Scratch Chess Engine - Game of Kings

qwertycodechamp90411

ArnoHu wrote:
qwertycodechamp90411 wrote:
Something I thought of, it would be nice to be able to disable castling in the FEN import on GoK

Yes, a known issue, will fix it soon.

When that gets fixed I'll try GoK against FStockfish with 2 kings (to see how it goes at unorthodox chess). I have played a few games against it before but it always finds a way to skewer my two kings or something lol

birdracerthree

qwertycodechamp90411 wrote:
ArnoHu wrote:
qwertycodechamp90411 wrote:
Something I thought of, it would be nice to be able to disable castling in the FEN import on GoK

Yes, a known issue, will fix it soon.
When that gets fixed I'll try GoK against FStockfish with 2 kings (to see how it goes at unorthodox chess). I have played a few games against it before but it always finds a way to skewer my two kings or something lol

Unfortunately, I believe that GoK does not support these types of game. However, I believe White Dove HCE handles these very well. Where we you finding these Stockfish variants?

qwertycodechamp90411

birdracerthree wrote:
qwertycodechamp90411 wrote:
ArnoHu wrote:
qwertycodechamp90411 wrote:
Something I thought of, it would be nice to be able to disable castling in the FEN import on GoK

Yes, a known issue, will fix it soon.
When that gets fixed I'll try GoK against FStockfish with 2 kings (to see how it goes at unorthodox chess). I have played a few games against it before but it always finds a way to skewer my two kings or something lol
Unfortunately, I believe that GoK does not support these types of game. However, I believe White Dove HCE handles these very well. Where we you finding these Stockfish variants?

1) It does with a FEN import
2) https://fairyground.vercel.app/advanced.html
Create a text file with the following contents

[cfour]
maxRank = 6
maxFile = 7
immobile = p
startFen = 7/7/7/7/7/7[PPPPPPPPPPPPPPPPPPPPPppppppppppppppppppppp] w - - 0 1
pieceDrops = true
enclosingDrop = top
doubleStep = false
castling = false
stalemateValue = draw
immobilityIllegal = false
connectN = 4
nMoveRule = 0

(What this does: 6 ranks, 7 files, pawns can't move, start with blank board and 21 pawns each to drop, you can drop pieces, they can only be dropped on top of each other, no double step, castling, stalemate is draw, getting into stalemate is allowed, connect 4, no 50moverule. Basically connect4 but with pawns as the pieces)

Then on the website, click ‘choose file’ next to variants.ini on the top of the page
The select Connect 4 in the dropdown next to “Name”
For the 2 kings variant, import the following

[twokings:extinction]
startFen = rnbkkbnr/pppppppp/8/8/8/8/PPPPPPPP/RNBKKBNR
extinctionPieceTypes = k
extinctionPieceCount = 1
extinctionPseudoRoyal = true

(What this does: replace the queen with king, if you reach 1 king you lose, treat losing next move as check and checkmate)

All functions: https://github.com/fairy-stockfish/Fairy-Stockfish/blob/master/src/variants.ini

To create your own pieces, use Betza notation: https://www.gnu.org/software/xboard/Betza.html

Example: No double step pawn is fmWfcF (fm - forward move, W - 1 square orthogonally, fc - forward capture, F - 1 square diagonally i.e. move forward 1 square, capture 1 square diagonally)

Lowercase letters are for directions (f = forward, b = backward, etc.) or move/capture (m = only move, not capture, c = capture only). Uppercase is for the actual movement (e.g. N = Knight, F = 1 square diagonally, F0 = B = as far as you want diagonally, F2 = 1 or 2 squares diagonally)

I believe the X-Board specific ones work on the website (I've seen i used).

Last edited by qwertycodechamp90411 (Jan. 7, 2025 21:57:05)

qwertycodechamp90411

Is there any way to get these scratch bots a bot account on lichess.org?

qwertycodechamp90411

Also how do you get your rating approximations for your chess engines?

ababoin07

qwertycodechamp90411 wrote:
Also how do you get your rating approximations for your chess engines?

How do you get means you would like know that, or you mean how to calculate it? It’s base with a looooot of game between the s chess engines, but it’s approximative…

qwertycodechamp90411

ababoin07 wrote:
qwertycodechamp90411 wrote:
Also how do you get your rating approximations for your chess engines?
How do you get means you would like know that, or you mean how to calculate it? It’s base with a looooot of game between the s chess engines, but it’s approximative…

Which chess engines do you match the bot against?

AZURUS41

https://lichess.org/7BiA220h#0
LA on the verge of winning to Verde ! Except it's LA…
1.88, which includes a new book, allowed the engine to reach its all-time highest nps on an endgame with sliding pieces : 165039 nps, for a total of 3.3 nodes in only 20 seconds !
Now I need to way to prevent stalemate

ababoin07

AZURUS41 wrote:
https://lichess.org/7BiA220h#0
LA on the verge of winning to Verde ! Except it's LA…
1.88, which includes a new book, allowed the engine to reach its all-time highest nps on an endgame with sliding pieces : 165039 nps, for a total of 3.3 nodes in only 20 seconds !
Now I need to way to prevent stalemate

I don't understand “165039 nps, for a total of 3.3 nodes in only 20 seconds”…

WhiteDoveOfficial

ababoin07 wrote:
AZURUS41 wrote:
https://lichess.org/7BiA220h#0
LA on the verge of winning to Verde ! Except it's LA…
1.88, which includes a new book, allowed the engine to reach its all-time highest nps on an endgame with sliding pieces : 165039 nps, for a total of 3.3 nodes in only 20 seconds !
Now I need to way to prevent stalemate
I don't understand “165039 nps, for a total of 3.3 nodes in only 20 seconds”…

I'm assuming that it's supposed to be 3.3M nodes in only 20 seconds.

(165039 nodes/seconds * 20 seconds = 3,300,780 nodes)

ababoin07

WhiteDoveOfficial wrote:
ababoin07 wrote:
AZURUS41 wrote:
https://lichess.org/7BiA220h#0
LA on the verge of winning to Verde ! Except it's LA…
1.88, which includes a new book, allowed the engine to reach its all-time highest nps on an endgame with sliding pieces : 165039 nps, for a total of 3.3 nodes in only 20 seconds !
Now I need to way to prevent stalemate
I don't understand “165039 nps, for a total of 3.3 nodes in only 20 seconds”…

I'm assuming that it's supposed to be 3.3M nodes in only 20 seconds.

(165039 nodes/seconds * 20 seconds = 3,300,780 nodes)

Oh!!!! Yeah, it makes sense

AZURUS41

ababoin07 wrote:
WhiteDoveOfficial wrote:
ababoin07 wrote:
AZURUS41 wrote:
https://lichess.org/7BiA220h#0
LA on the verge of winning to Verde ! Except it's LA…
1.88, which includes a new book, allowed the engine to reach its all-time highest nps on an endgame with sliding pieces : 165039 nps, for a total of 3.3 nodes in only 20 seconds !
Now I need to way to prevent stalemate
I don't understand “165039 nps, for a total of 3.3 nodes in only 20 seconds”…

I'm assuming that it's supposed to be 3.3M nodes in only 20 seconds.

(165039 nodes/seconds * 20 seconds = 3,300,780 nodes)
Oh!!!! Yeah, it makes sense

Oh yeah indeed, I forgot the “M”.
Still impressive numbers over here !
Other LA - Verde game, this time LA wins due to almost perfect play (and also Verde doesn't hendle en passant bruh) : https://lichess.org/dxVKxjYR#0
About the rankings, I think Verde will be put down to 7th place below ECE.

ababoin07

AZURUS41 wrote:
ababoin07 wrote:
WhiteDoveOfficial wrote:
ababoin07 wrote:
AZURUS41 wrote:
https://lichess.org/7BiA220h#0
LA on the verge of winning to Verde ! Except it's LA…
1.88, which includes a new book, allowed the engine to reach its all-time highest nps on an endgame with sliding pieces : 165039 nps, for a total of 3.3 nodes in only 20 seconds !
Now I need to way to prevent stalemate
I don't understand “165039 nps, for a total of 3.3 nodes in only 20 seconds”…

I'm assuming that it's supposed to be 3.3M nodes in only 20 seconds.

(165039 nodes/seconds * 20 seconds = 3,300,780 nodes)
Oh!!!! Yeah, it makes sense
Oh yeah indeed, I forgot the “M”.
Still impressive numbers over here !
Other LA - Verde game, this time LA wins due to almost perfect play (and also Verde doesn't hendle en passant bruh) : https://lichess.org/dxVKxjYR#0
About the rankings, I think Verde will be put down to 7th place below ECE.

Really cool! I ́m currently working on a completly new chess engine (NNUE (768+x (x<15), 128, 64, 1), non-list use for the minimax, NO str on the code (on the AI, not on the fen handling, yeah…)…)

Last edited by ababoin07 (Jan. 8, 2025 17:51:28)

ArnoHu

qwertycodechamp90411 wrote:
Is there any way to get these scratch bots a bot account on lichess.org?

Possibly, but I don't think anyone has done that yet. Best chance IMHO is TurboWarp with a TW extension.

ArnoHu

qwertycodechamp90411 wrote:
Also how do you get your rating approximations for your chess engines?

There is an outdated ranking which was created using the BayesELO tool and 60+ Scratch chess engine games: https://scratch-mit-edu.ezproxyberklee.flo.org/discuss/post/8074502/

Last edited by ArnoHu (Jan. 8, 2025 23:54:02)

ArnoHu

The Connect-4 Engine Studio can be found here: https://scratch-mit-edu.ezproxyberklee.flo.org/studios/36106829/

lucastezino

ababoin07 wrote:
AZURUS41 wrote:
ababoin07 wrote:
WhiteDoveOfficial wrote:
ababoin07 wrote:
AZURUS41 wrote:
https://lichess.org/7BiA220h#0
LA on the verge of winning to Verde ! Except it's LA…
1.88, which includes a new book, allowed the engine to reach its all-time highest nps on an endgame with sliding pieces : 165039 nps, for a total of 3.3 nodes in only 20 seconds !
Now I need to way to prevent stalemate
I don't understand “165039 nps, for a total of 3.3 nodes in only 20 seconds”…

I'm assuming that it's supposed to be 3.3M nodes in only 20 seconds.

(165039 nodes/seconds * 20 seconds = 3,300,780 nodes)
Oh!!!! Yeah, it makes sense
Oh yeah indeed, I forgot the “M”.
Still impressive numbers over here !
Other LA - Verde game, this time LA wins due to almost perfect play (and also Verde doesn't hendle en passant bruh) : https://lichess.org/dxVKxjYR#0
About the rankings, I think Verde will be put down to 7th place below ECE.
Really cool! I ́m currently working on a completly new chess engine (NNUE (768+x (x<15), 128, 64, 1), non-list use for the minimax, NO str on the code (on the AI, not on the fen handling, yeah…)…)

what???

lucastezino

ArnoHu wrote:
The Connect-4 Engine Studio can be found here: https://scratch-mit-edu.ezproxyberklee.flo.org/studios/36106829/

when this sprite clicked

change [v] by ([sqrt] of ((6) * (50))::operators)

ArnoHu

GoK Classic (Medium, white) vs. Delta (15s), 93% vs. 90%:
https://lichess.org/2W9J4RnE#141

GoK Classic (Medium, white) vs. White Dove (P2), 94% vs. 89%:
https://lichess.org/WlTeIsZ3#131

ababoin07

lucastezino wrote:
ArnoHu wrote:
The Connect-4 Engine Studio can be found here: https://scratch-mit-edu.ezproxyberklee.flo.org/studios/36106829/
when this sprite clicked
change [v] by ([sqrt] of ((6) * (50))::operators)

Please don’t spam this kind of message

Discuss Scratch