Amazing results of LeelaKnightOdds-Dev

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

Moderator: Ras

Father
Posts: 1800
Joined: Sun Mar 19, 2006 4:39 am
Location: Colombia
Full name: Pablo Ignacio Restrepo

Re: Amazing results of LeelaKnightOdds-Dev

Post by Father »

lkaufman wrote: Mon Oct 07, 2024 12:43 am Since getting upgraded just prior to Sept. 28, LeelaKnightOdds-Dev (trained specifically for knight odds play) has the following results on LiChess, counting only games against 2400+ rated humans and only with a minimum of 3 min base AND 2 sec increment. 164 wins, five draws, ZERO losses!! Most of the opponents in the 2500s, which is typically somewhere between FM and IM level. This works out to a performance rating approaching 3300, way above the blitz rating of Magnus Carlsen on LiChess!! It's risky to extrapolate from this that the bot would actually win a 3'2" match giving Magnus Carlsen knight odds, but that would be my bet. Moreover, Leela did this despite being internally set to play 1'1" bullet speed or even faster, and despite the single-GPU computer sometimes playing other odds games simultaneously! Of course in Rapid it MIGHT only take a 2700 FIDE level opponent to win a match, perhaps we'll find out soon.
If anyone knows of any players near or above 2700 FIDE who might be interested in a match for moderate prize money, or if anyone has contact info for such a player who might be receptive, please send me a PM. We have one interested so far, but we would like to have two or three opponents this year. If and when we show that we can win a Rapid match from a 2700 ballpark player or two, we can start to approach the top ten players in the world. The goal is to be able to win a Rapid knight odds match with any human in the world by next year sometime. If we can't get Magnus Carlsen to play we'll have to show a performance rating against top-10 players that exceeds his rating to claim success.
I recall the discussions here about a decade ago, with the general conclusion that the above would be forever impossible, although I recall that Bob Hyatt disagreed with this conclusion. Hopefully we'll find out before too long.
...I will now be silent...After I perform the psychoanalysis on the "LeelaknightOdds-Dev" algorithm, and taking my last 10 games as a starting point, my estimated elo would be 1-10/10*700 =x.
3300-x=y; If the trend continues, my performance in 169 games would be = x1, breaking all standards. The depth of our souls and the mysteries of the Creator are insoluble. Thank you again Mr. Larry Kaufman. I would like the help of someone in the forum for valuable collaboration in placing the game with Fritz playing the human without his queen based on a sacrifice against Odds, by the human being. That departure will help me regain hope, despite having the unwavering faith, that the Programmer of our spirits guides the history of humanity. "We call evil the good that we do not understand at the moment." For my part, I will always be grateful for existing and for my state of consciousness..." I think therefore I am."
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: 1800
Joined: Sun Mar 19, 2006 4:39 am
Location: Colombia
Full name: Pablo Ignacio Restrepo

Re: Amazing results of LeelaKnightOdds-Dev

Post by Father »

Father wrote: Mon Oct 14, 2024 3:54 pm
lkaufman wrote: Mon Oct 07, 2024 12:43 am Since getting upgraded just prior to Sept. 28, LeelaKnightOdds-Dev (trained specifically for knight odds play) has the following results on LiChess, counting only games against 2400+ rated humans and only with a minimum of 3 min base AND 2 sec increment. 164 wins, five draws, ZERO losses!! Most of the opponents in the 2500s, which is typically somewhere between FM and IM level. This works out to a performance rating approaching 3300, way above the blitz rating of Magnus Carlsen on LiChess!! It's risky to extrapolate from this that the bot would actually win a 3'2" match giving Magnus Carlsen knight odds, but that would be my bet. Moreover, Leela did this despite being internally set to play 1'1" bullet speed or even faster, and despite the single-GPU computer sometimes playing other odds games simultaneously! Of course in Rapid it MIGHT only take a 2700 FIDE level opponent to win a match, perhaps we'll find out soon.
If anyone knows of any players near or above 2700 FIDE who might be interested in a match for moderate prize money, or if anyone has contact info for such a player who might be receptive, please send me a PM. We have one interested so far, but we would like to have two or three opponents this year. If and when we show that we can win a Rapid match from a 2700 ballpark player or two, we can start to approach the top ten players in the world. The goal is to be able to win a Rapid knight odds match with any human in the world by next year sometime. If we can't get Magnus Carlsen to play we'll have to show a performance rating against top-10 players that exceeds his rating to claim success.
I recall the discussions here about a decade ago, with the general conclusion that the above would be forever impossible, although I recall that Bob Hyatt disagreed with this conclusion. Hopefully we'll find out before too long.
...I will now be silent...After I perform the psychoanalysis on the "LeelaknightOdds-Dev" algorithm, and taking my last 10 games as a starting point, my estimated elo would be 1-10/10*700 =x.
3300-x=y; If the trend continues, my performance in 169 games would be = x1, breaking all standards. The depth of our souls and the mysteries of the Creator are insoluble. Thank you again Mr. Larry Kaufman. I would like the help of someone in the forum for valuable collaboration in placing the game with Fritz playing the human without his queen based on a sacrifice against Odds, by the human being. That departure will help me regain hope, despite having the unwavering faith, that the Programmer of our spirits guides the history of humanity. "We call evil the good that we do not understand at the moment." For my part, I will always be grateful for existing and for my state of consciousness..." I think therefore I am."
**1-9/10…
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: 6217
Joined: Sun Jan 10, 2010 6:15 am
Location: Maryland USA
Full name: Larry Kaufman

Re: Amazing results of LeelaKnightOdds-Dev

Post by lkaufman »

Father wrote: Mon Oct 14, 2024 6:14 am
lkaufman wrote: Mon Oct 07, 2024 12:43 am Since getting upgraded just prior to Sept. 28, LeelaKnightOdds-Dev (trained specifically for knight odds play) has the following results on LiChess, counting only games against 2400+ rated humans and only with a minimum of 3 min base AND 2 sec increment. 164 wins, five draws, ZERO losses!! Most of the opponents in the 2500s, which is typically somewhere between FM and IM level. This works out to a performance rating approaching 3300, way above the blitz rating of Magnus Carlsen on LiChess!! It's risky to extrapolate from this that the bot would actually win a 3'2" match giving Magnus Carlsen knight odds, but that would be my bet. Moreover, Leela did this despite being internally set to play 1'1" bullet speed or even faster, and despite the single-GPU computer sometimes playing other odds games simultaneously! Of course in Rapid it MIGHT only take a 2700 FIDE level opponent to win a match, perhaps we'll find out soon.
If anyone knows of any players near or above 2700 FIDE who might be interested in a match for moderate prize money, or if anyone has contact info for such a player who might be receptive, please send me a PM. We have one interested so far, but we would like to have two or three opponents this year. If and when we show that we can win a Rapid match from a 2700 ballpark player or two, we can start to approach the top ten players in the world. The goal is to be able to win a Rapid knight odds match with any human in the world by next year sometime. If we can't get Magnus Carlsen to play we'll have to show a performance rating against top-10 players that exceeds his rating to claim success.
I recall the discussions here about a decade ago, with the general conclusion that the above would be forever impossible, although I recall that Bob Hyatt disagreed with this conclusion. Hopefully we'll find out before too long.
Mr. Larry Kaufman. I thank you again for teaching me about odds. I want to tell you that I have begun to understand LeelaKnightOdds-Dev. As you can deduce from reading my last ten pieces of the dance, I have changed the strategic step, and the computer could soon be in trouble, to the point that if the probabilistic tendency, the blind spot, occurs, it could lead to a superior performance. to the estimated statistician to rate my playing strength. I will continue from my silence, building the network and the cage to contain, from the philosophy of a simple mortal and ordinary human being, the beast of AI. Thank you again Mr. Larry Kaufman. For me, the greatest reward received is that you have invested part of your time in discussing my concerns about the game of science and computers.

[pgn] [Event "Casual blitz game"]
[Site "https://lichess.org/L04V0eFa"]
[Date "2024.10.14"]
[White "LeelaKnightOdds-dev"]
[Black "Catecan"]
[Result "1/2-1/2"]
[UTCDate "2024.10.14"]
[UTCTime "03:16:26"]
[WhiteElo "2000"]
[BlackElo "2082"]
[WhiteTitle "BOT"]
[Variant "From Position"]
[TimeControl "180+0"]
[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. Bf4 e6 3. Nc3 c6 4. e4 dxe4 5. Nxe4 Nf6 6. Nxf6+ Qxf6 7. Be5 Qd8 8. Qd3 Nd7 9. Bg3 Nf6 10. O-O-O Bd6 11. Be2 Bxg3 12. hxg3 Qe7 13. Qc3 Bd7 14. Qa5 Qd8 15. Qe5 Qb8 16. Qg5 Rg8 17. Rd3 Qd8 18. Qe5 Qb8 19. Qe3 Nd5 20. Qd2 Nf6 21. Ra3 Qd8 22. g4 h6 23. f4 Qc7 24. g5 hxg5 25. fxg5 Nd5 26. Rf1 a6 27. Bc4 O-O-O 28. Qe2 Rgf8 29. Rb3 b5 30. Bxd5 exd5 31. Qd2 g6 32. a4 Kb7 33. Re3 Be6 34. Rc3 Qb6 35. b3 Rd7 36. Kb2 Rfd8 37. Rc5 Ka7 38. Ra1 Kb7 39. Qc3 Rc7 40. Qg3 Rcd7 41. Qc3 Rc7 42. Qg3 Rcd7 43. Qc3 { The game is a draw. } 1/2-1/2[/pgn]
I'm glad you are enjoying playing these bots. But I just want to emphasize that the bot doesn't know you are rated way below its "target" rating in the GM range, so it doesn't especially try to avoid draws with you. Even if it knew your rating, it was trained only against GM level bots which aren't aiming for draws so it isn't especially strong at avoiding draws. Calculating performance ratings based on what percentage of the games you are able to draw may be misleading for these reasons. I think there will be some new developments with the bots in the near future.
Komodo rules!
Uri Blass
Posts: 10786
Joined: Thu Mar 09, 2006 12:37 am
Location: Tel-Aviv Israel

Re: Amazing results of LeelaKnightOdds-Dev

Post by Uri Blass »

I will try to play against lealaKnightOdds-Dev and I will see how many games I need to play for a win or draw(I do not plan to do special preperation except learning from previous games).

I do not plan to play more than one game per day(I prefer to learn more from previous games before I start a new game(I do not know if people who played many games against lc0 did it in this way or simply played one game after the other that means that they had no time to analyze the previous games to learn from mistakes).

Here is my first game that I lost(I trust Leela to be able to win a won endgame so I resigned).

My tactical chanhing result mistakes are 13...Qe7 (Qe8 or Be8 were winning) and 14...Rxf7(14...Be6 was the only move to keep the position equal)

2...dxe4 was probably bad based on stockfish that prefered 2...d4 but I wanted to trade material(not sure if 2...dxe4 was a bad practical decision)

I am also not sure if 4...Nxe4 was not a practical mistake(stockfish prefers 4...Nc6).
I am not sure if 8...exf4 was not a practical mistake(stockfish prefers 8...f5 and 9...e4.
It seems 11...Bd7 was a practical mistake and better is stockfish's move 11...Qd6(Bd7 miss the fork 12.Qb3 against f7 and b7 that for some reason is not stockfish's suggestion)



[pgn][Event "Casual blitz game"]
[Site "https://lichess.org/3kyOyG6W"]
[Date "2024.10.14"]
[White "LeelaKnightOdds-dev"]
[Black "uriblass"]
[Result "1-0"]
[UTCDate "2024.10.14"]
[UTCTime "19:33:50"]
[WhiteElo "2000"]
[BlackElo "2132"]
[WhiteTitle "BOT"]
[Variant "From Position"]
[TimeControl "300+3"]
[ECO "?"]
[Opening "?"]
[Termination "Normal"]
[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 e5 6. O-O Nc6 7. c3 Be7 8. f4 exf4 9. d4 Bd6 10. Bxf4 Bxf4 11. Rxf4 Bd7 12. Qb3 O-O 13. Raf1 Qe7 14. Rxf7 Rxf7 15. Rxf7 Be6 16. Rxe7 Bxb3 17. Bxc6 bxc6 18. axb3 Rb8 19. b4 a5 20. bxa5 Rxb2 21. Rxc7 Ra2 22. Rxc6 { Black resigns. } 1-0
[/pgn]
Father
Posts: 1800
Joined: Sun Mar 19, 2006 4:39 am
Location: Colombia
Full name: Pablo Ignacio Restrepo

Re: Amazing results of LeelaKnightOdds-Dev

Post by Father »

Uri Blass wrote: Mon Oct 14, 2024 10:01 pm I will try to play against lealaKnightOdds-Dev and I will see how many games I need to play for a win or draw(I do not plan to do special preperation except learning from previous games).

I do not plan to play more than one game per day(I prefer to learn more from previous games before I start a new game(I do not know if people who played many games against lc0 did it in this way or simply played one game after the other that means that they had no time to analyze the previous games to learn from mistakes).

Here is my first game that I lost(I trust Leela to be able to win a won endgame so I resigned).

My tactical chanhing result mistakes are 13...Qe7 (Qe8 or Be8 were winning) and 14...Rxf7(14...Be6 was the only move to keep the position equal)

2...dxe4 was probably bad based on stockfish that prefered 2...d4 but I wanted to trade material(not sure if 2...dxe4 was a bad practical decision)

I am also not sure if 4...Nxe4 was not a practical mistake(stockfish prefers 4...Nc6).
I am not sure if 8...exf4 was not a practical mistake(stockfish prefers 8...f5 and 9...e4.
It seems 11...Bd7 was a practical mistake and better is stockfish's move 11...Qd6(Bd7 miss the fork 12.Qb3 against f7 and b7 that for some reason is not stockfish's suggestion)



[pgn][Event "Casual blitz game"]
[Site "https://lichess.org/3kyOyG6W"]
[Date "2024.10.14"]
[White "LeelaKnightOdds-dev"]
[Black "uriblass"]
[Result "1-0"]
[UTCDate "2024.10.14"]
[UTCTime "19:33:50"]
[WhiteElo "2000"]
[BlackElo "2132"]
[WhiteTitle "BOT"]
[Variant "From Position"]
[TimeControl "300+3"]
[ECO "?"]
[Opening "?"]
[Termination "Normal"]
[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 e5 6. O-O Nc6 7. c3 Be7 8. f4 exf4 9. d4 Bd6 10. Bxf4 Bxf4 11. Rxf4 Bd7 12. Qb3 O-O 13. Raf1 Qe7 14. Rxf7 Rxf7 15. Rxf7 Be6 16. Rxe7 Bxb3 17. Bxc6 bxc6 18. axb3 Rb8 19. b4 a5 20. bxa5 Rxb2 21. Rxc7 Ra2 22. Rxc6 { Black resigns. } 1-0
[/pgn]
Good morning Mr. Uri Blass. I wish the best for you, for your city and for your people, happiness and love.
To your question, whether it is a better path or a worse path in the training system of man against chess machines, having a greater number of combats, my answer, given the experience of 44 years of battles, is without the slightest doubt, opt for the greatest number of battles. It is also interesting and useful in training to play matches of 10 games alternating colors. In this way and added product, you will outline strategic concepts that will give you an option to lead the algorithm to the blind spot. Your common human state of consciousness and your ability to anticipate philosophical thinking will defeat the mathematical essence of hardware and software.

Thank you very mutch for for your game against the computer. I like it a lot.
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: 1800
Joined: Sun Mar 19, 2006 4:39 am
Location: Colombia
Full name: Pablo Ignacio Restrepo

Re: Amazing results of LeelaKnightOdds-Dev

Post by Father »

lkaufman wrote: Mon Oct 14, 2024 8:40 pm
Father wrote: Mon Oct 14, 2024 6:14 am
lkaufman wrote: Mon Oct 07, 2024 12:43 am Since getting upgraded just prior to Sept. 28, LeelaKnightOdds-Dev (trained specifically for knight odds play) has the following results on LiChess, counting only games against 2400+ rated humans and only with a minimum of 3 min base AND 2 sec increment. 164 wins, five draws, ZERO losses!! Most of the opponents in the 2500s, which is typically somewhere between FM and IM level. This works out to a performance rating approaching 3300, way above the blitz rating of Magnus Carlsen on LiChess!! It's risky to extrapolate from this that the bot would actually win a 3'2" match giving Magnus Carlsen knight odds, but that would be my bet. Moreover, Leela did this despite being internally set to play 1'1" bullet speed or even faster, and despite the single-GPU computer sometimes playing other odds games simultaneously! Of course in Rapid it MIGHT only take a 2700 FIDE level opponent to win a match, perhaps we'll find out soon.
If anyone knows of any players near or above 2700 FIDE who might be interested in a match for moderate prize money, or if anyone has contact info for such a player who might be receptive, please send me a PM. We have one interested so far, but we would like to have two or three opponents this year. If and when we show that we can win a Rapid match from a 2700 ballpark player or two, we can start to approach the top ten players in the world. The goal is to be able to win a Rapid knight odds match with any human in the world by next year sometime. If we can't get Magnus Carlsen to play we'll have to show a performance rating against top-10 players that exceeds his rating to claim success.
I recall the discussions here about a decade ago, with the general conclusion that the above would be forever impossible, although I recall that Bob Hyatt disagreed with this conclusion. Hopefully we'll find out before too long.
Mr. Larry Kaufman. I thank you again for teaching me about odds. I want to tell you that I have begun to understand LeelaKnightOdds-Dev. As you can deduce from reading my last ten pieces of the dance, I have changed the strategic step, and the computer could soon be in trouble, to the point that if the probabilistic tendency, the blind spot, occurs, it could lead to a superior performance. to the estimated statistician to rate my playing strength. I will continue from my silence, building the network and the cage to contain, from the philosophy of a simple mortal and ordinary human being, the beast of AI. Thank you again Mr. Larry Kaufman. For me, the greatest reward received is that you have invested part of your time in discussing my concerns about the game of science and computers.

[pgn] [Event "Casual blitz game"]
[Site "https://lichess.org/L04V0eFa"]
[Date "2024.10.14"]
[White "LeelaKnightOdds-dev"]
[Black "Catecan"]
[Result "1/2-1/2"]
[UTCDate "2024.10.14"]
[UTCTime "03:16:26"]
[WhiteElo "2000"]
[BlackElo "2082"]
[WhiteTitle "BOT"]
[Variant "From Position"]
[TimeControl "180+0"]
[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. Bf4 e6 3. Nc3 c6 4. e4 dxe4 5. Nxe4 Nf6 6. Nxf6+ Qxf6 7. Be5 Qd8 8. Qd3 Nd7 9. Bg3 Nf6 10. O-O-O Bd6 11. Be2 Bxg3 12. hxg3 Qe7 13. Qc3 Bd7 14. Qa5 Qd8 15. Qe5 Qb8 16. Qg5 Rg8 17. Rd3 Qd8 18. Qe5 Qb8 19. Qe3 Nd5 20. Qd2 Nf6 21. Ra3 Qd8 22. g4 h6 23. f4 Qc7 24. g5 hxg5 25. fxg5 Nd5 26. Rf1 a6 27. Bc4 O-O-O 28. Qe2 Rgf8 29. Rb3 b5 30. Bxd5 exd5 31. Qd2 g6 32. a4 Kb7 33. Re3 Be6 34. Rc3 Qb6 35. b3 Rd7 36. Kb2 Rfd8 37. Rc5 Ka7 38. Ra1 Kb7 39. Qc3 Rc7 40. Qg3 Rcd7 41. Qc3 Rc7 42. Qg3 Rcd7 43. Qc3 { The game is a draw. } 1/2-1/2[/pgn]
I'm glad you are enjoying playing these bots. But I just want to emphasize that the bot doesn't know you are rated way below its "target" rating in the GM range, so it doesn't especially try to avoid draws with you. Even if it knew your rating, it was trained only against GM level bots which aren't aiming for draws so it isn't especially strong at avoiding draws. Calculating performance ratings based on what percentage of the games you are able to draw may be misleading for these reasons. I think there will be some new developments with the bots in the near future.
Good morning Mr. Larry Kaufman. Thank you very much for the development of odds and other software. You make our lives more joyful and bearable...referring to the movie, "The War of the Worlds" I dreamed and imagined that I was part of that movie... due to analogical parallels, I said to myself: "The aliens are the invading machines; the GMs and best human chess players are and represented by the weapons of the great nations that face the invading extraterrestrial machines, and I gave Father the role of the "microbe that managed to enter." to the CPU of the invaders. Really being or pretending to be an insignificant microbe makes me happy... especially when you aspire to be stronger in results than the GMs achieve and you aspire to defeat the infernal invading extraterrestrial machines.
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: 1800
Joined: Sun Mar 19, 2006 4:39 am
Location: Colombia
Full name: Pablo Ignacio Restrepo

Re: Amazing results of LeelaKnightOdds-Dev

Post by Father »

Father wrote: Tue Oct 15, 2024 4:56 pm
lkaufman wrote: Mon Oct 14, 2024 8:40 pm
Father wrote: Mon Oct 14, 2024 6:14 am
lkaufman wrote: Mon Oct 07, 2024 12:43 am Since getting upgraded just prior to Sept. 28, LeelaKnightOdds-Dev (trained specifically for knight odds play) has the following results on LiChess, counting only games against 2400+ rated humans and only with a minimum of 3 min base AND 2 sec increment. 164 wins, five draws, ZERO losses!! Most of the opponents in the 2500s, which is typically somewhere between FM and IM level. This works out to a performance rating approaching 3300, way above the blitz rating of Magnus Carlsen on LiChess!! It's risky to extrapolate from this that the bot would actually win a 3'2" match giving Magnus Carlsen knight odds, but that would be my bet. Moreover, Leela did this despite being internally set to play 1'1" bullet speed or even faster, and despite the single-GPU computer sometimes playing other odds games simultaneously! Of course in Rapid it MIGHT only take a 2700 FIDE level opponent to win a match, perhaps we'll find out soon.
If anyone knows of any players near or above 2700 FIDE who might be interested in a match for moderate prize money, or if anyone has contact info for such a player who might be receptive, please send me a PM. We have one interested so far, but we would like to have two or three opponents this year. If and when we show that we can win a Rapid match from a 2700 ballpark player or two, we can start to approach the top ten players in the world. The goal is to be able to win a Rapid knight odds match with any human in the world by next year sometime. If we can't get Magnus Carlsen to play we'll have to show a performance rating against top-10 players that exceeds his rating to claim success.
I recall the discussions here about a decade ago, with the general conclusion that the above would be forever impossible, although I recall that Bob Hyatt disagreed with this conclusion. Hopefully we'll find out before too long.
Mr. Larry Kaufman. I thank you again for teaching me about odds. I want to tell you that I have begun to understand LeelaKnightOdds-Dev. As you can deduce from reading my last ten pieces of the dance, I have changed the strategic step, and the computer could soon be in trouble, to the point that if the probabilistic tendency, the blind spot, occurs, it could lead to a superior performance. to the estimated statistician to rate my playing strength. I will continue from my silence, building the network and the cage to contain, from the philosophy of a simple mortal and ordinary human being, the beast of AI. Thank you again Mr. Larry Kaufman. For me, the greatest reward received is that you have invested part of your time in discussing my concerns about the game of science and computers.

[pgn] [Event "Casual blitz game"]
[Site "https://lichess.org/L04V0eFa"]
[Date "2024.10.14"]
[White "LeelaKnightOdds-dev"]
[Black "Catecan"]
[Result "1/2-1/2"]
[UTCDate "2024.10.14"]
[UTCTime "03:16:26"]
[WhiteElo "2000"]
[BlackElo "2082"]
[WhiteTitle "BOT"]
[Variant "From Position"]
[TimeControl "180+0"]
[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. Bf4 e6 3. Nc3 c6 4. e4 dxe4 5. Nxe4 Nf6 6. Nxf6+ Qxf6 7. Be5 Qd8 8. Qd3 Nd7 9. Bg3 Nf6 10. O-O-O Bd6 11. Be2 Bxg3 12. hxg3 Qe7 13. Qc3 Bd7 14. Qa5 Qd8 15. Qe5 Qb8 16. Qg5 Rg8 17. Rd3 Qd8 18. Qe5 Qb8 19. Qe3 Nd5 20. Qd2 Nf6 21. Ra3 Qd8 22. g4 h6 23. f4 Qc7 24. g5 hxg5 25. fxg5 Nd5 26. Rf1 a6 27. Bc4 O-O-O 28. Qe2 Rgf8 29. Rb3 b5 30. Bxd5 exd5 31. Qd2 g6 32. a4 Kb7 33. Re3 Be6 34. Rc3 Qb6 35. b3 Rd7 36. Kb2 Rfd8 37. Rc5 Ka7 38. Ra1 Kb7 39. Qc3 Rc7 40. Qg3 Rcd7 41. Qc3 Rc7 42. Qg3 Rcd7 43. Qc3 { The game is a draw. } 1/2-1/2[/pgn]
I'm glad you are enjoying playing these bots. But I just want to emphasize that the bot doesn't know you are rated way below its "target" rating in the GM range, so it doesn't especially try to avoid draws with you. Even if it knew your rating, it was trained only against GM level bots which aren't aiming for draws so it isn't especially strong at avoiding draws. Calculating performance ratings based on what percentage of the games you are able to draw may be misleading for these reasons. I think there will be some new developments with the bots in the near future.
Good morning Mr. Larry Kaufman. Thank you very much for the development of odds and other software. You make our lives more joyful and bearable...referring to the movie, "The War of the Worlds" I dreamed and imagined that I was part of that movie... due to analogical parallels, I said to myself: "The aliens are the invading machines; the GMs and best human chess players are and represented by the weapons of the great nations that face the invading extraterrestrial machines, and I gave Father the role of the "microbe that managed to enter." to the CPU of the invaders. Really being or pretending to be an insignificant microbe makes me happy... especially when you aspire to be stronger in results than the GMs achieve and you aspire to defeat the infernal invading extraterrestrial machines.
** Errata. Mr. Larry Kaufman. I must make a clarification, when I write, "when you aspire to" I mean "when I aspire to" I simply wrote in the third person. I don't want to give rise to misunderstandings or disrespect.
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.
Uri Blass
Posts: 10786
Joined: Thu Mar 09, 2006 12:37 am
Location: Tel-Aviv Israel

Re: Amazing results of LeelaKnightOdds-Dev

Post by Uri Blass »

lost second 5+3 game against lc0
[pgn]
Event "Casual blitz game"]
[Site "https://lichess.org/tvV8YqoU"]
[Date "2024.10.15"]
[White "LeelaKnightOdds-dev"]
[Black "uriblass"]
[Result "1-0"]
[UTCDate "2024.10.15"]
[UTCTime "18:48:39"]
[WhiteElo "2000"]
[BlackElo "2132"]
[WhiteTitle "BOT"]
[Variant "From Position"]
[TimeControl "300+3"]
[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 Nf6 5. Bg5 h6 6. Bh4 e6 7. Qd2 Nc6 8. O-O-O Be7 9. Bc4 Ne4 10. Nxe4 Bxh4 11. c3 O-O 12. Qe2 Qe7 13. g3 Bf6 14. h4 b6 15. g4 Bxh4 16. f4 f5 17. gxf5 Rxf5 18. Qg4 Bf6 19. Qxf5 { Black resigns. } 1-0
5+3 • Casual • FROM POSITION
[/pgn]
Uri Blass
Posts: 10786
Joined: Thu Mar 09, 2006 12:37 am
Location: Tel-Aviv Israel

Re: Amazing results of LeelaKnightOdds-Dev

Post by Uri Blass »

2 tactical changing result mistakes.

17...Kh8 was objectively winning and 18...Na5 probably the only drawing move.

I was at time trouble and did not see the tactics of the fork of lc0 against me.

Earlier in the game it seems that 2...Qxd5 it not best and better is 2...c6 or 2...Nf6
3...Qd8 was bad and better was 3...Qd4 4.Nb5 Qb6
5...e6 was slightly better than 5...h6
6...Ne4 force a trade but stockfish prefers 6...Nd5 or 6...Nb4 or 6...a6 but I think 6...Nd5 is better in human terms.
11...0-0 was bad and better was 11...e5 but the line that I played could transpose
12...Qe7 was a positional mistake and better was the pawn sacrifice 12...e5
14...Rd8 or 14...e5 were better than 14...b6
15...Bxh4 16.f4 f5 was the best line
Uri Blass
Posts: 10786
Joined: Thu Mar 09, 2006 12:37 am
Location: Tel-Aviv Israel

Re: Amazing results of LeelaKnightOdds-Dev

Post by Uri Blass »

game 3:

[pgn][Event "Casual blitz game"]
[Site "https://lichess.org/Sj346w8S"]
[Date "2024.10.15"]
[White "LeelaKnightOdds-dev"]
[Black "uriblass"]
[Result "1-0"]
[UTCDate "2024.10.15"]
[UTCTime "21:53:27"]
[WhiteElo "2000"]
[BlackElo "2132"]
[WhiteTitle "BOT"]
[Variant "From Position"]
[TimeControl "300+3"]
[ECO "?"]
[Opening "?"]
[Termination "Normal"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/RNBQKB1R w KQkq - 0 1"]
[SetUp "1"]
[Annotator "lichess.org"]

1. c4 d5 2. cxd5 Nf6 3. e4 e6 4. dxe6 Bxe6 5. d3 Nc6 6. Be2 Bb4+ 7. Nc3 Qd7 8. O-O O-O 9. Qe1 Nd4 10. Bd1 Rad8 11. Bg5 Be7 12. f4 g6 13. Qh4 Kg7 14. f5 Nxf5 15. exf5 Bxf5 16. Bh6+ Kg8 17. Bxf8 Kxf8 18. Kh1 Qxd3 19. Bb3 Nh5 20. Qe1 Qd2 21. g4 Qxe1 22. Raxe1 Bxg4 23. Rxf7+ Ke8 24. Rexe7# { White wins by checkmate. } 1-0
5+3 • Casual • FROM POSITION

[/pgn]

losing mistake was 20...Qd2 20...Nf6 was the only move to stay in the game
Earlier 12...g6 was a serious mistake. 12...Ne8 was better(I saw the threat f5 and did not see that Ne8 defends against the threat.
The problem with g6 is that it allows Qh4 that wins material and I have nothing better than Kg7 or Nxe4