Why there is no interest in Computer with odds Vs Humans match?

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

Moderators: hgm, Harvey Williamson, bob

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
Gary Internet
Posts: 37
Joined: Thu Jan 04, 2018 6:09 pm

Re: Why there is no interest in Computer with odds Vs Humans match?

Post by Gary Internet » Tue Jul 02, 2019 5:02 am

Speaking of Fritz Bahrain check out this Reddit thread - 15 Year of Chess Engine Development

lkaufman
Posts: 3538
Joined: Sun Jan 10, 2010 5:15 am
Location: Maryland USA
Contact:

Re: Why there is no interest in Computer with odds Vs Humans match?

Post by lkaufman » Tue Jul 02, 2019 6:01 am

Chessqueen wrote:
Tue Jul 02, 2019 3:29 am
lkaufman wrote:
Mon Jul 01, 2019 5:03 pm
Nordlandia wrote:
Mon Jul 01, 2019 3:21 pm
Maybe odds matches in Fischer Random is harder for the human side. Humans are used to the standard opening setup. The unfamiliarity of Chess960 may give the engine the upper hand?

Maybe more odds is needed to equalize the chances in Fischer Random.

What is your thought on this one Larry?
Well we already "proved" that this is so, John Meyer won 3 to 1 at regular knight odds vs Komodo, but lost 0-4 at frc knight odds, both 45' + 15". Komodo has no chance giving knight odds in normal chess to strong GMs at rapid (not blitz) time controls, but may have some chance at frc knight odds, depending on the time control and the rating of the GM. But even with frc, knight odds is a lot to give to a GM with enough time to think.
I was also wondering what would be the outcome of pitting Komodo versus The fritz version that played versus Kramnik about 10 years ago Deep Fritz Bahrain with the f3 pawn or the f7 pawn odds in 50 games game in 30 minutes TC. What would your prediction be ?
First of all, Deep Fritz Bahrain played Kramnik in 2002, 17 years ago, not 10, and drew the match on hardware of that time. Presumably on my i7 laptop that same Fritz on 8 threads would be at least a couple hundred elo stronger, over FIDE 3000 surely. But to answer your question, I can't actually do the test on our tester since I don't have that ancient engine and anyway need something that runs in Linux on current hardware, but perhaps I can run some sort of test on something similarly rated on CCRL or CEGT lists. I am quite certain that Komodo would win such a match at f7 handicap (f2 would just be a blowout); there would probably be a fair number of draws, but Komodo would win a lot more than Fritz. Evan at f7 plus second White move handicap Komodo would be favored. Knight odds would be too much though. If someone knows of a suitable engine with a free and reliable Linux version, please advise.
Komodo rules!

Chessqueen
Posts: 547
Joined: Wed Sep 05, 2018 12:16 am
Full name: Nancy M Pichardo

Re: Why there is no interest in Computer with odds Vs Humans match?

Post by Chessqueen » Tue Jul 02, 2019 2:34 pm

lkaufman wrote:
Tue Jul 02, 2019 6:01 am
Chessqueen wrote:
Tue Jul 02, 2019 3:29 am
lkaufman wrote:
Mon Jul 01, 2019 5:03 pm
Nordlandia wrote:
Mon Jul 01, 2019 3:21 pm
Maybe odds matches in Fischer Random is harder for the human side. Humans are used to the standard opening setup. The unfamiliarity of Chess960 may give the engine the upper hand?

Maybe more odds is needed to equalize the chances in Fischer Random.

What is your thought on this one Larry?
Well we already "proved" that this is so, John Meyer won 3 to 1 at regular knight odds vs Komodo, but lost 0-4 at frc knight odds, both 45' + 15". Komodo has no chance giving knight odds in normal chess to strong GMs at rapid (not blitz) time controls, but may have some chance at frc knight odds, depending on the time control and the rating of the GM. But even with frc, knight odds is a lot to give to a GM with enough time to think.
I was also wondering what would be the outcome of pitting Komodo versus The fritz version that played versus Kramnik about 10 years ago Deep Fritz Bahrain with the f3 pawn or the f7 pawn odds in 50 games game in 30 minutes TC. What would your prediction be ?
First of all, Deep Fritz Bahrain played Kramnik in 2002, 17 years ago, not 10, and drew the match on hardware of that time. Presumably on my i7 laptop that same Fritz on 8 threads would be at least a couple hundred elo stronger, over FIDE 3000 surely. But to answer your question, I can't actually do the test on our tester since I don't have that ancient engine and anyway need something that runs in Linux on current hardware, but perhaps I can run some sort of test on something similarly rated on CCRL or CEGT lists. I am quite certain that Komodo would win such a match at f7 handicap (f2 would just be a blowout); there would probably be a fair number of draws, but Komodo would win a lot more than Fritz. Evan at f7 plus second White move handicap Komodo would be favored. Knight odds would be too much though. If someone knows of a suitable engine with a free and reliable Linux version, please advise.

WOW it has been 17 years already! Another program that you can pit is Deep Junior the same that play versus Gary Kasparov back in 2003 which also ended in a Draw. Since DJ and Kasparov were close in rating back in 2003 and it is hard to get a good Grandmaster that wants to play versus Komodo. and Carlsen will never agree since he is too proud to lose versus a computer. We can get Deep Junior with a similar hardware that can hit the same MNPS Millions nodes per second and give Deep Junior a handicap of the f2 pawn, or the f7 pawn odds. If Komodo wins easy then 2 pawns odds which to me it would be very hard for Komodo to even get a draw versus another engine, with a human they have the tendency to get nervous and sometimes blunder versus computer.
https://www.youtube.com/watch?v=WWYaLg21Dss
https://en.chessbase.com/post/kasparov- ... -sacrifice

lkaufman
Posts: 3538
Joined: Sun Jan 10, 2010 5:15 am
Location: Maryland USA
Contact:

Re: Why there is no interest in Computer with odds Vs Humans match?

Post by lkaufman » Tue Jul 02, 2019 6:54 pm

Chessqueen wrote:
Tue Jul 02, 2019 2:34 pm
lkaufman wrote:
Tue Jul 02, 2019 6:01 am
Chessqueen wrote:
Tue Jul 02, 2019 3:29 am
lkaufman wrote:
Mon Jul 01, 2019 5:03 pm
Nordlandia wrote:
Mon Jul 01, 2019 3:21 pm
Maybe odds matches in Fischer Random is harder for the human side. Humans are used to the standard opening setup. The unfamiliarity of Chess960 may give the engine the upper hand?

Maybe more odds is needed to equalize the chances in Fischer Random.

What is your thought on this one Larry?
Well we already "proved" that this is so, John Meyer won 3 to 1 at regular knight odds vs Komodo, but lost 0-4 at frc knight odds, both 45' + 15". Komodo has no chance giving knight odds in normal chess to strong GMs at rapid (not blitz) time controls, but may have some chance at frc knight odds, depending on the time control and the rating of the GM. But even with frc, knight odds is a lot to give to a GM with enough time to think.
I was also wondering what would be the outcome of pitting Komodo versus The fritz version that played versus Kramnik about 10 years ago Deep Fritz Bahrain with the f3 pawn or the f7 pawn odds in 50 games game in 30 minutes TC. What would your prediction be ?
First of all, Deep Fritz Bahrain played Kramnik in 2002, 17 years ago, not 10, and drew the match on hardware of that time. Presumably on my i7 laptop that same Fritz on 8 threads would be at least a couple hundred elo stronger, over FIDE 3000 surely. But to answer your question, I can't actually do the test on our tester since I don't have that ancient engine and anyway need something that runs in Linux on current hardware, but perhaps I can run some sort of test on something similarly rated on CCRL or CEGT lists. I am quite certain that Komodo would win such a match at f7 handicap (f2 would just be a blowout); there would probably be a fair number of draws, but Komodo would win a lot more than Fritz. Evan at f7 plus second White move handicap Komodo would be favored. Knight odds would be too much though. If someone knows of a suitable engine with a free and reliable Linux version, please advise.

WOW it has been 17 years already! Another program that you can pit is Deep Junior the same that play versus Gary Kasparov back in 2003 which also ended in a Draw. Since DJ and Kasparov were close in rating back in 2003 and it is hard to get a good Grandmaster that wants to play versus Komodo. and Carlsen will never agree since he is too proud to lose versus a computer. We can get Deep Junior with a similar hardware that can hit the same MNPS Millions nodes per second and give Deep Junior a handicap of the f2 pawn, or the f7 pawn odds. If Komodo wins easy then 2 pawns odds which to me it would be very hard for Komodo to even get a draw versus another engine, with a human they have the tendency to get nervous and sometimes blunder versus computer.
https://www.youtube.com/watch?v=WWYaLg21Dss
https://en.chessbase.com/post/kasparov- ... -sacrifice
Okay, I've started the following test. I don't have any engines as weak as those old Fritz and Junior versions on our Linux tester, the weakest one I have on the tester is Arasan 18.0, which is of course not weak at all (2860 CCRL 40/40 single thread). I am reasonably sure that with 1/4 of the time on one thread it is stronger than the two mentioned old engines on whatever number of threads they ran on back then. They are too old to find on the list, but if you extrapolate backward you can see that they should be somewhere below 2700, so this should be correct. I picked a time limit of 5' + 2", which if quadrupled would give games of about the same total time as 30' sudden death. I am running latest Komodo on four threads against Arasan 18.0 on one thread, with Komodo giving the f7 pawn handicap (which in self-testing at one minute + .4" inc. is nearly a 500 elo handicap based on 700 games). Note that this test is very unfair to Komodo (compared to the one you ask for) because Komodo is also only getting 5' + 2", it should really get quadruple this time. Also Komodo is only given four threads, but in an actual match with a top GM we would use maybe 47 threads on a 48 core machine. Also of course Arasan (and Komodo) are running on modern hardware, not 2002/2003 hardware, which makes a huge difference. On the other hand, a human GM would know that he was playing a much stronger entity and would avoid complications and try to simplify, which we can't expect an older engine to do. So maybe these factors balance out. I'll report results when I have them.
Komodo rules!

Chessqueen
Posts: 547
Joined: Wed Sep 05, 2018 12:16 am
Full name: Nancy M Pichardo

Re: Why there is no interest in Computer with odds Vs Humans match?

Post by Chessqueen » Tue Jul 02, 2019 7:31 pm

lkaufman wrote:
Tue Jul 02, 2019 6:54 pm
Chessqueen wrote:
Tue Jul 02, 2019 2:34 pm
lkaufman wrote:
Tue Jul 02, 2019 6:01 am
Chessqueen wrote:
Tue Jul 02, 2019 3:29 am
lkaufman wrote:
Mon Jul 01, 2019 5:03 pm
Nordlandia wrote:
Mon Jul 01, 2019 3:21 pm
Maybe odds matches in Fischer Random is harder for the human side. Humans are used to the standard opening setup. The unfamiliarity of Chess960 may give the engine the upper hand?

Maybe more odds is needed to equalize the chances in Fischer Random.

What is your thought on this one Larry?
Well we already "proved" that this is so, John Meyer won 3 to 1 at regular knight odds vs Komodo, but lost 0-4 at frc knight odds, both 45' + 15". Komodo has no chance giving knight odds in normal chess to strong GMs at rapid (not blitz) time controls, but may have some chance at frc knight odds, depending on the time control and the rating of the GM. But even with frc, knight odds is a lot to give to a GM with enough time to think.
I was also wondering what would be the outcome of pitting Komodo versus The fritz version that played versus Kramnik about 10 years ago Deep Fritz Bahrain with the f3 pawn or the f7 pawn odds in 50 games game in 30 minutes TC. What would your prediction be ?
First of all, Deep Fritz Bahrain played Kramnik in 2002, 17 years ago, not 10, and drew the match on hardware of that time. Presumably on my i7 laptop that same Fritz on 8 threads would be at least a couple hundred elo stronger, over FIDE 3000 surely. But to answer your question, I can't actually do the test on our tester since I don't have that ancient engine and anyway need something that runs in Linux on current hardware, but perhaps I can run some sort of test on something similarly rated on CCRL or CEGT lists. I am quite certain that Komodo would win such a match at f7 handicap (f2 would just be a blowout); there would probably be a fair number of draws, but Komodo would win a lot more than Fritz. Evan at f7 plus second White move handicap Komodo would be favored. Knight odds would be too much though. If someone knows of a suitable engine with a free and reliable Linux version, please advise.

WOW it has been 17 years already! Another program that you can pit is Deep Junior the same that play versus Gary Kasparov back in 2003 which also ended in a Draw. Since DJ and Kasparov were close in rating back in 2003 and it is hard to get a good Grandmaster that wants to play versus Komodo. and Carlsen will never agree since he is too proud to lose versus a computer. We can get Deep Junior with a similar hardware that can hit the same MNPS Millions nodes per second and give Deep Junior a handicap of the f2 pawn, or the f7 pawn odds. If Komodo wins easy then 2 pawns odds which to me it would be very hard for Komodo to even get a draw versus another engine, with a human they have the tendency to get nervous and sometimes blunder versus computer.
https://www.youtube.com/watch?v=WWYaLg21Dss
https://en.chessbase.com/post/kasparov- ... -sacrifice
Okay, I've started the following test. I don't have any engines as weak as those old Fritz and Junior versions on our Linux tester, the weakest one I have on the tester is Arasan 18.0, which is of course not weak at all (2860 CCRL 40/40 single thread). I am reasonably sure that with 1/4 of the time on one thread it is stronger than the two mentioned old engines on whatever number of threads they ran on back then. They are too old to find on the list, but if you extrapolate backward you can see that they should be somewhere below 2700, so this should be correct. I picked a time limit of 5' + 2", which if quadrupled would give games of about the same total time as 30' sudden death. I am running latest Komodo on four threads against Arasan 18.0 on one thread, with Komodo giving the f7 pawn handicap (which in self-testing at one minute + .4" inc. is nearly a 500 elo handicap based on 700 games). Note that this test is very unfair to Komodo (compared to the one you ask for) because Komodo is also only getting 5' + 2", it should really get quadruple this time. Also Komodo is only given four threads, but in an actual match with a top GM we would use maybe 47 threads on a 48 core machine. Also of course Arasan (and Komodo) are running on modern hardware, not 2002/2003 hardware, which makes a huge difference. On the other hand, a human GM would know that he was playing a much stronger entity and would avoid complications and try to simplify, which we can't expect an older engine to do. So maybe these factors balance out. I'll report results when I have them.
I was thinking if you were going to use Linux about Demolito ==> https://github.com/lucasart/Demolito

I am also going to do some test using this version of Araran ==> https://www.arasanchess.org/

Using GCC on Linux, type: make. Otherwise, just have a look at the makefile, nothing fancy there. You should also be able to compile it on any platform (POSIX or Windows), using a C11 capable compiler (eg. GCC or Clang but not MSVC). But you'll have to figure out the exact commands for yourself, depending on your compiler, your libc, and your system.

lkaufman
Posts: 3538
Joined: Sun Jan 10, 2010 5:15 am
Location: Maryland USA
Contact:

Re: Why there is no interest in Computer with odds Vs Humans match?

Post by lkaufman » Tue Jul 02, 2019 8:19 pm

Chessqueen wrote:
Tue Jul 02, 2019 7:31 pm
lkaufman wrote:
Tue Jul 02, 2019 6:54 pm
Chessqueen wrote:
Tue Jul 02, 2019 2:34 pm
lkaufman wrote:
Tue Jul 02, 2019 6:01 am
Chessqueen wrote:
Tue Jul 02, 2019 3:29 am
lkaufman wrote:
Mon Jul 01, 2019 5:03 pm
Nordlandia wrote:
Mon Jul 01, 2019 3:21 pm
Maybe odds matches in Fischer Random is harder for the human side. Humans are used to the standard opening setup. The unfamiliarity of Chess960 may give the engine the upper hand?

Maybe more odds is needed to equalize the chances in Fischer Random.

What is your thought on this one Larry?
Well we already "proved" that this is so, John Meyer won 3 to 1 at regular knight odds vs Komodo, but lost 0-4 at frc knight odds, both 45' + 15". Komodo has no chance giving knight odds in normal chess to strong GMs at rapid (not blitz) time controls, but may have some chance at frc knight odds, depending on the time control and the rating of the GM. But even with frc, knight odds is a lot to give to a GM with enough time to think.
I was also wondering what would be the outcome of pitting Komodo versus The fritz version that played versus Kramnik about 10 years ago Deep Fritz Bahrain with the f3 pawn or the f7 pawn odds in 50 games game in 30 minutes TC. What would your prediction be ?
First of all, Deep Fritz Bahrain played Kramnik in 2002, 17 years ago, not 10, and drew the match on hardware of that time. Presumably on my i7 laptop that same Fritz on 8 threads would be at least a couple hundred elo stronger, over FIDE 3000 surely. But to answer your question, I can't actually do the test on our tester since I don't have that ancient engine and anyway need something that runs in Linux on current hardware, but perhaps I can run some sort of test on something similarly rated on CCRL or CEGT lists. I am quite certain that Komodo would win such a match at f7 handicap (f2 would just be a blowout); there would probably be a fair number of draws, but Komodo would win a lot more than Fritz. Evan at f7 plus second White move handicap Komodo would be favored. Knight odds would be too much though. If someone knows of a suitable engine with a free and reliable Linux version, please advise.

WOW it has been 17 years already! Another program that you can pit is Deep Junior the same that play versus Gary Kasparov back in 2003 which also ended in a Draw. Since DJ and Kasparov were close in rating back in 2003 and it is hard to get a good Grandmaster that wants to play versus Komodo. and Carlsen will never agree since he is too proud to lose versus a computer. We can get Deep Junior with a similar hardware that can hit the same MNPS Millions nodes per second and give Deep Junior a handicap of the f2 pawn, or the f7 pawn odds. If Komodo wins easy then 2 pawns odds which to me it would be very hard for Komodo to even get a draw versus another engine, with a human they have the tendency to get nervous and sometimes blunder versus computer.
https://www.youtube.com/watch?v=WWYaLg21Dss
https://en.chessbase.com/post/kasparov- ... -sacrifice
Okay, I've started the following test. I don't have any engines as weak as those old Fritz and Junior versions on our Linux tester, the weakest one I have on the tester is Arasan 18.0, which is of course not weak at all (2860 CCRL 40/40 single thread). I am reasonably sure that with 1/4 of the time on one thread it is stronger than the two mentioned old engines on whatever number of threads they ran on back then. They are too old to find on the list, but if you extrapolate backward you can see that they should be somewhere below 2700, so this should be correct. I picked a time limit of 5' + 2", which if quadrupled would give games of about the same total time as 30' sudden death. I am running latest Komodo on four threads against Arasan 18.0 on one thread, with Komodo giving the f7 pawn handicap (which in self-testing at one minute + .4" inc. is nearly a 500 elo handicap based on 700 games). Note that this test is very unfair to Komodo (compared to the one you ask for) because Komodo is also only getting 5' + 2", it should really get quadruple this time. Also Komodo is only given four threads, but in an actual match with a top GM we would use maybe 47 threads on a 48 core machine. Also of course Arasan (and Komodo) are running on modern hardware, not 2002/2003 hardware, which makes a huge difference. On the other hand, a human GM would know that he was playing a much stronger entity and would avoid complications and try to simplify, which we can't expect an older engine to do. So maybe these factors balance out. I'll report results when I have them.
I was thinking if you were going to use Linux about Demolito ==> https://github.com/lucasart/Demolito

I am also going to do some test using this version of Araran ==> https://www.arasanchess.org/

Using GCC on Linux, type: make. Otherwise, just have a look at the makefile, nothing fancy there. You should also be able to compile it on any platform (POSIX or Windows), using a C11 capable compiler (eg. GCC or Clang but not MSVC). But you'll have to figure out the exact commands for yourself, depending on your compiler, your libc, and your system.
Why would you want to use the latest Arasan for this rather than an older one like 18.0? Even 18.0 is way stronger than the engines of 2002/2003. Maybe you meant to use Arasan 14?
Anyway, the test above surprised me; Komodo won by an incredible 87 wins to 2 losses, with 33 draws, for a plus 320 elo result despite the nearly 500 elo handicap! So I raised the handicap to pawn and two moves (f7 off, e4 played, WTM) and this made a big difference, but still Komodo won decisively (54 to 13 with 41 draws, plus 144 elo). All this without even raising Contempt from the default value! So now I'm also giving Arasan the same four threads as Komodo, with the pawn and two move handicap; this should be enough for Arasan to close the gap. But then this is a simulation of playing against an entity roughly a class stronger than the Fritz and Junior in question.
Komodo rules!

lkaufman
Posts: 3538
Joined: Sun Jan 10, 2010 5:15 am
Location: Maryland USA
Contact:

Re: Why there is no interest in Computer with odds Vs Humans match?

Post by lkaufman » Tue Jul 02, 2019 10:13 pm

lkaufman wrote:
Tue Jul 02, 2019 8:19 pm
Chessqueen wrote:
Tue Jul 02, 2019 7:31 pm
lkaufman wrote:
Tue Jul 02, 2019 6:54 pm
Chessqueen wrote:
Tue Jul 02, 2019 2:34 pm
lkaufman wrote:
Tue Jul 02, 2019 6:01 am
Chessqueen wrote:
Tue Jul 02, 2019 3:29 am
lkaufman wrote:
Mon Jul 01, 2019 5:03 pm
Nordlandia wrote:
Mon Jul 01, 2019 3:21 pm
Maybe odds matches in Fischer Random is harder for the human side. Humans are used to the standard opening setup. The unfamiliarity of Chess960 may give the engine the upper hand?

Maybe more odds is needed to equalize the chances in Fischer Random.

What is your thought on this one Larry?
Well we already "proved" that this is so, John Meyer won 3 to 1 at regular knight odds vs Komodo, but lost 0-4 at frc knight odds, both 45' + 15". Komodo has no chance giving knight odds in normal chess to strong GMs at rapid (not blitz) time controls, but may have some chance at frc knight odds, depending on the time control and the rating of the GM. But even with frc, knight odds is a lot to give to a GM with enough time to think.
I was also wondering what would be the outcome of pitting Komodo versus The fritz version that played versus Kramnik about 10 years ago Deep Fritz Bahrain with the f3 pawn or the f7 pawn odds in 50 games game in 30 minutes TC. What would your prediction be ?
First of all, Deep Fritz Bahrain played Kramnik in 2002, 17 years ago, not 10, and drew the match on hardware of that time. Presumably on my i7 laptop that same Fritz on 8 threads would be at least a couple hundred elo stronger, over FIDE 3000 surely. But to answer your question, I can't actually do the test on our tester since I don't have that ancient engine and anyway need something that runs in Linux on current hardware, but perhaps I can run some sort of test on something similarly rated on CCRL or CEGT lists. I am quite certain that Komodo would win such a match at f7 handicap (f2 would just be a blowout); there would probably be a fair number of draws, but Komodo would win a lot more than Fritz. Evan at f7 plus second White move handicap Komodo would be favored. Knight odds would be too much though. If someone knows of a suitable engine with a free and reliable Linux version, please advise.

WOW it has been 17 years already! Another program that you can pit is Deep Junior the same that play versus Gary Kasparov back in 2003 which also ended in a Draw. Since DJ and Kasparov were close in rating back in 2003 and it is hard to get a good Grandmaster that wants to play versus Komodo. and Carlsen will never agree since he is too proud to lose versus a computer. We can get Deep Junior with a similar hardware that can hit the same MNPS Millions nodes per second and give Deep Junior a handicap of the f2 pawn, or the f7 pawn odds. If Komodo wins easy then 2 pawns odds which to me it would be very hard for Komodo to even get a draw versus another engine, with a human they have the tendency to get nervous and sometimes blunder versus computer.
https://www.youtube.com/watch?v=WWYaLg21Dss
https://en.chessbase.com/post/kasparov- ... -sacrifice
Okay, I've started the following test. I don't have any engines as weak as those old Fritz and Junior versions on our Linux tester, the weakest one I have on the tester is Arasan 18.0, which is of course not weak at all (2860 CCRL 40/40 single thread). I am reasonably sure that with 1/4 of the time on one thread it is stronger than the two mentioned old engines on whatever number of threads they ran on back then. They are too old to find on the list, but if you extrapolate backward you can see that they should be somewhere below 2700, so this should be correct. I picked a time limit of 5' + 2", which if quadrupled would give games of about the same total time as 30' sudden death. I am running latest Komodo on four threads against Arasan 18.0 on one thread, with Komodo giving the f7 pawn handicap (which in self-testing at one minute + .4" inc. is nearly a 500 elo handicap based on 700 games). Note that this test is very unfair to Komodo (compared to the one you ask for) because Komodo is also only getting 5' + 2", it should really get quadruple this time. Also Komodo is only given four threads, but in an actual match with a top GM we would use maybe 47 threads on a 48 core machine. Also of course Arasan (and Komodo) are running on modern hardware, not 2002/2003 hardware, which makes a huge difference. On the other hand, a human GM would know that he was playing a much stronger entity and would avoid complications and try to simplify, which we can't expect an older engine to do. So maybe these factors balance out. I'll report results when I have them.
I was thinking if you were going to use Linux about Demolito ==> https://github.com/lucasart/Demolito

I am also going to do some test using this version of Araran ==> https://www.arasanchess.org/

Using GCC on Linux, type: make. Otherwise, just have a look at the makefile, nothing fancy there. You should also be able to compile it on any platform (POSIX or Windows), using a C11 capable compiler (eg. GCC or Clang but not MSVC). But you'll have to figure out the exact commands for yourself, depending on your compiler, your libc, and your system.
Why would you want to use the latest Arasan for this rather than an older one like 18.0? Even 18.0 is way stronger than the engines of 2002/2003. Maybe you meant to use Arasan 14?
Anyway, the test above surprised me; Komodo won by an incredible 87 wins to 2 losses, with 33 draws, for a plus 320 elo result despite the nearly 500 elo handicap! So I raised the handicap to pawn and two moves (f7 off, e4 played, WTM) and this made a big difference, but still Komodo won decisively (54 to 13 with 41 draws, plus 144 elo). All this without even raising Contempt from the default value! So now I'm also giving Arasan the same four threads as Komodo, with the pawn and two move handicap; this should be enough for Arasan to close the gap. But then this is a simulation of playing against an entity roughly a class stronger than the Fritz and Junior in question.
With both Komodo and Arasan 18 running on four threads at pawn and two move handicap, the result was close, Komodo winning by 116 wins to 105 losses with 101 draws, +12 elo. Now I'm running the same match with Contempt = 50 to see if that helps.
Komodo rules!

Chessqueen
Posts: 547
Joined: Wed Sep 05, 2018 12:16 am
Full name: Nancy M Pichardo

Re: Why there is no interest in Computer with odds Vs Humans match?

Post by Chessqueen » Wed Jul 03, 2019 12:29 am

lkaufman wrote:
Tue Jul 02, 2019 10:13 pm
lkaufman wrote:
Tue Jul 02, 2019 8:19 pm
Chessqueen wrote:
Tue Jul 02, 2019 7:31 pm
lkaufman wrote:
Tue Jul 02, 2019 6:54 pm
Chessqueen wrote:
Tue Jul 02, 2019 2:34 pm
lkaufman wrote:
Tue Jul 02, 2019 6:01 am
Chessqueen wrote:
Tue Jul 02, 2019 3:29 am
lkaufman wrote:
Mon Jul 01, 2019 5:03 pm
Nordlandia wrote:
Mon Jul 01, 2019 3:21 pm
Maybe odds matches in Fischer Random is harder for the human side. Humans are used to the standard opening setup. The unfamiliarity of Chess960 may give the engine the upper hand?

Maybe more odds is needed to equalize the chances in Fischer Random.

What is your thought on this one Larry?
Well we already "proved" that this is so, John Meyer won 3 to 1 at regular knight odds vs Komodo, but lost 0-4 at frc knight odds, both 45' + 15". Komodo has no chance giving knight odds in normal chess to strong GMs at rapid (not blitz) time controls, but may have some chance at frc knight odds, depending on the time control and the rating of the GM. But even with frc, knight odds is a lot to give to a GM with enough time to think.
I was also wondering what would be the outcome of pitting Komodo versus The fritz version that played versus Kramnik about 10 years ago Deep Fritz Bahrain with the f3 pawn or the f7 pawn odds in 50 games game in 30 minutes TC. What would your prediction be ?
First of all, Deep Fritz Bahrain played Kramnik in 2002, 17 years ago, not 10, and drew the match on hardware of that time. Presumably on my i7 laptop that same Fritz on 8 threads would be at least a couple hundred elo stronger, over FIDE 3000 surely. But to answer your question, I can't actually do the test on our tester since I don't have that ancient engine and anyway need something that runs in Linux on current hardware, but perhaps I can run some sort of test on something similarly rated on CCRL or CEGT lists. I am quite certain that Komodo would win such a match at f7 handicap (f2 would just be a blowout); there would probably be a fair number of draws, but Komodo would win a lot more than Fritz. Evan at f7 plus second White move handicap Komodo would be favored. Knight odds would be too much though. If someone knows of a suitable engine with a free and reliable Linux version, please advise.

WOW it has been 17 years already! Another program that you can pit is Deep Junior the same that play versus Gary Kasparov back in 2003 which also ended in a Draw. Since DJ and Kasparov were close in rating back in 2003 and it is hard to get a good Grandmaster that wants to play versus Komodo. and Carlsen will never agree since he is too proud to lose versus a computer. We can get Deep Junior with a similar hardware that can hit the same MNPS Millions nodes per second and give Deep Junior a handicap of the f2 pawn, or the f7 pawn odds. If Komodo wins easy then 2 pawns odds which to me it would be very hard for Komodo to even get a draw versus another engine, with a human they have the tendency to get nervous and sometimes blunder versus computer.
https://www.youtube.com/watch?v=WWYaLg21Dss
https://en.chessbase.com/post/kasparov- ... -sacrifice
Okay, I've started the following test. I don't have any engines as weak as those old Fritz and Junior versions on our Linux tester, the weakest one I have on the tester is Arasan 18.0, which is of course not weak at all (2860 CCRL 40/40 single thread). I am reasonably sure that with 1/4 of the time on one thread it is stronger than the two mentioned old engines on whatever number of threads they ran on back then. They are too old to find on the list, but if you extrapolate backward you can see that they should be somewhere below 2700, so this should be correct. I picked a time limit of 5' + 2", which if quadrupled would give games of about the same total time as 30' sudden death. I am running latest Komodo on four threads against Arasan 18.0 on one thread, with Komodo giving the f7 pawn handicap (which in self-testing at one minute + .4" inc. is nearly a 500 elo handicap based on 700 games). Note that this test is very unfair to Komodo (compared to the one you ask for) because Komodo is also only getting 5' + 2", it should really get quadruple this time. Also Komodo is only given four threads, but in an actual match with a top GM we would use maybe 47 threads on a 48 core machine. Also of course Arasan (and Komodo) are running on modern hardware, not 2002/2003 hardware, which makes a huge difference. On the other hand, a human GM would know that he was playing a much stronger entity and would avoid complications and try to simplify, which we can't expect an older engine to do. So maybe these factors balance out. I'll report results when I have them.
I was thinking if you were going to use Linux about Demolito ==> https://github.com/lucasart/Demolito

I am also going to do some test using this version of Araran ==> https://www.arasanchess.org/

Using GCC on Linux, type: make. Otherwise, just have a look at the makefile, nothing fancy there. You should also be able to compile it on any platform (POSIX or Windows), using a C11 capable compiler (eg. GCC or Clang but not MSVC). But you'll have to figure out the exact commands for yourself, depending on your compiler, your libc, and your system.
Why would you want to use the latest Arasan for this rather than an older one like 18.0? Even 18.0 is way stronger than the engines of 2002/2003. Maybe you meant to use Arasan 14?
Anyway, the test above surprised me; Komodo won by an incredible 87 wins to 2 losses, with 33 draws, for a plus 320 elo result despite the nearly 500 elo handicap! So I raised the handicap to pawn and two moves (f7 off, e4 played, WTM) and this made a big difference, but still Komodo won decisively (54 to 13 with 41 draws, plus 144 elo). All this without even raising Contempt from the default value! So now I'm also giving Arasan the same four threads as Komodo, with the pawn and two move handicap; this should be enough for Arasan to close the gap. But then this is a simulation of playing against an entity roughly a class stronger than the Fritz and Junior in question.
With both Komodo and Arasan 18 running on four threads at pawn and two move handicap, the result was close, Komodo winning by 116 wins to 105 losses with 101 draws, +12 elo. Now I'm running the same match with Contempt = 50 to see if that helps.
I guess you need to find out if Komodo still can win with a Knight for a pawn odds that would be interesting to find out :shock:
It is strange to noticed that the only Chess Champions that benefited from playing versus Computer engines where Kramnik and Kasparov the most since he got paid top money first playing versus Deeper Blue and later playing versus Fritz XD3 and Deep Junior,

PS: I wonder why Carlsen does not benefit to accept the challenge of playing Odds versus Engines, definitively he would make more money than playing in any top humans tournament :roll:

lkaufman
Posts: 3538
Joined: Sun Jan 10, 2010 5:15 am
Location: Maryland USA
Contact:

Re: Why there is no interest in Computer with odds Vs Humans match?

Post by lkaufman » Wed Jul 03, 2019 1:05 am

Chessqueen wrote:
Wed Jul 03, 2019 12:29 am
lkaufman wrote:
Tue Jul 02, 2019 10:13 pm
lkaufman wrote:
Tue Jul 02, 2019 8:19 pm
Chessqueen wrote:
Tue Jul 02, 2019 7:31 pm
lkaufman wrote:
Tue Jul 02, 2019 6:54 pm
Chessqueen wrote:
Tue Jul 02, 2019 2:34 pm
lkaufman wrote:
Tue Jul 02, 2019 6:01 am
Chessqueen wrote:
Tue Jul 02, 2019 3:29 am
lkaufman wrote:
Mon Jul 01, 2019 5:03 pm
Nordlandia wrote:
Mon Jul 01, 2019 3:21 pm
Maybe odds matches in Fischer Random is harder for the human side. Humans are used to the standard opening setup. The unfamiliarity of Chess960 may give the engine the upper hand?

Maybe more odds is needed to equalize the chances in Fischer Random.

What is your thought on this one Larry?
Well we already "proved" that this is so, John Meyer won 3 to 1 at regular knight odds vs Komodo, but lost 0-4 at frc knight odds, both 45' + 15". Komodo has no chance giving knight odds in normal chess to strong GMs at rapid (not blitz) time controls, but may have some chance at frc knight odds, depending on the time control and the rating of the GM. But even with frc, knight odds is a lot to give to a GM with enough time to think.
I was also wondering what would be the outcome of pitting Komodo versus The fritz version that played versus Kramnik about 10 years ago Deep Fritz Bahrain with the f3 pawn or the f7 pawn odds in 50 games game in 30 minutes TC. What would your prediction be ?
First of all, Deep Fritz Bahrain played Kramnik in 2002, 17 years ago, not 10, and drew the match on hardware of that time. Presumably on my i7 laptop that same Fritz on 8 threads would be at least a couple hundred elo stronger, over FIDE 3000 surely. But to answer your question, I can't actually do the test on our tester since I don't have that ancient engine and anyway need something that runs in Linux on current hardware, but perhaps I can run some sort of test on something similarly rated on CCRL or CEGT lists. I am quite certain that Komodo would win such a match at f7 handicap (f2 would just be a blowout); there would probably be a fair number of draws, but Komodo would win a lot more than Fritz. Evan at f7 plus second White move handicap Komodo would be favored. Knight odds would be too much though. If someone knows of a suitable engine with a free and reliable Linux version, please advise.

WOW it has been 17 years already! Another program that you can pit is Deep Junior the same that play versus Gary Kasparov back in 2003 which also ended in a Draw. Since DJ and Kasparov were close in rating back in 2003 and it is hard to get a good Grandmaster that wants to play versus Komodo. and Carlsen will never agree since he is too proud to lose versus a computer. We can get Deep Junior with a similar hardware that can hit the same MNPS Millions nodes per second and give Deep Junior a handicap of the f2 pawn, or the f7 pawn odds. If Komodo wins easy then 2 pawns odds which to me it would be very hard for Komodo to even get a draw versus another engine, with a human they have the tendency to get nervous and sometimes blunder versus computer.
https://www.youtube.com/watch?v=WWYaLg21Dss
https://en.chessbase.com/post/kasparov- ... -sacrifice
Okay, I've started the following test. I don't have any engines as weak as those old Fritz and Junior versions on our Linux tester, the weakest one I have on the tester is Arasan 18.0, which is of course not weak at all (2860 CCRL 40/40 single thread). I am reasonably sure that with 1/4 of the time on one thread it is stronger than the two mentioned old engines on whatever number of threads they ran on back then. They are too old to find on the list, but if you extrapolate backward you can see that they should be somewhere below 2700, so this should be correct. I picked a time limit of 5' + 2", which if quadrupled would give games of about the same total time as 30' sudden death. I am running latest Komodo on four threads against Arasan 18.0 on one thread, with Komodo giving the f7 pawn handicap (which in self-testing at one minute + .4" inc. is nearly a 500 elo handicap based on 700 games). Note that this test is very unfair to Komodo (compared to the one you ask for) because Komodo is also only getting 5' + 2", it should really get quadruple this time. Also Komodo is only given four threads, but in an actual match with a top GM we would use maybe 47 threads on a 48 core machine. Also of course Arasan (and Komodo) are running on modern hardware, not 2002/2003 hardware, which makes a huge difference. On the other hand, a human GM would know that he was playing a much stronger entity and would avoid complications and try to simplify, which we can't expect an older engine to do. So maybe these factors balance out. I'll report results when I have them.
I was thinking if you were going to use Linux about Demolito ==> https://github.com/lucasart/Demolito

I am also going to do some test using this version of Araran ==> https://www.arasanchess.org/

Using GCC on Linux, type: make. Otherwise, just have a look at the makefile, nothing fancy there. You should also be able to compile it on any platform (POSIX or Windows), using a C11 capable compiler (eg. GCC or Clang but not MSVC). But you'll have to figure out the exact commands for yourself, depending on your compiler, your libc, and your system.
Why would you want to use the latest Arasan for this rather than an older one like 18.0? Even 18.0 is way stronger than the engines of 2002/2003. Maybe you meant to use Arasan 14?
Anyway, the test above surprised me; Komodo won by an incredible 87 wins to 2 losses, with 33 draws, for a plus 320 elo result despite the nearly 500 elo handicap! So I raised the handicap to pawn and two moves (f7 off, e4 played, WTM) and this made a big difference, but still Komodo won decisively (54 to 13 with 41 draws, plus 144 elo). All this without even raising Contempt from the default value! So now I'm also giving Arasan the same four threads as Komodo, with the pawn and two move handicap; this should be enough for Arasan to close the gap. But then this is a simulation of playing against an entity roughly a class stronger than the Fritz and Junior in question.
With both Komodo and Arasan 18 running on four threads at pawn and two move handicap, the result was close, Komodo winning by 116 wins to 105 losses with 101 draws, +12 elo. Now I'm running the same match with Contempt = 50 to see if that helps.
I guess you need to find out if Komodo still can win with a Knight for a pawn odds that would be interesting to find out :shock:
It is strange to noticed that the only Chess Champions that benefited from playing versus Computer engines where Kramnik and Kasparov the most since he got paid top money first playing versus Deeper Blue and later playing versus Fritz XD3 and Deep Junior,

PS: I wonder why Carlsen does not benefit to accept the challenge of playing Odds versus Engines, definitively he would make more money than playing in any top humans tournament :roll:
Setting Contempt to 50 increased Komodo's winning margin at pawn and two move odds by 22 elo (from 12 to 34 elo, 353 games). I started a new match at two pawn odds (c2 and f2); I predict that Komodo will win by more at this handicap than at pawn and two moves; the difference of two moves with the f pawn gone at the start is even bigger than the value of the "c" pawn, I think. I don't have a way to test knight for pawn in the tester now, I can't add a new type of test to it myself. I could do knight odds, but I would need a weaker opponent than Arasan 18. As for Carlsen, who is going to pay him big money to play a computer? There is only slight financial benefit to a chess software company like Komodo to prove what everyone already knows, that it is way beyond human level. The amount we pay to superstars like Nakamura and MVL for rapid or blitz matches is pretty small compared to what they get for tournaments, and even if we paid Carlsen twice as much that would still be true. The top players understand that they have nothing to lose and conversely that we have little to gain, so the fees are small. Not at all like 15 years ago.
Komodo rules!

User avatar
Nordlandia
Posts: 2353
Joined: Fri Sep 25, 2015 7:38 pm
Location: Sortland, Norway

Re: Why there is no interest in Computer with odds Vs Humans match?

Post by Nordlandia » Wed Jul 03, 2019 2:10 am

The world champions pride is too high to be hurt by losing to Komodo. I don't know if team Komodo have considered a match with him. The assumption that he decline the invitation is clearly there though. Alternatively Maxime Vachier-Lagrave or Caruana is also viable alternative.

Post Reply