lkaufman wrote: ↑Fri Oct 16, 2020 11:35 pm
Have any rated human players played games against Sargon on a modern computer to provide any data on the human rating it might earn today? If so at what time control?
OK, as best as I can figure out from all the posts, the elo rating of Sargon 1 in 1978 was about 1400, and its estimated elo vs. humans on an i7 now is around 1700. The hardware speedup is either 50 to 1 or 6000+ to 1 depending on which post you believe; if it is 50 to 1 this all makes sense, since 1700 elo is not too far out of line with 1254 CCRL Blitz if you contract rating differences by a third or so. I get 52knps on my very fast i7 which means about 40k on a typical one, so if there is some evidence that the original hardware got around 800 nps then everything fits. The 6000 to 1 figure is hard to credit, especially since it referred to using some old hardware, not an i7, so an i7 might be 10000 to 1 which means the original Sargon got 4 nodes per second?? That seems impossibly low.
I note that the 1700 elo mentioned was based on Shredder and SF versions set to 1600, but does anyone know whether those ratings were themselves based on human games or on games with CCRL rated engines?
You can estimate the speed up. The level times 1ply meaning level 1, 2 to 2..... level 6 for 6 ply took an average of 4 hours. Sargon did not have time controls only levels. Take a few positions and see how long your Sargon takes to search 6 ply. And compare that to 4 hours.This will give you a speed ratio compared to a TRS-80 computer.
I get 63knps using my brother very fast i9 computer, I forgot to mentioned if it makes any difference it is the intel i9 9900K, and the final score after 10 games I won 5 games to Sargon 3 games and two draw
What I do NOT understand neither my trainer online is why did Sargon decided to repeat the three fold and accept a draw instead of playing 43.Qxh6!
Chessqueen wrote: ↑Sat Jan 02, 2021 8:40 pm
What I do NOT understand neither my trainer online is why did Sargon decided to repeat the three fold and accept a draw instead of playing 43.Qxh6!
There is plenty of discussion of this up-thread, including a recent commitment from me to fix this repetition problem.
Bill Forster wrote: ↑Thu Dec 31, 2020 11:21 pm
Thanks to the three posters expressing interest. I will go ahead and work on a new V1.01 release with the goal of fixing the repetition problem whilst respecting the provenance of the program, i.e. without making any changes to the core Sargon 1978 assembly code. Improving the core Sargon 1978 code is an interesting possible project, but that's not what the current project is all about.
I will also add a new "timing" set to the suite of regression tests, to try to get a more satisfactory resolution of the "speed up factor" issue.
Sargon does this (concedes unnecessary repetition draws), basically because it's dumber than a cockroach. It doesn't understand that a bird in the hand is worth two in a bush, it is just as happy to know it retains the option of playing Qxh6 as it would be actually playing Qxh6 and pocketing that pawn. There is no code to detect that this decision condemns it to falling into an infinite loop - the original Z80 program would have happily continued repeating until the end of time because there was no three times repetition draw cutoff in the user interface.
I will (try to?) fix this by adding another layer over the top, when Sargon repeats like this with 43.Qg6 the new layer will reject the move (internally) and say, "No try again, and this time the move 43.Qg6 is not allowed". Only if Sargon's next best move leaves it worse (according to its own eval) will the updated version revert to 43.Qg6. Pretty sure that won't happen here and it will play 43.Qxh6 instead.
Incidentally if you read the old chess computing literature, this sort of thing was a common problem for early chess programs, see for example the famously funny conclusion to Coko v Genie in 1971.
Chessqueen wrote: ↑Sat Jan 02, 2021 8:40 pm
What I do NOT understand neither my trainer online is why did Sargon decided to repeat the three fold and accept a draw instead of playing 43.Qxh6!
There is plenty of discussion of this up-thread, including a recent commitment from me to fix this repetition problem.
Bill Forster wrote: ↑Thu Dec 31, 2020 11:21 pm
Thanks to the three posters expressing interest. I will go ahead and work on a new V1.01 release with the goal of fixing the repetition problem whilst respecting the provenance of the program, i.e. without making any changes to the core Sargon 1978 assembly code. Improving the core Sargon 1978 code is an interesting possible project, but that's not what the current project is all about.
I will also add a new "timing" set to the suite of regression tests, to try to get a more satisfactory resolution of the "speed up factor" issue.
Sargon does this (concedes unnecessary repetition draws), basically because it's dumber than a cockroach. It doesn't understand that a bird in the hand is worth two in a bush, it is just as happy to know it retains the option of playing Qxh6 as it would be actually playing Qxh6 and pocketing that pawn. There is no code to detect that this decision condemns it to falling into an infinite loop - the original Z80 program would have happily continued repeating until the end of time because there was no three times repetition draw cutoff in the user interface.
I will (try to?) fix this by adding another layer over the top, when Sargon repeats like this with 43.Qg6 the new layer will reject the move (internally) and say, "No try again, and this time the move 43.Qg6 is not allowed". Only if Sargon's next best move leaves it worse (according to its own eval) will the updated version revert to 43.Qg6. Pretty sure that won't happen here and it will play 43.Qxh6 instead.
Incidentally if you read the old chess computing literature, this sort of thing was a common problem for early chess programs, see for example the famously funny conclusion to Coko v Genie in 1971.
Since you have been given permission to work or fix Sargon Try to give it NN or NNU,E and fix the evaluation by giving it an extra 100+ ratings points
Is Socrate stronger than Sargon, and where can I get it, ask Larry Kaufman if you can work on it too?
Re: Sargon 1978 UCI Available
Posted: Sun Jan 03, 2021 1:04 pm
by mephisto
No for goodness sake leave Bill to just try and fix the repetition problem and leave the original playing strength as it is. That is what this is all about.
Re: Sargon 1978 UCI Available
Posted: Sun Jan 03, 2021 2:35 pm
by Paloma
+1
Re: Sargon 1978 UCI Available
Posted: Sun Jan 03, 2021 9:17 pm
by Chessqueen
mephisto wrote: ↑Sun Jan 03, 2021 1:04 pm
No for goodness sake leave Bill to just try and fix the repetition problem and leave the original playing strength as it is. That is what this is all about.
You are correct, this old goodie can beat about 70% of the members of this Forum Talkchess, and by just correcting the repetition bug in its algorithm should give it at least another 50 elo, specially when it is winning. But it would be nice if later ob Bill can add the NNUE as a separate version just to compare the Elo gain, which should be at least 150 -1
Chessqueen wrote: ↑Sat Jan 02, 2021 8:40 pm
What I do NOT understand neither my trainer online is why did Sargon decided to repeat the three fold and accept a draw instead of playing 43.Qxh6!
There is plenty of discussion of this up-thread, including a recent commitment from me to fix this repetition problem.
Bill Forster wrote: ↑Thu Dec 31, 2020 11:21 pm
Thanks to the three posters expressing interest. I will go ahead and work on a new V1.01 release with the goal of fixing the repetition problem whilst respecting the provenance of the program, i.e. without making any changes to the core Sargon 1978 assembly code. Improving the core Sargon 1978 code is an interesting possible project, but that's not what the current project is all about.
I will also add a new "timing" set to the suite of regression tests, to try to get a more satisfactory resolution of the "speed up factor" issue.
Sargon does this (concedes unnecessary repetition draws), basically because it's dumber than a cockroach. It doesn't understand that a bird in the hand is worth two in a bush, it is just as happy to know it retains the option of playing Qxh6 as it would be actually playing Qxh6 and pocketing that pawn. There is no code to detect that this decision condemns it to falling into an infinite loop - the original Z80 program would have happily continued repeating until the end of time because there was no three times repetition draw cutoff in the user interface.
I will (try to?) fix this by adding another layer over the top, when Sargon repeats like this with 43.Qg6 the new layer will reject the move (internally) and say, "No try again, and this time the move 43.Qg6 is not allowed". Only if Sargon's next best move leaves it worse (according to its own eval) will the updated version revert to 43.Qg6. Pretty sure that won't happen here and it will play 43.Qxh6 instead.
Incidentally if you read the old chess computing literature, this sort of thing was a common problem for early chess programs, see for example the famously funny conclusion to Coko v Genie in 1971.
In the meantime until you fix the repetition bug algorithm, I decided to have this match against Snowy rate 1998, but the first game was a hard fought, but I did manage to take it out of the Opening by playing the 1.h3 Clemenz Opening which not too many engines know how to play it best. I did messed up in the Opening by NOT playing 3. Nc3 which is better than 3. Nf3 to keep control of the center better, I also believe that in move 32 I should have taken with my Bishop to keep his pawns from advancing, but I realized it a few moves later when Snowy started to push its pawns chain. Now I will send this game to my trainer so he can analyze it and tell me where I went wrong, that is why I am paying him, so he can teach me from games that I play against either opponents that he pit me against from India even if they are just 7 to 8 years old, future GM, or chess Engines that are close to what he believe that is my current ratings, after 16 months of training with him online.
Chessqueen wrote: ↑Sat Jan 02, 2021 8:40 pm
What I do NOT understand neither my trainer online is why did Sargon decided to repeat the three fold and accept a draw instead of playing 43.Qxh6!
There is plenty of discussion of this up-thread, including a recent commitment from me to fix this repetition problem.
Bill Forster wrote: ↑Thu Dec 31, 2020 11:21 pm
Thanks to the three posters expressing interest. I will go ahead and work on a new V1.01 release with the goal of fixing the repetition problem whilst respecting the provenance of the program, i.e. without making any changes to the core Sargon 1978 assembly code. Improving the core Sargon 1978 code is an interesting possible project, but that's not what the current project is all about.
I will also add a new "timing" set to the suite of regression tests, to try to get a more satisfactory resolution of the "speed up factor" issue.
Sargon does this (concedes unnecessary repetition draws), basically because it's dumber than a cockroach. It doesn't understand that a bird in the hand is worth two in a bush, it is just as happy to know it retains the option of playing Qxh6 as it would be actually playing Qxh6 and pocketing that pawn. There is no code to detect that this decision condemns it to falling into an infinite loop - the original Z80 program would have happily continued repeating until the end of time because there was no three times repetition draw cutoff in the user interface.
I will (try to?) fix this by adding another layer over the top, when Sargon repeats like this with 43.Qg6 the new layer will reject the move (internally) and say, "No try again, and this time the move 43.Qg6 is not allowed". Only if Sargon's next best move leaves it worse (according to its own eval) will the updated version revert to 43.Qg6. Pretty sure that won't happen here and it will play 43.Qxh6 instead.
Incidentally if you read the old chess computing literature, this sort of thing was a common problem for early chess programs, see for example the famously funny conclusion to Coko v Genie in 1971.
In the meantime until you fix the repetition bug algorithm, I decided to have this match against Snowy rate 1998, but the first game was a hard fought, but I did manage to take it out of the Opening by playing the 1.h3 Clemenz Opening which not too many engines know how to play it best. I did messed up in the Opening by NOT playing 3. Nc3 which is better than 3. Nf3 to keep control of the center better, I also believe that in move 32 I should have taken with my Bishop to keep his pawns from advancing, but I realized it a few moves later when Snowy started to push its pawns chain. Now I will send this game to my trainer so he can analyze it and tell me where I went wrong, that is why I am paying him, so he can teach me from games that I play against either opponents that he pit me against from India even if they are just 7 to 8 years old, future GM, or chess Engines that are close to what he believe that is my current ratings, after 16 months of training with him online.
I would like to request a change from Tarrasch GUI, and that is to remove the Houdini engine since it is a clone and to replace it with Sargon since it is an improvement of the original specially after fixing the 3 repetition bug in the algorithm.
I would like to request a change from Tarrasch GUI, and that is to remove the Houdini engine since it is a clone and to replace it with Sargon since it is an improvement of the original specially after fixing the 3 repetition bug in the algorithm.[/pgn]
Chessqueen wrote: ↑Sat Jan 02, 2021 8:40 pm
What I do NOT understand neither my trainer online is why did Sargon decided to repeat the three fold and accept a draw instead of playing 43.Qxh6!
There is plenty of discussion of this up-thread, including a recent commitment from me to fix this repetition problem.
Bill Forster wrote: ↑Thu Dec 31, 2020 11:21 pm
Thanks to the three posters expressing interest. I will go ahead and work on a new V1.01 release with the goal of fixing the repetition problem whilst respecting the provenance of the program, i.e. without making any changes to the core Sargon 1978 assembly code. Improving the core Sargon 1978 code is an interesting possible project, but that's not what the current project is all about.
I will also add a new "timing" set to the suite of regression tests, to try to get a more satisfactory resolution of the "speed up factor" issue.
Sargon does this (concedes unnecessary repetition draws), basically because it's dumber than a cockroach. It doesn't understand that a bird in the hand is worth two in a bush, it is just as happy to know it retains the option of playing Qxh6 as it would be actually playing Qxh6 and pocketing that pawn. There is no code to detect that this decision condemns it to falling into an infinite loop - the original Z80 program would have happily continued repeating until the end of time because there was no three times repetition draw cutoff in the user interface.
I will (try to?) fix this by adding another layer over the top, when Sargon repeats like this with 43.Qg6 the new layer will reject the move (internally) and say, "No try again, and this time the move 43.Qg6 is not allowed". Only if Sargon's next best move leaves it worse (according to its own eval) will the updated version revert to 43.Qg6. Pretty sure that won't happen here and it will play 43.Qxh6 instead.
Incidentally if you read the old chess computing literature, this sort of thing was a common problem for early chess programs, see for example the famously funny conclusion to Coko v Genie in 1971.
Bill how much longer will it take you to correct the repetition bug in Sargon so I can play another 10 games, since Sargon is so close to my rating ?
Re: Sargon 1978 UCI Available
Posted: Sat Jan 09, 2021 9:14 pm
by Chessqueen
Here is another example where I wanted to test Sargon to see if it knew the opposition principle, it took the bait of my pawn offering and it ended in a draw