| | CCC FRC semifinal: all games draw so far.. | |
| | Author | Message |
---|
mwyoung
Posts : 880 Join date : 2020-11-25 Location : USA
| Subject: CCC FRC semifinal: all games draw so far.. Tue Oct 19, 2021 9:22 pm | |
| - Code:
-
Re: CCC FRC semifinal: all games draw so far.. Post by lkaufman » Tue Oct 12, 2021 5:41 pm
Well, all draws among the top 3, but overall in the semifinals, with six of the best engines, Komodo Dragon has 18 wins to 0 losses, Stockfish 17 wins to 0 losses, and Lc0 16 wins to 0 losses as of this writing. So FRC does seem to allow for plenty of wins when there is some strength disparity, even at 3500+ levels. But I suspect that with long time limits, the top engines on this giant hardware would indeed draw 99+% of the games. Standard chess with no specified openings is even more drawish. The best engines are far from perfect, but they probably play well enough at long time limits on big hardware that they should never practically never lose a game unless forced to play an inferior opening. The draw margin in chess is just much larger than the errors made at this level. We should really find a better way to address this problem than just having each organizer pick bad openings arbitrarily. We need a solution that provides variety, decisive results, and makes some kind of sense logically. It is curious that people can see the truth. But will not take the next logical step. 1. Yes chess engines are far from perfect. 2. And yes the top engines have a high draw rate when playing each other. And not just at long time controls, or on powerful hardware. We are always stuck in today with our logic. What do I mean. 1. Chess engines from just 1 year ago also had a very high draw rate. And this is true for 2 years ago, and 3 years ago and so on. But if you play a 1 year old chess engine, or a 2 year old chess engine, or a 3 year old chess engine. With today's best chess engines. All the older chess engines draw rate falls when playing the best of todays chess engines. As they lose more games, to the better chess engines of today. And this is not really hard to understand. Chess is a 2 player game, and chess engines are far from perfect. But it takes a better chess engine to refute the mistakes. As seen when the older chess engines are playing the best chess engines today. And this will also be the result of todays best chess engines. As better chess engines are developed. | |
| | | Uri Blass
Posts : 193 Join date : 2020-11-28
| Subject: Re: CCC FRC semifinal: all games draw so far.. Wed Oct 20, 2021 3:22 pm | |
| - mwyoung wrote:
-
- Code:
-
Re: CCC FRC semifinal: all games draw so far.. Post by lkaufman » Tue Oct 12, 2021 5:41 pm
Well, all draws among the top 3, but overall in the semifinals, with six of the best engines, Komodo Dragon has 18 wins to 0 losses, Stockfish 17 wins to 0 losses, and Lc0 16 wins to 0 losses as of this writing. So FRC does seem to allow for plenty of wins when there is some strength disparity, even at 3500+ levels. But I suspect that with long time limits, the top engines on this giant hardware would indeed draw 99+% of the games. Standard chess with no specified openings is even more drawish. The best engines are far from perfect, but they probably play well enough at long time limits on big hardware that they should never practically never lose a game unless forced to play an inferior opening. The draw margin in chess is just much larger than the errors made at this level. We should really find a better way to address this problem than just having each organizer pick bad openings arbitrarily. We need a solution that provides variety, decisive results, and makes some kind of sense logically.
It is curious that people can see the truth. But will not take the next logical step.
1. Yes chess engines are far from perfect. 2. And yes the top engines have a high draw rate when playing each other. And not just at long time controls, or on powerful hardware.
We are always stuck in today with our logic. What do I mean.
1. Chess engines from just 1 year ago also had a very high draw rate. And this is true for 2 years ago, and 3 years ago and so on.
But if you play a 1 year old chess engine, or a 2 year old chess engine, or a 3 year old chess engine. With today's best chess engines. All the older chess engines draw rate falls when playing the best of todays chess engines. As they lose more games, to the better chess engines of today.
And this is not really hard to understand. Chess is a 2 player game, and chess engines are far from perfect. But it takes a better chess engine to refute the mistakes. As seen when the older chess engines are playing the best chess engines today.
And this will also be the result of todays best chess engines. As better chess engines are developed. The situation today is different because we have higher draw rate then the draw rate that we had 3 years ago and chess engines clearly have problems to beat top engines that are one years old I remember the match when stockfish14 won against stockfish13 only 3-0 and 997 draws. | |
| | | mwyoung
Posts : 880 Join date : 2020-11-25 Location : USA
| Subject: Re: CCC FRC semifinal: all games draw so far.. Wed Oct 20, 2021 9:27 pm | |
| - Uri Blass wrote:
- mwyoung wrote:
-
- Code:
-
Re: CCC FRC semifinal: all games draw so far.. Post by lkaufman » Tue Oct 12, 2021 5:41 pm
Well, all draws among the top 3, but overall in the semifinals, with six of the best engines, Komodo Dragon has 18 wins to 0 losses, Stockfish 17 wins to 0 losses, and Lc0 16 wins to 0 losses as of this writing. So FRC does seem to allow for plenty of wins when there is some strength disparity, even at 3500+ levels. But I suspect that with long time limits, the top engines on this giant hardware would indeed draw 99+% of the games. Standard chess with no specified openings is even more drawish. The best engines are far from perfect, but they probably play well enough at long time limits on big hardware that they should never practically never lose a game unless forced to play an inferior opening. The draw margin in chess is just much larger than the errors made at this level. We should really find a better way to address this problem than just having each organizer pick bad openings arbitrarily. We need a solution that provides variety, decisive results, and makes some kind of sense logically.
It is curious that people can see the truth. But will not take the next logical step.
1. Yes chess engines are far from perfect. 2. And yes the top engines have a high draw rate when playing each other. And not just at long time controls, or on powerful hardware.
We are always stuck in today with our logic. What do I mean.
1. Chess engines from just 1 year ago also had a very high draw rate. And this is true for 2 years ago, and 3 years ago and so on.
But if you play a 1 year old chess engine, or a 2 year old chess engine, or a 3 year old chess engine. With today's best chess engines. All the older chess engines draw rate falls when playing the best of todays chess engines. As they lose more games, to the better chess engines of today.
And this is not really hard to understand. Chess is a 2 player game, and chess engines are far from perfect. But it takes a better chess engine to refute the mistakes. As seen when the older chess engines are playing the best chess engines today.
And this will also be the result of todays best chess engines. As better chess engines are developed. The situation today is different because we have higher draw rate then the draw rate that we had 3 years ago and chess engines clearly have problems to beat top engines that are one years old
I remember the match when stockfish14 won against stockfish13 only 3-0 and 997 draws. "I remember the match when stockfish14 won against stockfish13 only 3-0 and 997 draws." Yes but the engines are not that much different in age. And Fishtest clearly has a issue in testing. But this is not surprising as if you only test at bullet for improvements. That is what you are most likely to get. Or sometimes worse when testing at longer time controls. Here is the latest run, SF 14 vs Stockfish 19/10/21. Score of Stockfish 19/10/21 vs Stockfish 14: 0 - 2 - 144 [0.493] ... Stockfish 19/10/21 playing White: 0 - 0 - 73 [0.500] 73 ... Stockfish 19/10/21 playing Black: 0 - 2 - 71 [0.486] 73 ... White vs Black: 2 - 0 - 144 [0.507] 146 Elo difference: -4.8 +/- 6.5, LOS: 7.9 %, DrawRatio: 98.6 % 146 of 500 games finished. But here is what your logic means. You are claiming the draw rate of today's engines will remain the same. In 1 year, 3 years and forever. When playing newer engines. You are basically claiming Stockfish has softly solved chess. Do you want to bet money on your conclusions. | |
| | | Uri Blass
Posts : 193 Join date : 2020-11-28
| Subject: Re: CCC FRC semifinal: all games draw so far.. Wed Oct 20, 2021 10:06 pm | |
| - mwyoung wrote:
- Uri Blass wrote:
- mwyoung wrote:
-
- Code:
-
Re: CCC FRC semifinal: all games draw so far.. Post by lkaufman » Tue Oct 12, 2021 5:41 pm
Well, all draws among the top 3, but overall in the semifinals, with six of the best engines, Komodo Dragon has 18 wins to 0 losses, Stockfish 17 wins to 0 losses, and Lc0 16 wins to 0 losses as of this writing. So FRC does seem to allow for plenty of wins when there is some strength disparity, even at 3500+ levels. But I suspect that with long time limits, the top engines on this giant hardware would indeed draw 99+% of the games. Standard chess with no specified openings is even more drawish. The best engines are far from perfect, but they probably play well enough at long time limits on big hardware that they should never practically never lose a game unless forced to play an inferior opening. The draw margin in chess is just much larger than the errors made at this level. We should really find a better way to address this problem than just having each organizer pick bad openings arbitrarily. We need a solution that provides variety, decisive results, and makes some kind of sense logically.
It is curious that people can see the truth. But will not take the next logical step.
1. Yes chess engines are far from perfect. 2. And yes the top engines have a high draw rate when playing each other. And not just at long time controls, or on powerful hardware.
We are always stuck in today with our logic. What do I mean.
1. Chess engines from just 1 year ago also had a very high draw rate. And this is true for 2 years ago, and 3 years ago and so on.
But if you play a 1 year old chess engine, or a 2 year old chess engine, or a 3 year old chess engine. With today's best chess engines. All the older chess engines draw rate falls when playing the best of todays chess engines. As they lose more games, to the better chess engines of today.
And this is not really hard to understand. Chess is a 2 player game, and chess engines are far from perfect. But it takes a better chess engine to refute the mistakes. As seen when the older chess engines are playing the best chess engines today.
And this will also be the result of todays best chess engines. As better chess engines are developed. The situation today is different because we have higher draw rate then the draw rate that we had 3 years ago and chess engines clearly have problems to beat top engines that are one years old
I remember the match when stockfish14 won against stockfish13 only 3-0 and 997 draws. "I remember the match when stockfish14 won against stockfish13 only 3-0 and 997 draws."
Yes but the engines are not that much different in age. And Fishtest clearly has a issue in testing. But this is not surprising as if you only test at bullet for improvements. That is what you are most likely to get. Or sometimes worse when testing at longer time controls.
Here is the latest run, SF 14 vs Stockfish 19/10/21. Score of Stockfish 19/10/21 vs Stockfish 14: 0 - 2 - 144 [0.493] ... Stockfish 19/10/21 playing White: 0 - 0 - 73 [0.500] 73 ... Stockfish 19/10/21 playing Black: 0 - 2 - 71 [0.486] 73 ... White vs Black: 2 - 0 - 144 [0.507] 146 Elo difference: -4.8 +/- 6.5, LOS: 7.9 %, DrawRatio: 98.6 % 146 of 500 games finished.
But here is what your logic means. You are claiming the draw rate of today's engines will remain the same. In 1 year, 3 years and forever. When playing newer engines. You are basically claiming Stockfish has softly solved chess.
Do you want to bet money on your conclusions. Nם I do not claim that forever engines of today are not going to lose but only that the situation today is different I do not believe that engines are perfect but I believe that even with unequal time control they usually do not make mistakes in games between them simply because they do not play in the right way to help the opponent to make mistakes so practically you will get more than 90% draws even if you play unequal time control like 5 minutes per game+5 seconds per move against 50 minutes per game+50 minutes per move. In other words more time help the engine to avoid losing mistakes but does not help the engine to play moves that cause the opponent to make losing mistakes. | |
| | | mwyoung
Posts : 880 Join date : 2020-11-25 Location : USA
| Subject: Re: CCC FRC semifinal: all games draw so far.. Wed Oct 20, 2021 10:13 pm | |
| - Uri Blass wrote:
- mwyoung wrote:
- Uri Blass wrote:
- mwyoung wrote:
-
- Code:
-
Re: CCC FRC semifinal: all games draw so far.. Post by lkaufman » Tue Oct 12, 2021 5:41 pm
Well, all draws among the top 3, but overall in the semifinals, with six of the best engines, Komodo Dragon has 18 wins to 0 losses, Stockfish 17 wins to 0 losses, and Lc0 16 wins to 0 losses as of this writing. So FRC does seem to allow for plenty of wins when there is some strength disparity, even at 3500+ levels. But I suspect that with long time limits, the top engines on this giant hardware would indeed draw 99+% of the games. Standard chess with no specified openings is even more drawish. The best engines are far from perfect, but they probably play well enough at long time limits on big hardware that they should never practically never lose a game unless forced to play an inferior opening. The draw margin in chess is just much larger than the errors made at this level. We should really find a better way to address this problem than just having each organizer pick bad openings arbitrarily. We need a solution that provides variety, decisive results, and makes some kind of sense logically.
It is curious that people can see the truth. But will not take the next logical step.
1. Yes chess engines are far from perfect. 2. And yes the top engines have a high draw rate when playing each other. And not just at long time controls, or on powerful hardware.
We are always stuck in today with our logic. What do I mean.
1. Chess engines from just 1 year ago also had a very high draw rate. And this is true for 2 years ago, and 3 years ago and so on.
But if you play a 1 year old chess engine, or a 2 year old chess engine, or a 3 year old chess engine. With today's best chess engines. All the older chess engines draw rate falls when playing the best of todays chess engines. As they lose more games, to the better chess engines of today.
And this is not really hard to understand. Chess is a 2 player game, and chess engines are far from perfect. But it takes a better chess engine to refute the mistakes. As seen when the older chess engines are playing the best chess engines today.
And this will also be the result of todays best chess engines. As better chess engines are developed. The situation today is different because we have higher draw rate then the draw rate that we had 3 years ago and chess engines clearly have problems to beat top engines that are one years old
I remember the match when stockfish14 won against stockfish13 only 3-0 and 997 draws. "I remember the match when stockfish14 won against stockfish13 only 3-0 and 997 draws."
Yes but the engines are not that much different in age. And Fishtest clearly has a issue in testing. But this is not surprising as if you only test at bullet for improvements. That is what you are most likely to get. Or sometimes worse when testing at longer time controls.
Here is the latest run, SF 14 vs Stockfish 19/10/21. Score of Stockfish 19/10/21 vs Stockfish 14: 0 - 2 - 144 [0.493] ... Stockfish 19/10/21 playing White: 0 - 0 - 73 [0.500] 73 ... Stockfish 19/10/21 playing Black: 0 - 2 - 71 [0.486] 73 ... White vs Black: 2 - 0 - 144 [0.507] 146 Elo difference: -4.8 +/- 6.5, LOS: 7.9 %, DrawRatio: 98.6 % 146 of 500 games finished.
But here is what your logic means. You are claiming the draw rate of today's engines will remain the same. In 1 year, 3 years and forever. When playing newer engines. You are basically claiming Stockfish has softly solved chess.
Do you want to bet money on your conclusions. Nם I do not claim that forever engines of today are not going to lose but only that the situation today is different
I do not believe that engines are perfect but I believe that even with unequal time control they usually do not make mistakes in games between them simply because they do not play in the right way to help the opponent to make mistakes so practically you will get more than 90% draws even if you play unequal time control like 5 minutes per game+5 seconds per move against 50 minutes per game+50 minutes per move.
In other words more time help the engine to avoid losing mistakes but does not help the engine to play moves that cause the opponent to make losing mistakes. "I do not claim that forever engines of today are not going to lose but only that the situation today is different." Then it is not different today as I said. As this is how it has always been! | |
| | | Uri Blass
Posts : 193 Join date : 2020-11-28
| Subject: Re: CCC FRC semifinal: all games draw so far.. Wed Oct 20, 2021 10:54 pm | |
| - mwyoung wrote:
- Uri Blass wrote:
- mwyoung wrote:
- Uri Blass wrote:
- mwyoung wrote:
-
- Code:
-
Re: CCC FRC semifinal: all games draw so far.. Post by lkaufman » Tue Oct 12, 2021 5:41 pm
Well, all draws among the top 3, but overall in the semifinals, with six of the best engines, Komodo Dragon has 18 wins to 0 losses, Stockfish 17 wins to 0 losses, and Lc0 16 wins to 0 losses as of this writing. So FRC does seem to allow for plenty of wins when there is some strength disparity, even at 3500+ levels. But I suspect that with long time limits, the top engines on this giant hardware would indeed draw 99+% of the games. Standard chess with no specified openings is even more drawish. The best engines are far from perfect, but they probably play well enough at long time limits on big hardware that they should never practically never lose a game unless forced to play an inferior opening. The draw margin in chess is just much larger than the errors made at this level. We should really find a better way to address this problem than just having each organizer pick bad openings arbitrarily. We need a solution that provides variety, decisive results, and makes some kind of sense logically.
It is curious that people can see the truth. But will not take the next logical step.
1. Yes chess engines are far from perfect. 2. And yes the top engines have a high draw rate when playing each other. And not just at long time controls, or on powerful hardware.
We are always stuck in today with our logic. What do I mean.
1. Chess engines from just 1 year ago also had a very high draw rate. And this is true for 2 years ago, and 3 years ago and so on.
But if you play a 1 year old chess engine, or a 2 year old chess engine, or a 3 year old chess engine. With today's best chess engines. All the older chess engines draw rate falls when playing the best of todays chess engines. As they lose more games, to the better chess engines of today.
And this is not really hard to understand. Chess is a 2 player game, and chess engines are far from perfect. But it takes a better chess engine to refute the mistakes. As seen when the older chess engines are playing the best chess engines today.
And this will also be the result of todays best chess engines. As better chess engines are developed. The situation today is different because we have higher draw rate then the draw rate that we had 3 years ago and chess engines clearly have problems to beat top engines that are one years old
I remember the match when stockfish14 won against stockfish13 only 3-0 and 997 draws. "I remember the match when stockfish14 won against stockfish13 only 3-0 and 997 draws."
Yes but the engines are not that much different in age. And Fishtest clearly has a issue in testing. But this is not surprising as if you only test at bullet for improvements. That is what you are most likely to get. Or sometimes worse when testing at longer time controls.
Here is the latest run, SF 14 vs Stockfish 19/10/21. Score of Stockfish 19/10/21 vs Stockfish 14: 0 - 2 - 144 [0.493] ... Stockfish 19/10/21 playing White: 0 - 0 - 73 [0.500] 73 ... Stockfish 19/10/21 playing Black: 0 - 2 - 71 [0.486] 73 ... White vs Black: 2 - 0 - 144 [0.507] 146 Elo difference: -4.8 +/- 6.5, LOS: 7.9 %, DrawRatio: 98.6 % 146 of 500 games finished.
But here is what your logic means. You are claiming the draw rate of today's engines will remain the same. In 1 year, 3 years and forever. When playing newer engines. You are basically claiming Stockfish has softly solved chess.
Do you want to bet money on your conclusions. Nם I do not claim that forever engines of today are not going to lose but only that the situation today is different
I do not believe that engines are perfect but I believe that even with unequal time control they usually do not make mistakes in games between them simply because they do not play in the right way to help the opponent to make mistakes so practically you will get more than 90% draws even if you play unequal time control like 5 minutes per game+5 seconds per move against 50 minutes per game+50 minutes per move.
In other words more time help the engine to avoid losing mistakes but does not help the engine to play moves that cause the opponent to make losing mistakes. "I do not claim that forever engines of today are not going to lose but only that the situation today is different."
Then it is not different today as I said. As this is how it has always been! My point is that if you play stockfish11 at time control 1+1 against stockfish11 at time control 10+10 time you are going to get significantly less draws than you get if you do the same with stockfish14 at time control 1+1 against stockfish14 at time control 10+10 | |
| | | mwyoung
Posts : 880 Join date : 2020-11-25 Location : USA
| Subject: Re: CCC FRC semifinal: all games draw so far.. Thu Oct 21, 2021 12:12 am | |
| - Uri Blass wrote:
- mwyoung wrote:
- Uri Blass wrote:
- mwyoung wrote:
- Uri Blass wrote:
- mwyoung wrote:
-
- Code:
-
Re: CCC FRC semifinal: all games draw so far.. Post by lkaufman » Tue Oct 12, 2021 5:41 pm
Well, all draws among the top 3, but overall in the semifinals, with six of the best engines, Komodo Dragon has 18 wins to 0 losses, Stockfish 17 wins to 0 losses, and Lc0 16 wins to 0 losses as of this writing. So FRC does seem to allow for plenty of wins when there is some strength disparity, even at 3500+ levels. But I suspect that with long time limits, the top engines on this giant hardware would indeed draw 99+% of the games. Standard chess with no specified openings is even more drawish. The best engines are far from perfect, but they probably play well enough at long time limits on big hardware that they should never practically never lose a game unless forced to play an inferior opening. The draw margin in chess is just much larger than the errors made at this level. We should really find a better way to address this problem than just having each organizer pick bad openings arbitrarily. We need a solution that provides variety, decisive results, and makes some kind of sense logically.
It is curious that people can see the truth. But will not take the next logical step.
1. Yes chess engines are far from perfect. 2. And yes the top engines have a high draw rate when playing each other. And not just at long time controls, or on powerful hardware.
We are always stuck in today with our logic. What do I mean.
1. Chess engines from just 1 year ago also had a very high draw rate. And this is true for 2 years ago, and 3 years ago and so on.
But if you play a 1 year old chess engine, or a 2 year old chess engine, or a 3 year old chess engine. With today's best chess engines. All the older chess engines draw rate falls when playing the best of todays chess engines. As they lose more games, to the better chess engines of today.
And this is not really hard to understand. Chess is a 2 player game, and chess engines are far from perfect. But it takes a better chess engine to refute the mistakes. As seen when the older chess engines are playing the best chess engines today.
And this will also be the result of todays best chess engines. As better chess engines are developed. The situation today is different because we have higher draw rate then the draw rate that we had 3 years ago and chess engines clearly have problems to beat top engines that are one years old
I remember the match when stockfish14 won against stockfish13 only 3-0 and 997 draws. "I remember the match when stockfish14 won against stockfish13 only 3-0 and 997 draws."
Yes but the engines are not that much different in age. And Fishtest clearly has a issue in testing. But this is not surprising as if you only test at bullet for improvements. That is what you are most likely to get. Or sometimes worse when testing at longer time controls.
Here is the latest run, SF 14 vs Stockfish 19/10/21. Score of Stockfish 19/10/21 vs Stockfish 14: 0 - 2 - 144 [0.493] ... Stockfish 19/10/21 playing White: 0 - 0 - 73 [0.500] 73 ... Stockfish 19/10/21 playing Black: 0 - 2 - 71 [0.486] 73 ... White vs Black: 2 - 0 - 144 [0.507] 146 Elo difference: -4.8 +/- 6.5, LOS: 7.9 %, DrawRatio: 98.6 % 146 of 500 games finished.
But here is what your logic means. You are claiming the draw rate of today's engines will remain the same. In 1 year, 3 years and forever. When playing newer engines. You are basically claiming Stockfish has softly solved chess.
Do you want to bet money on your conclusions. Nם I do not claim that forever engines of today are not going to lose but only that the situation today is different
I do not believe that engines are perfect but I believe that even with unequal time control they usually do not make mistakes in games between them simply because they do not play in the right way to help the opponent to make mistakes so practically you will get more than 90% draws even if you play unequal time control like 5 minutes per game+5 seconds per move against 50 minutes per game+50 minutes per move.
In other words more time help the engine to avoid losing mistakes but does not help the engine to play moves that cause the opponent to make losing mistakes. "I do not claim that forever engines of today are not going to lose but only that the situation today is different."
Then it is not different today as I said. As this is how it has always been! My point is that if you play stockfish11 at time control 1+1 against stockfish11 at time control 10+10 time you are going to get significantly less draws than you get if you do the same with stockfish14 at time control 1+1 against stockfish14 at time control 10+10
So what, If that is the results of the scaling Stockfish 14. Then that is the results. That would be a issue with NNUE. But has nothing to do with anything else. Why must you change the draw rate to suite your desire for more wins. That is not testing. If Stockfish NNUE can not beat Stockfish NNUE with more time. Then that is the results! | |
| | | Admin Admin
Posts : 2522 Join date : 2020-11-17 Location : Netherlands
| Subject: Re: CCC FRC semifinal: all games draw so far.. Thu Oct 21, 2021 9:51 pm | |
| Still low draw rates with 20 cores. Maybe unbalanced -- but playable -- positions is the future, at least for the entertaining part. | |
| | | mwyoung
Posts : 880 Join date : 2020-11-25 Location : USA
| Subject: Re: CCC FRC semifinal: all games draw so far.. Fri Oct 22, 2021 2:35 am | |
| - Admin wrote:
Still low draw rates with 20 cores.
Maybe unbalanced -- but playable -- positions is the future, at least for the entertaining part. It maybe playable, but what would happen if you could only force one side to play your line. And the other engine is not forced into your line of play. But could play what it thinks is the best moves. Hmmm. And as I have said many times. Draw Rate is a meaningless metric. All I need to know is wins, losses, and draws. As 50 wins and 50 losses and 0 draws. Is equal to 0 wins and 0 losses and 100 draws. | |
| | | Admin Admin
Posts : 2522 Join date : 2020-11-17 Location : Netherlands
| Subject: Re: CCC FRC semifinal: all games draw so far.. Fri Oct 22, 2021 9:46 am | |
| - mwyoung wrote:
- Admin wrote:
Still low draw rates with 20 cores.
Maybe unbalanced -- but playable -- positions is the future, at least for the entertaining part. It maybe playable, but what would happen if you could only force one side to play your line. And the other engine is not forced into your line of play. But could play what it thinks is the best moves. Hmmm.
And as I have said many times. Draw Rate is a meaningless metric.
All I need to know is wins, losses, and draws. As 50 wins and 50 losses and 0 draws. Is equal to 0 wins and 0 losses and 100 draws. It's not the same, the former result is an indication of two low rated engines with lots of progress ahead while the latter can only happen between top engines on hardware like CCC or TCEC. | |
| | | Sponsored content
| Subject: Re: CCC FRC semifinal: all games draw so far.. | |
| |
| | | | CCC FRC semifinal: all games draw so far.. | |
|
Similar topics | |
|
| Permissions in this forum: | You cannot reply to topics in this forum
| |
| |
| |