Amazing results of LeelaKnightOdds-Dev

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

Moderators: hgm, chrisw, Rebel

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: Sat Nov 23, 2024 7:53 am
Uri Blass wrote: Sat Nov 23, 2024 2:21 am The main problem with the queen odd bots is that there are players who do not take it seriously.
One player lost most of his games but he told me by a private message that the reason is that he used a computer of a friend that make often mouse slips.

I asked him why he insisted to play with a bad mouse and he replied that it is only for fun and he is tired.
Why is that a big problem? Do you mean that it is a problem if we rely on all the games played for determining the formula for the leaderboard? I think that the formula allows somewhat for the fact that not all players play seriously. As long as we can make a reasonably good guess as to how much this affects results, we can compensate for it. Other than this, just ignore results that are clearly below average, there are always going to be players who play when sleepy or drunk or who just don't care, but I think they are a minority. There are other players like you who go to the opposite extreme and try to memorize their games in the hope of getting repeated openings which they have learned from. So maybe overall it's a fair test of players who try their best in each game without actually trying to repeat previous ones.
Practically the number of positions in the book that I memorized is less than 60 moves so relatively it was not a big effort.
I did not try to memorize complete games that I played but only until the first move Wasp suggested something else.

In normal chess I memorize significantly more than 60 moves.
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 »

I can add that father played a lot of 1+0 games against Leela in the last 24 hours.

It seems that he drew 15 games with white out of 50 and 10 games with black out of 193.

Not sure about the exact number of games and it is probably 10 games with black out of 192 because for some reason he resigned at move 2 in the only 5+3 game he played against the bot in the last 24 hours(maybe because games when the player resign in the second move are not for rating).

https://lichess.org/XCu2R0KS/black

search in lichess seems to have bugs but I guess it is 10 out of 192 games not including this game.
cbash
Posts: 3
Joined: Fri Nov 22, 2024 7:12 pm
Full name: Caleb Bash

Re: Amazing results of LeelaKnightOdds-Dev

Post by cbash »

Marcus91 wrote: Thu Nov 21, 2024 11:27 am
Marcus91 wrote: Thu Nov 21, 2024 10:40 am
Uri Blass wrote: Thu Nov 21, 2024 9:02 am
lkaufman wrote: Thu Nov 21, 2024 8:06 am
Uri Blass wrote: Thu Nov 21, 2024 7:51 am Something is clearly wrong with the bot and it seems not to use the opponent time.

https://lichess.org/Zz9FRZ3J#163

What is wrong with 82...e5 when the opponent has only 2.4 seconds on the clock in 1+0 game instead of allowing a draw with the 50 move rule?
The excuse of not knowing who is the opponent is not convincing.
Even if you assume that you play against a strong GM I believe 82...e5 is winning when the strong GM has only 2.4 seconds to finish the game.
You are correct that the bot does not take into consideration how much time the opponent has. In principle that should help, but it is not simple to use this information and other things have higher priority now. Simply reducing draws in general is much more important, and has already been done for the queenforknight bot. The goal has been to win matches against "fair" opponents, not to run up the score by avoiding draws against weaker than fair opponents, but now clearly most opponents of all bots are much below "fair".
I think that maybe you need some leaderboard tables:
one for bullet time control,one for blitz time control and one for rapid time control.
Well, I like to find a relationship between time control and game strength, however drawing in quick tc is much easier than win in slower tc, players like "Catecan" or "Brunetticus" seem to do just that, so maybe I should add an arbitrary rule, like halve 'K' for all draws?
All draws have 'K' halved now, play for the win!
Other changes: K=40 for the first 30 games, K = 20 up to 150 games, K=10 for subsequent games
Hello, I would like to say that I have been enjoying the odds bots, and I think the queenoddsbot leader board was a good idea. Acknowledging that we should view the leader board as in an open beta subject to change. Nonetheless it seems you inadvertently caused some sour feelings with this change. Whether or not it was a good change or not I don't actually think was the issue. I think it was that unlike the previous rating updates this one changed the relative rankings of the players. I have a simple compromise that will hopefully alleviate this. Add a Grandfather clause to the new k factor so they only apply to games played after the rating system changed. Given enough time and active participation it will all equalize to how you want it without robbing players of there 'earned' rating and let us return to focusing on having fun and exploring the strong play from these bots.
Father
Posts: 1660
Joined: Sun Mar 19, 2006 4:39 am
Location: Colombia
Full name: Pablo Ignacio Restrepo

Re: Amazing results of LeelaKnightOdds-Dev

Post by Father »

cbash wrote: Sat Nov 23, 2024 2:56 pm
Marcus91 wrote: Thu Nov 21, 2024 11:27 am
Marcus91 wrote: Thu Nov 21, 2024 10:40 am
Uri Blass wrote: Thu Nov 21, 2024 9:02 am
lkaufman wrote: Thu Nov 21, 2024 8:06 am
Uri Blass wrote: Thu Nov 21, 2024 7:51 am Something is clearly wrong with the bot and it seems not to use the opponent time.

https://lichess.org/Zz9FRZ3J#163

What is wrong with 82...e5 when the opponent has only 2.4 seconds on the clock in 1+0 game instead of allowing a draw with the 50 move rule?
The excuse of not knowing who is the opponent is not convincing.
Even if you assume that you play against a strong GM I believe 82...e5 is winning when the strong GM has only 2.4 seconds to finish the game.
You are correct that the bot does not take into consideration how much time the opponent has. In principle that should help, but it is not simple to use this information and other things have higher priority now. Simply reducing draws in general is much more important, and has already been done for the queenforknight bot. The goal has been to win matches against "fair" opponents, not to run up the score by avoiding draws against weaker than fair opponents, but now clearly most opponents of all bots are much below "fair".
I think that maybe you need some leaderboard tables:
one for bullet time control,one for blitz time control and one for rapid time control.
Well, I like to find a relationship between time control and game strength, however drawing in quick tc is much easier than win in slower tc, players like "Catecan" or "Brunetticus" seem to do just that, so maybe I should add an arbitrary rule, like halve 'K' for all draws?
All draws have 'K' halved now, play for the win!
Other changes: K=40 for the first 30 games, K = 20 up to 150 games, K=10 for subsequent games
Hello, I would like to say that I have been enjoying the odds bots, and I think the queenoddsbot leader board was a good idea. Acknowledging that we should view the leader board as in an open beta subject to change. Nonetheless it seems you inadvertently caused some sour feelings with this change. Whether or not it was a good change or not I don't actually think was the issue. I think it was that unlike the previous rating updates this one changed the relative rankings of the players. I have a simple compromise that will hopefully alleviate this. Add a Grandfather clause to the new k factor so they only apply to games played after the rating system changed. Given enough time and active participation it will all equalize to how you want it without robbing players of there 'earned' rating and let us return to focusing on having fun and exploring the strong play from these bots.

cbash thank you very much for the message.
I had decided not to play anymore or write more in this post. It was thanks to a friend, that I am here again. He told me and advised me to continue playing and having fun, that those titles were not what was truly important. Also, last night I felt a little sad, because I thought that I might have committed myself, causing mortification to Uri and Mr. Larry Kaufman, and that it would be better to give them the joy of knowing that I am still playing and learning from the bots and of all of you. Thanks again everyone. To tell the truth, I know that there is great work behind all this technological development and artificial intelligence, and it is precisely to that chain that has participated throughout history and created the possibility of playing against machines, who deserve the first prize. Thank you.
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: 10632
Joined: Thu Mar 09, 2006 12:37 am
Location: Tel-Aviv Israel

Re: Amazing results of LeelaKnightOdds-Dev

Post by Uri Blass »

I also decided to go back to play and I play against the weakest opponent to win.
I won 2.5-0.5 against LeelaQueenOdds as black 15+10 (Reference) 2027.3

First game was a bad game when I decided to accept a draw by repetition in some not simple winning position because I was afraid to make a mistake and lose in time trouble.

Second and Third game were relatively better.

This is the leaderboard that does not include my 3 games.
I believe that it may be easier for me to get rating points by winning at slow time control(and accept a draw only if I already did not play well) and not by trying to draw at fast time control.

LeelaQueenOdds as white 1+0 (Reference) 2895.2
LeelaQueenOdds as black 1+0 (Reference) 2795.2
LeelaQueenOdds as white 1+1 (Reference) 2634.5
LeelaQueenOdds as white 3+0 (Reference) 2619.5
LeelaQueenOdds as black 1+1 (Reference) 2534.5
LeelaQueenOdds as black 3+0 (Reference) 2519.5
1 Master_Unknown 2463.3 38
LeelaQueenOdds as white 3+2 (Reference) 2450.0
2 Orichess88 2442.7 33
3 Ushkothebear 2441.9 144
4 Omnivor 2406.8 30
LeelaQueenOdds as white 5+3 (Reference) 2353.7
LeelaQueenOdds as black 3+2 (Reference) 2350.0
5 Respectful_Dave 2256.9 68
LeelaQueenOdds as black 5+3 (Reference) 2253.7
6 Catecan 2223.7 703
LeelaQueenOdds as white 10+5 (Reference) 2220.3
7 Ajktulm 2210.6 199
8 iWannaSeeYouSmile 2209.8 73
9 AxelBoldt 2208.4 175
10 uriblass 2193.8 136
11 lyxsgdzt 2179.1 49
12 Hissha 2177.6 64
13 wael96 2171.0 53
14 TheLlamaLord 2153.7 22
15 nn1133 2152.3 6167
16 Forgive1219 2144.2 90
17 mapascual 2140.3 42
18 Jagen 2134.5 56
LeelaQueenOdds as white 15+10 (Reference) 2127.3
LeelaQueenOdds as black 10+5 (Reference) 2120.3
19 edgebottle 2084.2 56
20 Brunetticus 2070.2 434
21 Punypawn315 2064.9 19
22 wireless78 2064.4 21
23 xHadyx 2057.1 61
24 vital303 2044.8 96
25 RhinoTank 2035.2 36
26 Superhond 2032.4 30
27 box_boxed 2029.9 28
LeelaQueenOdds as black 15+10 (Reference) 2027.3
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: Sat Nov 23, 2024 3:22 pm
cbash wrote: Sat Nov 23, 2024 2:56 pm
Marcus91 wrote: Thu Nov 21, 2024 11:27 am
Marcus91 wrote: Thu Nov 21, 2024 10:40 am
Uri Blass wrote: Thu Nov 21, 2024 9:02 am
lkaufman wrote: Thu Nov 21, 2024 8:06 am
Uri Blass wrote: Thu Nov 21, 2024 7:51 am Something is clearly wrong with the bot and it seems not to use the opponent time.

https://lichess.org/Zz9FRZ3J#163

What is wrong with 82...e5 when the opponent has only 2.4 seconds on the clock in 1+0 game instead of allowing a draw with the 50 move rule?
The excuse of not knowing who is the opponent is not convincing.
Even if you assume that you play against a strong GM I believe 82...e5 is winning when the strong GM has only 2.4 seconds to finish the game.
You are correct that the bot does not take into consideration how much time the opponent has. In principle that should help, but it is not simple to use this information and other things have higher priority now. Simply reducing draws in general is much more important, and has already been done for the queenforknight bot. The goal has been to win matches against "fair" opponents, not to run up the score by avoiding draws against weaker than fair opponents, but now clearly most opponents of all bots are much below "fair".
I think that maybe you need some leaderboard tables:
one for bullet time control,one for blitz time control and one for rapid time control.
Well, I like to find a relationship between time control and game strength, however drawing in quick tc is much easier than win in slower tc, players like "Catecan" or "Brunetticus" seem to do just that, so maybe I should add an arbitrary rule, like halve 'K' for all draws?
All draws have 'K' halved now, play for the win!
Other changes: K=40 for the first 30 games, K = 20 up to 150 games, K=10 for subsequent games
Hello, I would like to say that I have been enjoying the odds bots, and I think the queenoddsbot leader board was a good idea. Acknowledging that we should view the leader board as in an open beta subject to change. Nonetheless it seems you inadvertently caused some sour feelings with this change. Whether or not it was a good change or not I don't actually think was the issue. I think it was that unlike the previous rating updates this one changed the relative rankings of the players. I have a simple compromise that will hopefully alleviate this. Add a Grandfather clause to the new k factor so they only apply to games played after the rating system changed. Given enough time and active participation it will all equalize to how you want it without robbing players of there 'earned' rating and let us return to focusing on having fun and exploring the strong play from these bots.

cbash thank you very much for the message.
I had decided not to play anymore or write more in this post. It was thanks to a friend, that I am here again. He told me and advised me to continue playing and having fun, that those titles were not what was truly important. Also, last night I felt a little sad, because I thought that I might have committed myself, causing mortification to Uri and Mr. Larry Kaufman, and that it would be better to give them the joy of knowing that I am still playing and learning from the bots and of all of you. Thanks again everyone. To tell the truth, I know that there is great work behind all this technological development and artificial intelligence, and it is precisely to that chain that has participated throughout history and created the possibility of playing against machines, who deserve the first prize. Thank you.
Thank you for pointing that out. I’m sorry if halving 'K' for the draws came across as an unfair decision. Over the past few days, many factors on this leaderboard have been adjusted multiple times. Personally, I find the data from your games against the bot very insightful for understanding how time control impacts overall strength and the strategies adopted by his players as well.
I’m not in a position to decide what’s fair or not—we should focus on simply enjoying the challenge of playing the bot. No offense is intended by what the leaderboard reflect. That said, I want to give special recognition to 'Father' climbing through the rankings despite everithing, managing to draw against such a strong bot in 1+0. Truly remarkable!
Father
Posts: 1660
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: Sun Nov 24, 2024 1:55 am
Father wrote: Sat Nov 23, 2024 3:22 pm
cbash wrote: Sat Nov 23, 2024 2:56 pm
Marcus91 wrote: Thu Nov 21, 2024 11:27 am
Marcus91 wrote: Thu Nov 21, 2024 10:40 am
Uri Blass wrote: Thu Nov 21, 2024 9:02 am
lkaufman wrote: Thu Nov 21, 2024 8:06 am
Uri Blass wrote: Thu Nov 21, 2024 7:51 am Something is clearly wrong with the bot and it seems not to use the opponent time.

https://lichess.org/Zz9FRZ3J#163

What is wrong with 82...e5 when the opponent has only 2.4 seconds on the clock in 1+0 game instead of allowing a draw with the 50 move rule?
The excuse of not knowing who is the opponent is not convincing.
Even if you assume that you play against a strong GM I believe 82...e5 is winning when the strong GM has only 2.4 seconds to finish the game.
You are correct that the bot does not take into consideration how much time the opponent has. In principle that should help, but it is not simple to use this information and other things have higher priority now. Simply reducing draws in general is much more important, and has already been done for the queenforknight bot. The goal has been to win matches against "fair" opponents, not to run up the score by avoiding draws against weaker than fair opponents, but now clearly most opponents of all bots are much below "fair".
I think that maybe you need some leaderboard tables:
one for bullet time control,one for blitz time control and one for rapid time control.
Well, I like to find a relationship between time control and game strength, however drawing in quick tc is much easier than win in slower tc, players like "Catecan" or "Brunetticus" seem to do just that, so maybe I should add an arbitrary rule, like halve 'K' for all draws?
All draws have 'K' halved now, play for the win!
Other changes: K=40 for the first 30 games, K = 20 up to 150 games, K=10 for subsequent games
Hello, I would like to say that I have been enjoying the odds bots, and I think the queenoddsbot leader board was a good idea. Acknowledging that we should view the leader board as in an open beta subject to change. Nonetheless it seems you inadvertently caused some sour feelings with this change. Whether or not it was a good change or not I don't actually think was the issue. I think it was that unlike the previous rating updates this one changed the relative rankings of the players. I have a simple compromise that will hopefully alleviate this. Add a Grandfather clause to the new k factor so they only apply to games played after the rating system changed. Given enough time and active participation it will all equalize to how you want it without robbing players of there 'earned' rating and let us return to focusing on having fun and exploring the strong play from these bots.

cbash thank you very much for the message.
I had decided not to play anymore or write more in this post. It was thanks to a friend, that I am here again. He told me and advised me to continue playing and having fun, that those titles were not what was truly important. Also, last night I felt a little sad, because I thought that I might have committed myself, causing mortification to Uri and Mr. Larry Kaufman, and that it would be better to give them the joy of knowing that I am still playing and learning from the bots and of all of you. Thanks again everyone. To tell the truth, I know that there is great work behind all this technological development and artificial intelligence, and it is precisely to that chain that has participated throughout history and created the possibility of playing against machines, who deserve the first prize. Thank you.
Thank you for pointing that out. I’m sorry if halving 'K' for the draws came across as an unfair decision. Over the past few days, many factors on this leaderboard have been adjusted multiple times. Personally, I find the data from your games against the bot very insightful for understanding how time control impacts overall strength and the strategies adopted by his players as well.
I’m not in a position to decide what’s fair or not—we should focus on simply enjoying the challenge of playing the bot. No offense is intended by what the leaderboard reflect. That said, I want to give special recognition to 'Father' climbing through the rankings despite everithing, managing to draw against such a strong bot in 1+0. Truly remarkable!
Marcus91, thanks for the message. I would believe that the important thing is to build bridges instead of dividing with walls. If you watch the games by the minute, you will find many mistakes on my part, given the relentless force of the Bot. But still, you will be able to observe a true mental dialogue between the Bot and my mind, for example, along the way I have been able to understand the tendency and brand of the Bot's learning, which creates paths on the part of the cybernetic mind and my mind. It's new for me. This is something incredible, and in games with long time controls there would not be enough time to do it, since there would be very few games and it would require a lot of investment of energy and time. For me it is easier to face a computer in classic controls, or with increased time per move, but it bores me, I consider that I learn more from the repeated bullet samples of "error hit error construction". Later, if you want, I could play with more extensive time controls, then the conclusive bullet guidelines would be implemented there. I find the strength and power of LeelaQueenOdds fantastic and at times it frustrates me. It is easier for me to face a computer with all the chips on the computer's side than to face LeelaQueenOdds in quick games. Machines for me were previously very predictable, for example in the first decade of the 21st century, I had the pleasure of playing repeatedly against a multiprocessor computer that ran at 200,000 kilondops per second at 60 minutes plus 15 seconds per play. That one, an anaconda, LeelaQueenOdds, a swarm of wasps. I will fight for a position on all tables if there are any and against standard computers. The machines overwhelm us, but not yet completely.
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: 1660
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: Sun Nov 24, 2024 1:55 am
Father wrote: Sat Nov 23, 2024 3:22 pm
cbash wrote: Sat Nov 23, 2024 2:56 pm
Marcus91 wrote: Thu Nov 21, 2024 11:27 am
Marcus91 wrote: Thu Nov 21, 2024 10:40 am
Uri Blass wrote: Thu Nov 21, 2024 9:02 am
lkaufman wrote: Thu Nov 21, 2024 8:06 am
Uri Blass wrote: Thu Nov 21, 2024 7:51 am Something is clearly wrong with the bot and it seems not to use the opponent time.

https://lichess.org/Zz9FRZ3J#163

What is wrong with 82...e5 when the opponent has only 2.4 seconds on the clock in 1+0 game instead of allowing a draw with the 50 move rule?
The excuse of not knowing who is the opponent is not convincing.
Even if you assume that you play against a strong GM I believe 82...e5 is winning when the strong GM has only 2.4 seconds to finish the game.
You are correct that the bot does not take into consideration how much time the opponent has. In principle that should help, but it is not simple to use this information and other things have higher priority now. Simply reducing draws in general is much more important, and has already been done for the queenforknight bot. The goal has been to win matches against "fair" opponents, not to run up the score by avoiding draws against weaker than fair opponents, but now clearly most opponents of all bots are much below "fair".
I think that maybe you need some leaderboard tables:
one for bullet time control,one for blitz time control and one for rapid time control.
Well, I like to find a relationship between time control and game strength, however drawing in quick tc is much easier than win in slower tc, players like "Catecan" or "Brunetticus" seem to do just that, so maybe I should add an arbitrary rule, like halve 'K' for all draws?
All draws have 'K' halved now, play for the win!
Other changes: K=40 for the first 30 games, K = 20 up to 150 games, K=10 for subsequent games
Hello, I would like to say that I have been enjoying the odds bots, and I think the queenoddsbot leader board was a good idea. Acknowledging that we should view the leader board as in an open beta subject to change. Nonetheless it seems you inadvertently caused some sour feelings with this change. Whether or not it was a good change or not I don't actually think was the issue. I think it was that unlike the previous rating updates this one changed the relative rankings of the players. I have a simple compromise that will hopefully alleviate this. Add a Grandfather clause to the new k factor so they only apply to games played after the rating system changed. Given enough time and active participation it will all equalize to how you want it without robbing players of there 'earned' rating and let us return to focusing on having fun and exploring the strong play from these bots.

cbash thank you very much for the message.
I had decided not to play anymore or write more in this post. It was thanks to a friend, that I am here again. He told me and advised me to continue playing and having fun, that those titles were not what was truly important. Also, last night I felt a little sad, because I thought that I might have committed myself, causing mortification to Uri and Mr. Larry Kaufman, and that it would be better to give them the joy of knowing that I am still playing and learning from the bots and of all of you. Thanks again everyone. To tell the truth, I know that there is great work behind all this technological development and artificial intelligence, and it is precisely to that chain that has participated throughout history and created the possibility of playing against machines, who deserve the first prize. Thank you.
Thank you for pointing that out. I’m sorry if halving 'K' for the draws came across as an unfair decision. Over the past few days, many factors on this leaderboard have been adjusted multiple times. Personally, I find the data from your games against the bot very insightful for understanding how time control impacts overall strength and the strategies adopted by his players as well.
I’m not in a position to decide what’s fair or not—we should focus on simply enjoying the challenge of playing the bot. No offense is intended by what the leaderboard reflect. That said, I want to give special recognition to 'Father' climbing through the rankings despite everithing, managing to draw against such a strong bot in 1+0. Truly remarkable!
Odds bots represent an extraordinary catalyst in the fight between the human mind and the computer. They allow forces to be adjusted so that the paradigms of chess purely between humans can be addressed or not with those of cybernetic thinking. As interesting as chess itself is the depth of autonomous and hybrid thinking. Machines come to humiliate and crush us to the point that it seems that they tyrannize us, but this requires unsuspected forces of will to emerge, a true overcoming. Man against machine, more than a stupid and meaningless fight, it is really true poetry.
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: 1660
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: Sun Nov 24, 2024 1:55 am
Father wrote: Sat Nov 23, 2024 3:22 pm
cbash wrote: Sat Nov 23, 2024 2:56 pm
Marcus91 wrote: Thu Nov 21, 2024 11:27 am
Marcus91 wrote: Thu Nov 21, 2024 10:40 am
Uri Blass wrote: Thu Nov 21, 2024 9:02 am
lkaufman wrote: Thu Nov 21, 2024 8:06 am
Uri Blass wrote: Thu Nov 21, 2024 7:51 am Something is clearly wrong with the bot and it seems not to use the opponent time.

https://lichess.org/Zz9FRZ3J#163

What is wrong with 82...e5 when the opponent has only 2.4 seconds on the clock in 1+0 game instead of allowing a draw with the 50 move rule?
The excuse of not knowing who is the opponent is not convincing.
Even if you assume that you play against a strong GM I believe 82...e5 is winning when the strong GM has only 2.4 seconds to finish the game.
You are correct that the bot does not take into consideration how much time the opponent has. In principle that should help, but it is not simple to use this information and other things have higher priority now. Simply reducing draws in general is much more important, and has already been done for the queenforknight bot. The goal has been to win matches against "fair" opponents, not to run up the score by avoiding draws against weaker than fair opponents, but now clearly most opponents of all bots are much below "fair".
I think that maybe you need some leaderboard tables:
one for bullet time control,one for blitz time control and one for rapid time control.
Well, I like to find a relationship between time control and game strength, however drawing in quick tc is much easier than win in slower tc, players like "Catecan" or "Brunetticus" seem to do just that, so maybe I should add an arbitrary rule, like halve 'K' for all draws?
All draws have 'K' halved now, play for the win!
Other changes: K=40 for the first 30 games, K = 20 up to 150 games, K=10 for subsequent games
Hello, I would like to say that I have been enjoying the odds bots, and I think the queenoddsbot leader board was a good idea. Acknowledging that we should view the leader board as in an open beta subject to change. Nonetheless it seems you inadvertently caused some sour feelings with this change. Whether or not it was a good change or not I don't actually think was the issue. I think it was that unlike the previous rating updates this one changed the relative rankings of the players. I have a simple compromise that will hopefully alleviate this. Add a Grandfather clause to the new k factor so they only apply to games played after the rating system changed. Given enough time and active participation it will all equalize to how you want it without robbing players of there 'earned' rating and let us return to focusing on having fun and exploring the strong play from these bots.

cbash thank you very much for the message.
I had decided not to play anymore or write more in this post. It was thanks to a friend, that I am here again. He told me and advised me to continue playing and having fun, that those titles were not what was truly important. Also, last night I felt a little sad, because I thought that I might have committed myself, causing mortification to Uri and Mr. Larry Kaufman, and that it would be better to give them the joy of knowing that I am still playing and learning from the bots and of all of you. Thanks again everyone. To tell the truth, I know that there is great work behind all this technological development and artificial intelligence, and it is precisely to that chain that has participated throughout history and created the possibility of playing against machines, who deserve the first prize. Thank you.
Thank you for pointing that out. I’m sorry if halving 'K' for the draws came across as an unfair decision. Over the past few days, many factors on this leaderboard have been adjusted multiple times. Personally, I find the data from your games against the bot very insightful for understanding how time control impacts overall strength and the strategies adopted by his players as well.
I’m not in a position to decide what’s fair or not—we should focus on simply enjoying the challenge of playing the bot. No offense is intended by what the leaderboard reflect. That said, I want to give special recognition to 'Father' climbing through the rankings despite everithing, managing to draw against such a strong bot in 1+0. Truly remarkable!
I feel a bit exhausted, I've been trying to play with LeelaQueenOdds but it seems the computer doesn't want anything to do with me. He refuses to play. Only accept some encounters. To all, I wish you a happy night and day and above all, the best.
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: Sun Nov 24, 2024 3:06 am
I feel a bit exhausted, I've been trying to play with LeelaQueenOdds but it seems the computer doesn't want anything to do with me. He refuses to play. Only accept some encounters. To all, I wish you a happy night and day and above all, the best.
All the games of all the LeelaOdds bots are played on a single computer with just one GPU. In order to prevent it from crashing, there is currently a limit of four games at once for queen odds, three for knight odds, two for rook odds, and two for queen for knight. If a bot refuses to play with you, it normally means that already the limit (4 games at once for queen odds) has been reached. Normally you won't have to wait long until one of the four games ends. So not only is the bot giving queen odds to strong players and moving in about a tenth of a second per move, it is sometimes playing ten games at once at various odds, all on one GPU, and scoring around 80% at queen odds, much more at lesser odds!

This is perhaps the most amazing thing I have seen in my long life, not only in the chess world but in general. It is about as incredible as finding intelligent space aliens among us, or bringing dinosaurs back to life, or almost anything from science fiction. I think it gives us a peak at the future, when bots will seem like gods to us, all-knowing and all-powerful. Frightening, but at least we chess players get more warning than most people.
Komodo rules!