Amazing results of LeelaKnightOdds-Dev

Discussion of anything and everything relating to chess playing software and machines.

Moderators: hgm, chrisw, Rebel

Father
Posts: 1664
Joined: Sun Mar 19, 2006 4:39 am
Location: Colombia
Full name: Pablo Ignacio Restrepo

Re: Amazing results of LeelaKnightOdds-Dev

Post by Father »

Marcus91 wrote: Tue Nov 12, 2024 12:10 am
Uri Blass wrote: Mon Nov 11, 2024 10:14 pm today not a good day against lc0 and I won only 3.5-2.5

Note that I did not do serious analysis to prepare but my conclusion that if I get a good position lc0 allow easily to trade pieces so serious preperation should be about getting a good position to learn from my positional mistakes in the opening.

Tactical mistakes when I already made some positional mistake are less important.

I had 2122.3 after 48 games when lc0 with white has 2099.6
I expect some small rating increase based on the 6 games.

total score so far is 35-19 for me(35-18 if I do not include one loss on time because of internet disconnection).
Get ready for the upgrade in the next few days, I can tell you It's going to be really tough!
...Marcus91: Thank you very much for your information and collaboration towards all of us, the chess player community; Your news about the increased gaming power of LeelaQueenOdds is a musical melody to my ears. You are mathematicians and scientists, I am just a learner of the interpretation of the notes of my heart, those notes that Our Programmer has placed in the hearts of his children: (the human being). From my heart I understand and have "a hunch" that LeelaQueenOdds will have the level of play in time controls of 3-0 equivalent to 2500 Elo Fide. And that could be soon... I wonder: Will the elite come to dance 3-0 with LeelaQueenOdds, or will they leave her sitting on the bench during the dance and party? ...Marcus91: Do you think that one day computers will have or possess free will? ... Regarding Mr. Uri Blass: I will be happy when he or someone else takes my second place, or moves me to lower positions, in the standings of the best 100 players who face LeelaQueenOdds. This will give me joy.
I am thinking chess is in a coin.Human beings for ever playing in one face.Now I am playing in the other face:"Antichess". Computers are as a fortres where owner forgot to close a little door behind. You must enter across this door.Forget the front.
Father
Posts: 1664
Joined: Sun Mar 19, 2006 4:39 am
Location: Colombia
Full name: Pablo Ignacio Restrepo

Re: Amazing results of LeelaKnightOdds-Dev

Post by Father »

Marcus91 wrote: Tue Nov 12, 2024 12:10 am
Uri Blass wrote: Mon Nov 11, 2024 10:14 pm today not a good day against lc0 and I won only 3.5-2.5

Note that I did not do serious analysis to prepare but my conclusion that if I get a good position lc0 allow easily to trade pieces so serious preperation should be about getting a good position to learn from my positional mistakes in the opening.

Tactical mistakes when I already made some positional mistake are less important.

I had 2122.3 after 48 games when lc0 with white has 2099.6
I expect some small rating increase based on the 6 games.

total score so far is 35-19 for me(35-18 if I do not include one loss on time because of internet disconnection).
Get ready for the upgrade in the next few days, I can tell you It's going to be really tough!
... We must remember: LeelaQueenOdds is the call to be the queen of the Odds. The icon. If all of you manage to raise LeelaQueenOdds to GM level in any blitz time control or higher, this computer will go from being a beautiful cheerleader to being Miss World of chess algorithms. If you manage to raise LeelaQueenOdds' level of play to the GM level in 90 min time controls with increment per move, and beat a GM, LeelaQueenOdds will automatically enter the chess program algorithm with the title of Miss Universe.
I am thinking chess is in a coin.Human beings for ever playing in one face.Now I am playing in the other face:"Antichess". Computers are as a fortres where owner forgot to close a little door behind. You must enter across this door.Forget the front.
lkaufman
Posts: 6108
Joined: Sun Jan 10, 2010 6:15 am
Location: Maryland USA

Re: Amazing results of LeelaKnightOdds-Dev

Post by lkaufman »

Father wrote: Tue Nov 12, 2024 4:06 am
Marcus91 wrote: Tue Nov 12, 2024 12:10 am
Uri Blass wrote: Mon Nov 11, 2024 10:14 pm today not a good day against lc0 and I won only 3.5-2.5

Note that I did not do serious analysis to prepare but my conclusion that if I get a good position lc0 allow easily to trade pieces so serious preperation should be about getting a good position to learn from my positional mistakes in the opening.

Tactical mistakes when I already made some positional mistake are less important.

I had 2122.3 after 48 games when lc0 with white has 2099.6
I expect some small rating increase based on the 6 games.

total score so far is 35-19 for me(35-18 if I do not include one loss on time because of internet disconnection).
Get ready for the upgrade in the next few days, I can tell you It's going to be really tough!
...Marcus91: Thank you very much for your information and collaboration towards all of us, the chess player community; Your news about the increased gaming power of LeelaQueenOdds is a musical melody to my ears. You are mathematicians and scientists, I am just a learner of the interpretation of the notes of my heart, those notes that Our Programmer has placed in the hearts of his children: (the human being). From my heart I understand and have "a hunch" that LeelaQueenOdds will have the level of play in time controls of 3-0 equivalent to 2500 Elo Fide. And that could be soon... I wonder: Will the elite come to dance 3-0 with LeelaQueenOdds, or will they leave her sitting on the bench during the dance and party? ...Marcus91: Do you think that one day computers will have or possess free will? ... Regarding Mr. Uri Blass: I will be happy when he or someone else takes my second place, or moves me to lower positions, in the standings of the best 100 players who face LeelaQueenOdds. This will give me joy.
LeelaQueenOdds has been upgraded to use the net specifically trained for queen odds. Effective 4:30 a.m. UTC Nov 12. Hopefully we'll have a fairly accurate estimate of the elo gain over the current knight-odds trained net within a few days, we believe it will be substantial.
Komodo rules!
Marcus91
Posts: 58
Joined: Sun Jul 10, 2022 11:00 am
Full name: Marco Giorgio

Re: Amazing results of LeelaKnightOdds-Dev

Post by Marcus91 »

Father wrote: Tue Nov 12, 2024 4:06 am
Marcus91 wrote: Tue Nov 12, 2024 12:10 am
Uri Blass wrote: Mon Nov 11, 2024 10:14 pm today not a good day against lc0 and I won only 3.5-2.5

Note that I did not do serious analysis to prepare but my conclusion that if I get a good position lc0 allow easily to trade pieces so serious preperation should be about getting a good position to learn from my positional mistakes in the opening.

Tactical mistakes when I already made some positional mistake are less important.

I had 2122.3 after 48 games when lc0 with white has 2099.6
I expect some small rating increase based on the 6 games.

total score so far is 35-19 for me(35-18 if I do not include one loss on time because of internet disconnection).
Get ready for the upgrade in the next few days, I can tell you It's going to be really tough!
...Marcus91: Thank you very much for your information and collaboration towards all of us, the chess player community; Your news about the increased gaming power of LeelaQueenOdds is a musical melody to my ears. You are mathematicians and scientists, I am just a learner of the interpretation of the notes of my heart, those notes that Our Programmer has placed in the hearts of his children: (the human being). From my heart I understand and have "a hunch" that LeelaQueenOdds will have the level of play in time controls of 3-0 equivalent to 2500 Elo Fide. And that could be soon... I wonder: Will the elite come to dance 3-0 with LeelaQueenOdds, or will they leave her sitting on the bench during the dance and party? ...Marcus91: Do you think that one day computers will have or possess free will? ... Regarding Mr. Uri Blass: I will be happy when he or someone else takes my second place, or moves me to lower positions, in the standings of the best 100 players who face LeelaQueenOdds. This will give me joy.
I agree, the potential of LeelaQueenOdds is exciting, and it’s amazing how fast things are evolving. While 2500 FIDE in 3-0 might be a bit optimistic, it’s not too far off—and just a few months ago, even approaching that level seemed unthinkable. As for the question of free will, it’s a complex philosophical one—right now, we’re focused on making AI more powerful in specific tasks, but the idea of machines with independent will is something for the future. Let’s see where this journey takes us!
Uri Blass
Posts: 10632
Joined: Thu Mar 09, 2006 12:37 am
Location: Tel-Aviv Israel

Re: Amazing results of LeelaKnightOdds-Dev

Post by Uri Blass »

Marcus91 wrote: Tue Nov 12, 2024 10:12 am
Father wrote: Tue Nov 12, 2024 4:06 am
Marcus91 wrote: Tue Nov 12, 2024 12:10 am
Uri Blass wrote: Mon Nov 11, 2024 10:14 pm today not a good day against lc0 and I won only 3.5-2.5

Note that I did not do serious analysis to prepare but my conclusion that if I get a good position lc0 allow easily to trade pieces so serious preperation should be about getting a good position to learn from my positional mistakes in the opening.

Tactical mistakes when I already made some positional mistake are less important.

I had 2122.3 after 48 games when lc0 with white has 2099.6
I expect some small rating increase based on the 6 games.

total score so far is 35-19 for me(35-18 if I do not include one loss on time because of internet disconnection).
Get ready for the upgrade in the next few days, I can tell you It's going to be really tough!
...Marcus91: Thank you very much for your information and collaboration towards all of us, the chess player community; Your news about the increased gaming power of LeelaQueenOdds is a musical melody to my ears. You are mathematicians and scientists, I am just a learner of the interpretation of the notes of my heart, those notes that Our Programmer has placed in the hearts of his children: (the human being). From my heart I understand and have "a hunch" that LeelaQueenOdds will have the level of play in time controls of 3-0 equivalent to 2500 Elo Fide. And that could be soon... I wonder: Will the elite come to dance 3-0 with LeelaQueenOdds, or will they leave her sitting on the bench during the dance and party? ...Marcus91: Do you think that one day computers will have or possess free will? ... Regarding Mr. Uri Blass: I will be happy when he or someone else takes my second place, or moves me to lower positions, in the standings of the best 100 players who face LeelaQueenOdds. This will give me joy.
I agree, the potential of LeelaQueenOdds is exciting, and it’s amazing how fast things are evolving. While 2500 FIDE in 3-0 might be a bit optimistic, it’s not too far off—and just a few months ago, even approaching that level seemed unthinkable. As for the question of free will, it’s a complex philosophical one—right now, we’re focused on making AI more powerful in specific tasks, but the idea of machines with independent will is something for the future. Let’s see where this journey takes us!
Humans do not train seriously for queen odds and my guess is that 2500 against unprepared humans at some time control may be only 2300 against prepared humans.

I do not believe that it is possible to get fide rating of 2300 at 3+0 against prepared humans.
Uri Blass
Posts: 10632
Joined: Thu Mar 09, 2006 12:37 am
Location: Tel-Aviv Israel

Re: Amazing results of LeelaKnightOdds-Dev

Post by Uri Blass »

lkaufman wrote: Fri Nov 08, 2024 9:43 pm
Father wrote: Fri Nov 08, 2024 9:35 pm
Father wrote: Fri Nov 08, 2024 5:34 pm
lkaufman wrote: Fri Nov 08, 2024 5:13 pm
Father wrote: Fri Nov 08, 2024 4:13 pm
Father wrote: Fri Nov 08, 2024 3:01 pm
lkaufman wrote: Wed Nov 06, 2024 6:11 pm
Uri Blass wrote: Wed Nov 06, 2024 5:14 pm I wonder if lc0 calculates based on probability that I do not see the threats of it.


https://lichess.org/u4WvN5Ly

What is the idea behind the blunder 27.Bxc6 bxc6 28.Rxe5+?
Does leela hope that I do not see Rxa5
Maybe it is the best practical chance but I usually do not blunder simple captures like that and I played Qa5 for a reason to defend e5 after trading the knight.

I am leading 16-9(today I won 10-7 that is worse than my score yesterday that was 6-2)
It is a blunder in the sense that massive trading is a blunder when way behind. Most likely the net, being trained on knight odds games only, hasn't learned much about how to play when down by a queen. I expect that a net trained for queen odds will eventually appear in the bot and that should show another large jump in performance. One such net already exists but is not available for use or testing.

So now number 1 on the list is a player with a very high bullet rating, probably a master, number 2 is "father", number 3 is a famous Backgammon grandmaster (and a close personal friend of mine), and number 4 is Uri!
In summary: "We are facing a great challenge regarding the expectations about where the evolutionary journey of the great LeelaQueenOdds and the human opponents, in "Man Against Machine", open tournament aspects, and other alternatives of battlefields.
LeelaQueenOdds versus 🆚 a player superior to 2600 elo, some months ago, when LeelaQueenOdds was wirh minus power than Newdays …

[pgn] Event "Casual bullet game"]
[Site "https://lichess.org/bDYxQU35"]
[Date "2024.06.24"]
[White "LeelaQueenOdds"]
[Black "JievveTheNotable"]
[Result "1-0"]
[UTCDate "2024.06.24"]
[UTCTime "02:18:47"]
[WhiteElo "2000"]
[BlackElo "2616"]
[WhiteTitle "BOT"]
[Variant "From Position"]
[TimeControl "60+1"]
[ECO "?"]
[Opening "?"]
[Termination "Normal"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/RNB1KBNR w KQkq - 0 1"]
[SetUp "1"]
[Annotator "lichess.org"]

1. d4 c6 2. e4 d5 3. a3 dxe4 4. Nc3 Nf6 5. h3 Qxd4 6. Nge2 Qb6 7. Ng3 e5 8. Bc4 Bc5 9. O-O e3 10. Na4 exf2+ 11. Kh2 Qa5 12. Nc3 Qc7 13. Nge4 Nxe4 14. Nxe4 Bb6 15. a4 a5 16. Rd1 O-O 17. Ra3 Bf5 18. Ng3 Bg6 19. Rh1 Nd7 20. Ne2 e4+ 21. Bf4 Ne5 22. Rc3 Qe7 23. Ba2 Rfd8 24. Rg3 Kh8 25. Bg5 f6 26. Bc1 Nf7 27. Nf4 Ng5 28. h4 Nf7 29. h5 Bf5 30. b4 Ng5 31. h6 gxh6 32. bxa5 Bxa5 33. Bb2 e3 34. Rxe3 Qg7 35. Nh5 Qg6 36. Bxf6+ Qxf6 37. Nxf6 Kg7 38. Nh5+ Kg6 39. Nf4+ Kg7 40. Re7+ Kf6 41. Rxb7 Rab8 42. Nh5+ Kg6 43. Rg7+ Kxh5 44. Kg3+ Nh3 45. Rxh3+ Bxh3 46. Bf7# { White wins by checkmate. } 1-0[/pgn]
It has indeed won some bullet games against players rated above 2600, even 2700, while in blitz it has won against at most 2503 rating. I find that it is too easy for me to win against it at 3'2" (that's about the fastest I can play decently), but that probably won't be true once it gets upgraded to a net trained specifically for queen odds, which may happen this month. Meanwhile the knight odds bot has done amazing things, playing 190 games at 1' + 1" against a player rated over 3100 without a single loss (!!), just 26 draws.
...these are two great news for everyone, Mr. Larry Kaufman. Thank you for allowing us to enter the combat of men against machines. In this way, now I will fight to occupy a position among the hundred best players who face LeelaKnightOdds. Of course that does not mean in any way that I am going to abandon the fight against modules in their "standard forms." I wonder: There is already a list, Mr. Larry Kaufman, that informs us about the Top 100 human contenders of LeelaKnight. If this list exists, I will fight to be part of it. !! My current priorities in chess warfare objectives are called: "Oddsmachines."
... "Winds of chess war of man against machine." LeelaKnight against Catecan (Father) in time level of 1+1. The first steps of Father's fight against LeelanigntOdds have begun. Today is November 8, 2024; For now it has only been skirmishes. Today there were 12 combats; 10 wins for the computer and two 2 draws. Consequently, my effectiveness range for the day is 2 draws in 12, which would represent just over 31 draws on a base of 190 games. The silicon horse "LeelaKnightOdds" will surely go to the manger to rest and feed, in the meantime, I will be preparing new battles. Good night I wish you.

[pgn][Event "Casual bullet game"]
[Site "https://lichess.org/jXsGOaQG"]
[Date "2024.11.08"]
[White "LeelaKnightOdds"]
[Black "Catecan"]
[Result "1/2-1/2"]
[UTCDate "2024.11.08"]
[UTCTime "18:17:47"]
[WhiteElo "2000"]
[BlackElo "2030"]
[WhiteTitle "BOT"]
[Variant "From Position"]
[TimeControl "60+1"]
[ECO "?"]
[Opening "?"]
[Termination "Normal"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/RNBQKB1R w KQkq - 0 1"]
[SetUp "1"]
[Annotator "lichess.org"]

1. d4 d5 2. Nc3 e6 3. Bf4 f5 4. Qd2 Bd6 5. g4 Bxf4 6. Qxf4 Qd6 7. Qg5 Qe7 8. Qf4 Qd6 9. Qg5 Qe7 10. Qf4 Qd6 { The game is a draw. } 1/2-1/2[/pgn]

[pgn][Event "Casual bullet game"]
[Site "https://lichess.org/EWJPMx1z"]
[Date "2024.11.08"]
[White "LeelaKnightOdds"]
[Black "Catecan"]
[Result "1/2-1/2"]
[UTCDate "2024.11.08"]
[UTCTime "18:21:33"]
[WhiteElo "2000"]
[BlackElo "2030"]
[WhiteTitle "BOT"]
[Variant "From Position"]
[TimeControl "60+1"]
[ECO "?"]
[Opening "?"]
[Termination "Normal"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/RNBQKB1R w KQkq - 0 1"]
[SetUp "1"]
[Annotator "lichess.org"]

1. e4 d5 2. Nc3 dxe4 3. Nxe4 Nf6 4. Bd3 Nxe4 5. Bxe4 Nd7 6. O-O Nf6 7. Bf3 e6 8. d4 Be7 9. Bf4 O-O 10. Be5 Bd6 11. Qe2 Be7 12. Rad1 c6 13. c4 Bd6 14. Rd3 Qc7 15. Re1 Nd7 16. Bxd6 Qxd6 17. Qd2 Nf6 18. Bd1 Re8 19. c5 Qd8 20. Bc2 Bd7 21. Rg3 g6 22. Re5 Nd5 23. h3 Qf6 24. Rf3 Qg7 25. Bb3 a6 26. Bc4 Re7 27. Rb3 Be8 28. Re4 Rc8 29. a4 Rcc7 30. Rf3 Red7 31. Rb3 Re7 32. Rf3 Red7 33. Rb3 Re7 { The game is a draw. } 1/2-1/2[/pgn]

[pgn][Event "Casual bullet game"]
[Site "https://lichess.org/sktNijn4"]
[Date "2024.11.08"]
[White "LeelaKnightOdds"]
[Black "Catecan"]
[Result "1-0"]
[UTCDate "2024.11.08"]
[UTCTime "17:52:00"]
[WhiteElo "2000"]
[BlackElo "2030"]
[WhiteTitle "BOT"]
[Variant "From Position"]
[TimeControl "60+1"]
[ECO "?"]
[Opening "?"]
[Termination "Time forfeit"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/RNBQKB1R w KQkq - 0 1"]
[SetUp "1"]
[Annotator "lichess.org"]

1. Nc3 d5 2. e4 dxe4 3. Nxe4 Nf6 4. Bd3 Nxe4 5. Bxe4 e6 6. Qe2 Nd7 7. d3 Nf6 8. Bf3 c6 9. Bd2 Be7 10. O-O-O O-O 11. h4 a5 12. g4 Nd5 13. g5 Bd6 14. d4 Bf4 15. c4 Bxd2+ 16. Rxd2 Nf4 17. Qe5 Ng6 18. Qh2 Bd7 19. h5 Ne7 20. Be4 Qb8 21. Qh4 Nf5 22. Qg4 Re8 23. Kb1 Rf8 24. Ka1 Re8 25. Bb1 Rf8 26. Qf3 Re8 27. a3 Rf8 28. Re1 Re8 29. Red1 Rf8 30. d5 cxd5 31. cxd5 Qc7 32. dxe6 Bxe6 33. Bxf5 Bxf5 34. Qxf5 Rad8 35. Rd5 Rxd5 36. Qxd5 { White wins on time. } 1-0[/pgn]

[pgn][Event "Casual bullet game"]
[Site "https://lichess.org/Fi0EZMuo"]
[Date "2024.11.08"]
[White "LeelaKnightOdds"]
[Black "Catecan"]
[Result "1-0"]
[UTCDate "2024.11.08"]
[UTCTime "17:37:41"]
[WhiteElo "2000"]
[BlackElo "2030"]
[WhiteTitle "BOT"]
[Variant "From Position"]
[TimeControl "60+1"]
[ECO "?"]
[Opening "?"]
[Termination "Normal"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/RNBQKB1R w KQkq - 0 1"]
[SetUp "1"]
[Annotator "lichess.org"]

1. e4 d5 2. exd5 Qxd5 3. Nc3 Qd8 4. d4 e6 5. Be3 Nf6 6. a3 c6 7. Bd3 Be7 8. h4 Nbd7 9. Qe2 Qa5 10. h5 h6 11. Bd2 Qc7 12. Qf3 Qd8 13. Qg3 Rg8 14. O-O-O Qb6 15. Rh4 Qd8 16. Re1 Nb6 17. Nd1 Qd6 18. Bf4 Qd8 19. Bc7 Qd7 20. c4 Bd8 21. Be5 Be7 22. Nc3 Qd8 23. Kb1 Qd7 24. Ka1 Qd8 25. Rh3 Qd7 26. Qf4 Qd8 27. Rf3 Qd7 28. g4 Qd8 29. Ne4 Nbd7 30. Bc7 { Black resigns. } 1-0[/pgn]
I wouldn't expect a Leaderboard for LeelaKnightOdds, because almost no one ever wins a game (occasionally either an IM or GM or a cheater wins), so it would just be a contest to see who could induce more draws by repetition, which isn't very interesting. A future version might try much harder to avoid such reps.
I wonder what about leaderboard for Rookodds.

A fide master with fide rating 2345 believe he will almost always win against queen odds at time control of 10+0 regardless of the program when rook odds may be more interesting.
Uri Blass
Posts: 10632
Joined: Thu Mar 09, 2006 12:37 am
Location: Tel-Aviv Israel

Re: Amazing results of LeelaKnightOdds-Dev

Post by Uri Blass »

somebody asked me where to download the version without a knight for analysis of chess positions.

I guess it can help humans to know the move that gives them the best practical chance in a losing position.
Uri Blass
Posts: 10632
Joined: Thu Mar 09, 2006 12:37 am
Location: Tel-Aviv Israel

Re: Amazing results of LeelaKnightOdds-Dev

Post by Uri Blass »

For some reason nobody played against Leela in the last 9 hours(I tried to play and after some time of waiting gave up of playing and I guess the problem is that lc0 is not available for some reason.
lkaufman
Posts: 6108
Joined: Sun Jan 10, 2010 6:15 am
Location: Maryland USA

Re: Amazing results of LeelaKnightOdds-Dev

Post by lkaufman »

Uri Blass wrote: Tue Nov 12, 2024 2:43 pm For some reason nobody played against Leela in the last 9 hours(I tried to play and after some time of waiting gave up of playing and I guess the problem is that lc0 is not available for some reason.
There was an internet outage last night, before a single game was played following the update. It has been fixed and the bots are all back online, with the update for the queen odds bot. It has been given a rating 200 above the previous one, tentatively.
Komodo rules!
lkaufman
Posts: 6108
Joined: Sun Jan 10, 2010 6:15 am
Location: Maryland USA

Re: Amazing results of LeelaKnightOdds-Dev

Post by lkaufman »

Uri Blass wrote: Tue Nov 12, 2024 2:10 pm somebody asked me where to download the version without a knight for analysis of chess positions.

I guess it can help humans to know the move that gives them the best practical chance in a losing position.
It is not currently available to the public. It is likely that using normal Leela with Contempt is a better way to judge practical chances in positions that are losing but just barely so, like a pawn down. The knight odds net should be better when the deficit approaches a piece.
Komodo rules!