Destructor_chess wrote:

coooolboy55 wrote:

I have stopped working on my Chess AI. I really don't like creating it.

Ah… You can ask others, I said that too many times. But well, chess programming is too much attractive, you'll see.

coooolboy55 wrote:

Destructor_chess wrote:

coooolboy55 wrote:

I have stopped working on my Chess AI. I really don't like creating it.

Ah… You can ask others, I said that too many times. But well, chess programming is too much attractive, you'll see.

Yes. I am planning to create one in a text-based programming language after learning one. (I don't know how that will turn out, as I usually set my goals too high)

ArnoHu wrote:

GoK Classic vs. Wolverine 2, 87% vs. 90% (draw by stalemate after GoK lead, caused by endgame horizon effects due to check sequences and looming Wolverine promotion):
https://lichess.org/study/PJTsaIt9/iHyqy9jv

GoK Classic vs. Wolverine 2, 94% vs. 91%:
https://lichess.org/study/PJTsaIt9/ZBOc7qEL

GoK Classic vs. White Dove , 97% vs. 93%:
https://lichess.org/study/PJTsaIt9/zRVjzS92

GoK Classic vs. GoK NNUE , 91% vs. 97%:
https://lichess.org/study/PJTsaIt9/ZBOc7qEL

GoK NNUE vs. Black Crow, 98% vs. 92%:
https://lichess.org/study/PJTsaIt9/GODLGJ5F

ArnoHu wrote:

GoK Classic vs. Wolverine 2, 87% vs. 90% (draw by stalemate after GoK lead, caused by endgame horizon effects due to check sequences and looming Wolverine promotion):
https://lichess.org/study/PJTsaIt9/iHyqy9jv

waabooboo wrote:

ArnoHu wrote:

GoK Classic vs. Wolverine 2, 87% vs. 90% (draw by stalemate after GoK lead, caused by endgame horizon effects due to check sequences and looming Wolverine promotion):
https://lichess.org/study/PJTsaIt9/iHyqy9jv

87% vs 90% is pretty generous by Lichess, that game was a mess…

7r/2NP4/8/1PPk4/P7/6p1/1K6/8 b - - 2 70 Wolverine actually has a win here, Stockfish gives this crazy mating net with Kc4 Ne8 g2 d8=Q Rh2. I can't blame Wolverine for not finding that. But I will blame Wolverine for that atrocious opening play…

ArnoHu wrote:

waabooboo wrote:

ArnoHu wrote:

GoK Classic vs. Wolverine 2, 87% vs. 90% (draw by stalemate after GoK lead, caused by endgame horizon effects due to check sequences and looming Wolverine promotion):
https://lichess.org/study/PJTsaIt9/iHyqy9jv

87% vs 90% is pretty generous by Lichess, that game was a mess…

7r/2NP4/8/1PPk4/P7/6p1/1K6/8 b - - 2 70 Wolverine actually has a win here, Stockfish gives this crazy mating net with Kc4 Ne8 g2 d8=Q Rh2. I can't blame Wolverine for not finding that. But I will blame Wolverine for that atrocious opening play…

And I thought only GoK Classic has opening issues

waabooboo wrote:

ArnoHu wrote:

waabooboo wrote:

ArnoHu wrote:

GoK Classic vs. Wolverine 2, 87% vs. 90% (draw by stalemate after GoK lead, caused by endgame horizon effects due to check sequences and looming Wolverine promotion):
https://lichess.org/study/PJTsaIt9/iHyqy9jv

87% vs 90% is pretty generous by Lichess, that game was a mess…

7r/2NP4/8/1PPk4/P7/6p1/1K6/8 b - - 2 70 Wolverine actually has a win here, Stockfish gives this crazy mating net with Kc4 Ne8 g2 d8=Q Rh2. I can't blame Wolverine for not finding that. But I will blame Wolverine for that atrocious opening play…

And I thought only GoK Classic has opening issues

Ha, well Wolverine doesn't even have an opening book, probably it wouldn't hurt to add one. But I feel like a strong program should be able to figure out the opening on the fly – bad opening play is indicative of bad understanding more generally…

r1bqkr2/pp1npp2/3p2p1/3P2Pp/3R1Pn1/2N2Q2/PPP4P/2K2BNR w q - 1 14
I mean, this is silly stuff.

0.265: 5 : Search start, depth = 5
0.329: 5 : 90 : -266
0.335: 5 : 418 : -248
0.337: PV: [FEN "r1bqk2r/pp1npp2/3p2p1/3P2Pp/3R1Pn1/2N2Q2/PPP4P/2K2BNR b"] Qd8b6 Ng1e2 Nd7c5 h2h3 e7e5 d5e6 Nc5xe6 

ArnoHu wrote:

waabooboo wrote:

ArnoHu wrote:

waabooboo wrote:

ArnoHu wrote:

GoK Classic vs. Wolverine 2, 87% vs. 90% (draw by stalemate after GoK lead, caused by endgame horizon effects due to check sequences and looming Wolverine promotion):
https://lichess.org/study/PJTsaIt9/iHyqy9jv

87% vs 90% is pretty generous by Lichess, that game was a mess…

7r/2NP4/8/1PPk4/P7/6p1/1K6/8 b - - 2 70 Wolverine actually has a win here, Stockfish gives this crazy mating net with Kc4 Ne8 g2 d8=Q Rh2. I can't blame Wolverine for not finding that. But I will blame Wolverine for that atrocious opening play…

And I thought only GoK Classic has opening issues

Ha, well Wolverine doesn't even have an opening book, probably it wouldn't hurt to add one. But I feel like a strong program should be able to figure out the opening on the fly – bad opening play is indicative of bad understanding more generally…

r1bqkr2/pp1npp2/3p2p1/3P2Pp/3R1Pn1/2N2Q2/PPP4P/2K2BNR w q - 1 14
I mean, this is silly stuff.

OK, at least Classic finds a decent move at depth 5 (if that's what you mean?)

0.265: 5 : Search start, depth = 5
0.329: 5 : 90 : -266
0.335: 5 : 418 : -248
0.337: PV: [FEN "r1bqk2r/pp1npp2/3p2p1/3P2Pp/3R1Pn1/2N2Q2/PPP4P/2K2BNR b"] Qd8b6 Ng1e2 Nd7c5 h2h3 e7e5 d5e6 Nc5xe6 

ArnoHu wrote:

ArnoHu wrote:

GoK Classic vs. Wolverine 2, 87% vs. 90% (draw by stalemate after GoK lead, caused by endgame horizon effects due to check sequences and looming Wolverine promotion):
https://lichess.org/study/PJTsaIt9/iHyqy9jv

GoK Classic vs. Wolverine 2, 94% vs. 91%:
https://lichess.org/study/PJTsaIt9/ZBOc7qEL

GoK Classic vs. White Dove , 97% vs. 93%:
https://lichess.org/study/PJTsaIt9/zRVjzS92

GoK Classic vs. GoK NNUE, 91% vs. 97%:
https://lichess.org/study/PJTsaIt9/ZBOc7qEL

GoK NNUE vs. Black Crow, 98% vs. 92%:
https://lichess.org/study/PJTsaIt9/GODLGJ5F

Alternative ending for game #1 after adding extensions and preventing reductions on dangerous passed pawn pushes:

GoK Classic vs. Wolverine 2, 98% vs. 95%:
https://lichess.org/study/PJTsaIt9/UyHM8ytZ

coooolboy55 wrote:

Do you think a perfect Connect 4 AI will ever be possible on scratch?

waabooboo wrote:

coooolboy55 wrote:

Do you think a perfect Connect 4 AI will ever be possible on scratch?

Well sure, it's a solved game after all. I could do it by adding an opening book to my Connect 4 program, if I had the time and the interest. Probably a few thousand memorized positions would be enough, and the proof number search would figure the rest out from there.

coooolboy55 wrote:

Do you think a perfect Connect 4 AI will ever be possible on scratch?

ArnoHu wrote:

coooolboy55 wrote:

Do you think a perfect Connect 4 AI will ever be possible on scratch?

I highly doubt it. While perfect solutions have been implemented,they rely on a full-coverage opening book for the first moves, which is orders of magnitude larger than what would fit into a Scratch project.

waabooboo wrote:

ArnoHu wrote:

coooolboy55 wrote:

Do you think a perfect Connect 4 AI will ever be possible on scratch?

I highly doubt it. While perfect solutions have been implemented,they rely on a full-coverage opening book for the first moves, which is orders of magnitude larger than what would fit into a Scratch project.

Are we talking about a strong solution or a weak solution?

If the task is just “computer goes first and always wins”, that seems quite a bit easier than a full-coverage opening book that would solve any position, including ones that result from suboptimal play…

ArnoHu wrote:

GoK NNUE vs. Black Crow, 97% vs. 90%:
https://lichess.org/study/PJTsaIt9/C8DKkH6w

GoK Classic vs. Wolverine 2, 97% vs. 92%:
https://lichess.org/study/PJTsaIt9/gDzJWeCE