ArnoHu wrote:

birdracerthree wrote:

How does GoK apply stand pat when running check extensions in quiescence? I’m removing stand pat for checks and check evasions, but it is not working.

Standing pat is simply not applied then, and also GoK looks for a check evasion during quiescence search.

Standing pat is not applied during when you are in check and when you are checking the opponent, correct?

By the way, are there any types of puzzles you are interested in? I need more puzzles for the test positions study.

birdracerthree wrote:

ArnoHu wrote:

birdracerthree wrote:

How does GoK apply stand pat when running check extensions in quiescence? I’m removing stand pat for checks and check evasions, but it is not working.

Standing pat is simply not applied then, and also GoK looks for a check evasion during quiescence search.

Standing pat is not applied during when you are in check and when you are checking the opponent, correct?

By the way, are there any types of puzzles you are interested in? I need more puzzles for the test positions study.

Only for the side that is in check, on its next move. I give it a checkmate base evaluation, and it has to find a way out.

Great puzzles, yes, I already used them before.

ArnoHu wrote:

birdracerthree wrote:

ArnoHu wrote:

birdracerthree wrote:

How does GoK apply stand pat when running check extensions in quiescence? I’m removing stand pat for checks and check evasions, but it is not working.

Standing pat is simply not applied then, and also GoK looks for a check evasion during quiescence search.

Standing pat is not applied during when you are in check and when you are checking the opponent, correct?

By the way, are there any types of puzzles you are interested in? I need more puzzles for the test positions study.

Only for the side that is in check, on its next move. I give it a checkmate base evaluation, and it has to find a way out.

Great puzzles, yes, I already used them before.

So that’s only on check evasions? Is that equivalent to removing stand pat, or is there a difference?

I was asking if you need a different type of puzzle that is not in the study.

birdracerthree wrote:

ArnoHu wrote:

birdracerthree wrote:

ArnoHu wrote:

birdracerthree wrote:

How does GoK apply stand pat when running check extensions in quiescence? I’m removing stand pat for checks and check evasions, but it is not working.

Standing pat is simply not applied then, and also GoK looks for a check evasion during quiescence search.

Standing pat is not applied during when you are in check and when you are checking the opponent, correct?

By the way, are there any types of puzzles you are interested in? I need more puzzles for the test positions study.

Only for the side that is in check, on its next move. I give it a checkmate base evaluation, and it has to find a way out.

Great puzzles, yes, I already used them before.

So that’s only on check evasions? Is that equivalent to removing stand pat, or is there a difference?

I was asking if you need a different type of puzzle that is not in the study.

If you'd only remove standing pat, I think you would only escape by captures, not by evasions.

ArnoHu wrote:

birdracerthree wrote:

ArnoHu wrote:

birdracerthree wrote:

ArnoHu wrote:

birdracerthree wrote:

How does GoK apply stand pat when running check extensions in quiescence? I’m removing stand pat for checks and check evasions, but it is not working.

Standing pat is simply not applied then, and also GoK looks for a check evasion during quiescence search.

Standing pat is not applied during when you are in check and when you are checking the opponent, correct?

By the way, are there any types of puzzles you are interested in? I need more puzzles for the test positions study.

Only for the side that is in check, on its next move. I give it a checkmate base evaluation, and it has to find a way out.

Great puzzles, yes, I already used them before.

So that’s only on check evasions? Is that equivalent to removing stand pat, or is there a difference?

I was asking if you need a different type of puzzle that is not in the study.

If you'd only remove standing pat, I think you would only escape by captures, not by evasions.

So you have to lower alpha as well as remove standing pat, or do I only lower alpha?

birdracerthree wrote:

ArnoHu wrote:

birdracerthree wrote:

ArnoHu wrote:

birdracerthree wrote:

ArnoHu wrote:

birdracerthree wrote:

How does GoK apply stand pat when running check extensions in quiescence? I’m removing stand pat for checks and check evasions, but it is not working.

Standing pat is simply not applied then, and also GoK looks for a check evasion during quiescence search.

Standing pat is not applied during when you are in check and when you are checking the opponent, correct?

By the way, are there any types of puzzles you are interested in? I need more puzzles for the test positions study.

Only for the side that is in check, on its next move. I give it a checkmate base evaluation, and it has to find a way out.

Great puzzles, yes, I already used them before.

So that’s only on check evasions? Is that equivalent to removing stand pat, or is there a difference?

I was asking if you need a different type of puzzle that is not in the study.

If you'd only remove standing pat, I think you would only escape by captures, not by evasions.

So you have to lower alpha as well as remove standing pat, or do I only lower alpha?

I do not apply standing pat, and return a checkmate evaluation, if there is no better move found.

ArnoHu wrote:

birdracerthree wrote:

ArnoHu wrote:

birdracerthree wrote:

ArnoHu wrote:

birdracerthree wrote:

ArnoHu wrote:

birdracerthree wrote:

How does GoK apply stand pat when running check extensions in quiescence? I’m removing stand pat for checks and check evasions, but it is not working.

Standing pat is simply not applied then, and also GoK looks for a check evasion during quiescence search.

Standing pat is not applied during when you are in check and when you are checking the opponent, correct?

By the way, are there any types of puzzles you are interested in? I need more puzzles for the test positions study.

Only for the side that is in check, on its next move. I give it a checkmate base evaluation, and it has to find a way out.

Great puzzles, yes, I already used them before.

So that’s only on check evasions? Is that equivalent to removing stand pat, or is there a difference?

I was asking if you need a different type of puzzle that is not in the study.

If you'd only remove standing pat, I think you would only escape by captures, not by evasions.

So you have to lower alpha as well as remove standing pat, or do I only lower alpha?

I do not apply standing pat, and return a checkmate evaluation, if there is no better move found.

Alright, thank you. Hopefully I will get check extensions and en passant into Element fairly soon.

GoK is becoming too powerful. GoK 7.000 will be coming

I put the code of my bot on github, so even if this isn't scratch, it can now be used for test games : https://github.com/Azurus41/PyChess !
It now has an opening book, and you can export the FEN/pgn ( the pgn is in uci format, convert it here : https://www.dcode.fr/uci-chess-notation )
( It's not scratch but as my bot is now, it has a lower nps so it's a fair oponent )

AZURUS41 wrote:

I put the code of my bot on github, so even if this isn't scratch, it can now be used for test games : https://github.com/Azurus41/PyChess !
It now has an opening book, and you can export the FEN/pgn ( the pgn is in uci format, convert it here : https://www.dcode.fr/uci-chess-notation )
( It's not scratch but as my bot is now, it has a lower nps so it's a fair oponent )

My program now can do self-play !
Also I'm trying to implement endgame tables, do you have any good pesto table for this ?

AZURUS41 wrote:

I put the code of my bot on github, so even if this isn't scratch, it can now be used for test games : https://github.com/Azurus41/PyChess !
It now has an opening book, and you can export the FEN/pgn ( the pgn is in uci format, convert it here : https://www.dcode.fr/uci-chess-notation )
( It's not scratch but as my bot is now, it has a lower nps so it's a fair oponent )

Thx, is the bot running online somewhere? For a frontend you'd have to connect it to a UCI GUI? What is the expected rating?

I took a quick look at the code, from the perspective of performance I recommend switching to bitboards and bitwise ooerations, not to rely on strings for underlying data structures (instead use primitives like ints and arrays / lists of ints, maps where necessary), and avoid creating objects like Pieces during search (esp. in the innermost code), as this can cause a lot of overhead and object churn.

ArnoHu wrote:

ArnoHu wrote:

Epic battle between GoK (black) and White Dove, both at 10sec think time, 92% vs. 87% accuracy: https://lichess.org/XLJKeCN7#166

Rematch, good game with high accuracy, GoK wins with 97% vs. 92%: https://lichess.org/u5Q1AjXm#104

Switch to Scratch 3, GoK Medium vs. WD P2, back and forth during midgame, GoK slowly destroying pawn structure, ends up at 89% vs. 86% accuracy: https://lichess.org/Vp5uKhHW#131

Game #4 on TurboWarp, GoK wins at 94% vs. 90% accuracy: https://lichess.org/W5pPE6jG#130

I read a few pages about the CPW, and I can say that I learned a lot, which will help me for my bot

ababoin07 wrote:

I read a few pages about the CPW, and I can say that I learned a lot, which will help me for my bot

Have you verified this one?

GoK Blitz 2 (black) vs. Destructor, at move 18 GoK is up 6.6, and Destructor did not move any more (was checked at the time): https://lichess.org/vLuZ1jwy#36

ArnoHu wrote:

Great effort, 94% accuracy looks really good.

well I guess if we're just talking about accuracy, you should have a look at this:

SB dev (depth 4+0) vs Scurious 2.2 (5 ply) https://lichess.org/5la9DXU9#75

I found it completely by accident. it blundered it's queen in round 7 of HasiLover's 960 tournament because of a bug and after I fixed that a couple days after the tournament, I wanted to do a rematch with that position.
needless to say the first time I saw it completely crush Scurious my jaw dropped.
originally I had it set to 4+3, it got 97% accuracy. (as well as chess.com's game review thingy giving it 2 back to back brilliant moves!)
https://lichess.org/roqGzzwK
and from there I just kept on lowering DSDepth.

just a massive coincidence of course, but still incredibly cool!

-089- wrote:

ArnoHu wrote:

Great effort, 94% accuracy looks really good.

well I guess if we're just talking about accuracy, you should have a look at this:

SB dev (depth 4+0) vs Scurious 2.2 (5 ply) https://lichess.org/5la9DXU9#75

I found it completely by accident. it blundered it's queen in round 7 of HasiLover's 960 tournament because of a bug and after I fixed that a couple days after the tournament, I wanted to do a rematch with that position.
needless to say the first time I saw it completely crush Scurious my jaw dropped.
originally I had it set to 4+3, it got 97% accuracy. (as well as chess.com's game review thingy giving it 2 back to back brilliant moves!)
https://lichess.org/roqGzzwK
and from there I just kept on lowering DSDepth.

just a massive coincidence of course, but still incredibly cool!

Very impressive, esp. at relatively low search depth. What depth setting do you recommend for 10sec think time on TurboWarp? Do you think it could get to depth 4 or 5 AND quasi-unlimited quiescence search? If so, by which means?

PyChess 3 - Destructor : https://lichess.org/Fiy7LIwR
Both engines are making good progress, it was an alomst equal game until Destructor let me take his king for no reason after 40.Qe8+
How could I make my bot putting his rooks on open file more often ?

(I'm talking about this project.)

ArnoHu wrote:

What depth setting do you recommend for 10sec think time on TurboWarp?

I'm really not sure, it depends a lot on your system… even 4+0 sometimes takes 10 seconds for me in extreme cases.
I just added a highest and average move time variable to the debug menu, if you want to you can just try out some different settings and
have SB play on automatic for a couple moves (for example until it has developed all the minor pieces for both sides).
to do that, just go to settings, set “engine” to 2 and then change the turn, after that it'll start playing against itself. to stop it, press and hold x.
and well after that just press r to show move rankings & debug, it'll show the total time it has spent, the highest time and the average time it spent to make a move.
besides that, since I know that your system is faster than mine, my best guess would be 4+1. anything above that probably isn't really an option

ArnoHu wrote:

Do you think it could get to depth 4 or 5 AND quasi-unlimited quiescence search? If so, by which means?

no way.
oh also, even if dynamic search depth can in theory achieve all the things quiescence can, it is not a real quiescence search. it's just a 1 ply extension of the regular search for some of the nodes. it is comically slow sometimes.
like really slow.

as an example, here's the first one of the ‘mate-in-scenarios’ from 2 days ago
r2qr1k1/pp3p1p/1b3p1B/2p1p3/2B1P3/2PP1P1b/P4P1P/R2QR2K w - - 3 16
it finds M7 with 1+13 in 7.5s, as compared to GoK with 0.06s

ArnoHu wrote:

ababoin07 wrote:

I read a few pages about the CPW, and I can say that I learned a lot, which will help me for my bot

Have you verified this one?

GoK Blitz 2 (black) vs. Destructor, at move 18 GoK is up 6.6, and Destructor did not move any more (was checked at the time): https://lichess.org/vLuZ1jwy#36

In reality, I resumed the courses and I do not have too much time to update m'y chess engine, I think I would check on the old versions if there is the bug and I would advise accordingly wednesday if I have the Time else friday (I don't use a translator so m'y english is Bad)

I just managed to run GoK as an Android app on my mobile phone (via TurboWarp packager, then open it as embedded .html in Android WebViewer). It might even end up on Google PlayStore.

ArnoHu wrote:

I just managed to run GoK as an Android app on my mobile phone (via TurboWarp packager, then open it as embedded .html in Android WebViewer). It might even end up on Google PlayStore.

That would be incredible! Just make sure GoK is ready for human opponents . We don’t need GoK playing Nf3-Ng1-Nf3-Ng1

Are you planning to do this?