8 yesterday

I forgot this forum, lol, you can divide the eval by 2 if itns the 2nd rime and set the eval to 0 if itns the 3rd time, it's better

birdracerthree

https://lichess.org/broadcast/10sec-scratch-chess-engine-tournament/10sec-scratch-chess-final/lW9ytYmT#boards

Hello, I am @birdracerthree , the creator of the fourth strongest chess engine on Scratch, Element
Element’s approximate rating is 1800 FIDE or 2050 chesscom.

ArnoHu

CONGRATULATIONS TO WHITE DOVE (P2), for its first win against GoK Classic (Medium, white) on TW and standard chess, playing a near-perfect game (one mistake in final endgame stage when everything was over), 97% vs. 93%. After many close-to-win matches, this victory indeed is well deserved! It also uncovered a GoK Classic bug, that led it to accept a positional disadvantageous RBP-vs-Q trade:

https://lichess.org/rHv9rnH8#126

The rematch, same opening, updated GoK Classic (white), 97% vs. 90% accuracy:

https://lichess.org/BrmpDAwR#111

HasiLover

ArnoHu wrote:
CONGRATULATIONS TO WHITE DOVE (P2), for its first win against GoK Classic (Medium, white) on TW and standard chess, playing a near-perfect game (one mistake in final endgame stage when everything was over), 97% vs. 93%. After many close-to-win matches, this victory indeed is well deserved! It also uncovered a GoK Classic bug, that led it to accept a positional disadvantageous RBP-vs-Q trade:

https://lichess.org/rHv9rnH8#126

The rematch, same opening, updated GoK Classic (white), 97% vs. 90% accuracy:

https://lichess.org/BrmpDAwR#111

impressive Games.

ArnoHu

Important update:

GoK NNUE is now running as the main GoK project at https://scratch-mit-edu.ezproxy.canberra.edu.au/projects/148769358/ The previous NNUE stable project became NNUE dev project.

GoK Classic has moved to its previous dev project location: https://scratch-mit-edu.ezproxy.canberra.edu.au/projects/828094886/

ArnoHu

HasiLover wrote:
ArnoHu wrote:
CONGRATULATIONS TO WHITE DOVE (P2), for its first win against GoK Classic (Medium, white) on TW and standard chess, playing a near-perfect game (one mistake in final endgame stage when everything was over), 97% vs. 93%. After many close-to-win matches, this victory indeed is well deserved! It also uncovered a GoK Classic bug, that led it to accept a positional disadvantageous RBP-vs-Q trade:

https://lichess.org/rHv9rnH8#126

The rematch, same opening, updated GoK Classic (white), 97% vs. 90% accuracy:

https://lichess.org/BrmpDAwR#111
impressive Games.

I think so too, and what can you do against an opponent running at 97% accuracy (or actually 100% until move 46).

Last edited by ArnoHu (Yesterday 06:36:08)

ArnoHu

Scratch Chess Engine Olympics October 2024

The Scratch Chess Engine Olympics October 2024 Tournament at

https://lichess.org/broadcast/scratch-chess-engine-olympics-october-2024/sce-olympics-super-cup-semifinal-and-final/HzPo7LKG#teams

ended with the following results:

Champions Cup:
1. Austria
2. Germany

Conference Cup:
1. Canada
2. France

Super Cup:
1. Austria
2. Germany

ArnoHu

10-Seconds Scratch Chess Engine Tournament

GoK NNUE won the 10-Seconds Scratch Chess Engine Tournament at

https://lichess.org/broadcast/10sec-scratch-chess-engine-tournament/10sec-scratch-chess-final/lW9ytYmT#boards

Final results:

1. GoK NNUE
2. GoK Classic
3. White Dove
3. Shallow Blue

5. Scurious
5. LA Chess
5. Delta
5. Element

Last edited by ArnoHu (Yesterday 07:03:37)

ArnoHu

5-Seconds Blitz Scratch Chess Engine Tournament

And finally, one more recent tournament:

https://lichess.org/broadcast/5-seconds-blitz-chess-scratch-engine-tournament-october-2024/blitz-chess-scratch-engines-round-6/ETVqvKlZ#players

Results:

1. GoK NNUE
2. GoK Classic
3. White Dove
4. Delta
5. LA Chess

ArnoHu

GoK NNUE (Medium (!), white, TurboWarp experimental version) wins against Komodo 24 (chess.com “super grandmaster”) in a perfect game, with 99% vs. 94%: https://lichess.org/pDuaRaE7#125

Last edited by ArnoHu (Yesterday 12:17:16)

birdracerthree

I have shared the old transposition table dev if anyone wants to take a look

Hello, I am @birdracerthree , the creator of the fourth strongest chess engine on Scratch, Element
Element’s approximate rating is 1800 FIDE or 2050 chesscom.

MoosenMooALot

ArnoHu wrote:
GoK NNUE (Medium (!), white, TurboWarp experimental version) wins against Komodo 24 (chess.com “super grandmaster”) in a perfect game, with 99% vs. 94%: https://lichess.org/pDuaRaE7#125

Nice! I'm surprised by how Komodo24 blundered its queen.

ArnoHu

MoosenMooALot wrote:
ArnoHu wrote:
GoK NNUE (Medium (!), white, TurboWarp experimental version) wins against Komodo 24 (chess.com “super grandmaster”) in a perfect game, with 99% vs. 94%: https://lichess.org/pDuaRaE7#125
Nice! I'm surprised by how Komodo24 blundered its queen.

GoK also won the next one: https://lichess.org/study/ldM7RmXh/IHmynOhY

Discuss Scratch