Looking for a last moment operator for Olithink 4.1.3 for CC

[Dann Corbit]

Geschrieben von: / Posted by: Dann Corbit at 26 January 2004 22:16:28:

There have been some recent changes to Olithink (mostly suggestions from Michel Langeveld due to his work on TSCPGothic).
Due to these changes, Olithink is playing surprisingly well.
Is there someone with a fast machine who would be available to operate Olithink?
Dr. Brausch will be avaialable sometimes to kibits along with the games, but he would not be able to attend all of the games.



my ftp site {remove http:// unless you like error messages}

[Jose Carlos]

Geschrieben von: / Posted by: Jose Carlos at 26 January 2004 22:51:34:
Als Antwort auf: / In reply to: Looking for a last moment operator for Olithink 4.1.3 for CCT-6 geschrieben von: / posted by: Dann Corbit at 26 January 2004 22:16:28:

There have been some recent changes to Olithink (mostly suggestions from Michel Langeveld due to his work on TSCPGothic).
Due to these changes, Olithink is playing surprisingly well.
Is there someone with a fast machine who would be available to operate Olithink?
Dr. Brausch will be avaialable sometimes to kibits along with the games, but he would not be able to attend all of the games.

Hi Dann, I have only one PC to run Averno, but anyway, isn't time to sign in finished since yesterday?
José C.

[Dann Corbit]

Geschrieben von: / Posted by: Dann Corbit at 26 January 2004 22:53:37:
Als Antwort auf: / In reply to: Re: Looking for a last moment operator for Olithink 4.1.3 for CCT-6 geschrieben von: / posted by: Jose Carlos at 26 January 2004 22:51:34:

There have been some recent changes to Olithink (mostly suggestions from Michel Langeveld due to his work on TSCPGothic).
Due to these changes, Olithink is playing surprisingly well.
Is there someone with a fast machine who would be available to operate Olithink?
Dr. Brausch will be avaialable sometimes to kibits along with the games, but he would not be able to attend all of the games.

Hi Dann, I have only one PC to run Averno, but anyway, isn't time to sign in finished since yesterday?

I saw a post that said 8 hours were left. Maybe I did not look carefully at when it was posted.
The interesting thing is that Olithink had about 10 lines changed or added at most and it added perhaps 250 Elo.
I think with a really fast machine it would have even some chances to win the tournament.



my ftp site {remove http:// unless you like error messages}

[Uri Blass]

Geschrieben von: / Posted by: Uri Blass at 26 January 2004 23:06:56:
Als Antwort auf: / In reply to: Re: Looking for a last moment operator for Olithink 4.1.3 for CCT-6 geschrieben von: / posted by: Dann Corbit at 26 January 2004 22:53:37:

There have been some recent changes to Olithink (mostly suggestions from Michel Langeveld due to his work on TSCPGothic).
Due to these changes, Olithink is playing surprisingly well.
Is there someone with a fast machine who would be available to operate Olithink?
Dr. Brausch will be avaialable sometimes to kibits along with the games, but he would not be able to attend all of the games.

Hi Dann, I have only one PC to run Averno, but anyway, isn't time to sign in finished since yesterday?

I saw a post that said 8 hours were left. Maybe I did not look carefully at when it was posted.
The interesting thing is that Olithink had about 10 lines changed or added at most and it added perhaps 250 Elo.
I think with a really fast machine it would have even some chances to win the tournament.

250 elo?
It seems too big improvement for 10 lines change in olithink because olithink is already a good program relative to its size.
Is it based on games against itself or games against other opponents?
I believe that 100 or 150 elo improvement may be possible for it with 10 lines change but not 250 elo improvement.
Olithink has not a lot of knowledge with its small evaluation and by 10 lines you cannot change it.
I guess that there are improvements in the search but you also cannot have complicated search rules with changing only 10 lines.
Uri

[Dann Corbit]

Geschrieben von: / Posted by: Dann Corbit at 26 January 2004 23:09:53:
Als Antwort auf: / In reply to: Re: Looking for a last moment operator for Olithink 4.1.3 for CCT-6 geschrieben von: / posted by: Uri Blass at 26 January 2004 23:06:56:

There have been some recent changes to Olithink (mostly suggestions from Michel Langeveld due to his work on TSCPGothic).
Due to these changes, Olithink is playing surprisingly well.
Is there someone with a fast machine who would be available to operate Olithink?
Dr. Brausch will be avaialable sometimes to kibits along with the games, but he would not be able to attend all of the games.

Hi Dann, I have only one PC to run Averno, but anyway, isn't time to sign in finished since yesterday?

I saw a post that said 8 hours were left. Maybe I did not look carefully at when it was posted.
The interesting thing is that Olithink had about 10 lines changed or added at most and it added perhaps 250 Elo.
I think with a really fast machine it would have even some chances to win the tournament.

250 elo?
It seems too big improvement for 10 lines change in olithink because olithink is already a good program relative to its size.
Is it based on games against itself or games against other opponents?
I believe that 100 or 150 elo improvement may be possible for it with 10 lines change but not 250 elo improvement.
Olithink has not a lot of knowledge with its small evaluation and by 10 lines you cannot change it.
I guess that there are improvements in the search but you also cannot have complicated search rules with changing only 10 lines.
Uri

The improvement was in the search. The hash table was underutilized and IID was missing. The latest 4.12 was already pretty strong. It could be considered as bug fixes for the hash, because deep hash searches were ignored and exact hash results were never used.

I have seen it beat Ruffian and SmarThink. Of course, it was a very tiny sample of games.


my ftp site {remove http:// unless you like error messages}

[Uri Blass]

Geschrieben von: / Posted by: Uri Blass at 26 January 2004 23:21:44:
Als Antwort auf: / In reply to: Re: Looking for a last moment operator for Olithink 4.1.3 for CCT-6 geschrieben von: / posted by: Dann Corbit at 26 January 2004 23:09:53:

There have been some recent changes to Olithink (mostly suggestions from Michel Langeveld due to his work on TSCPGothic).
Due to these changes, Olithink is playing surprisingly well.
Is there someone with a fast machine who would be available to operate Olithink?
Dr. Brausch will be avaialable sometimes to kibits along with the games, but he would not be able to attend all of the games.

Hi Dann, I have only one PC to run Averno, but anyway, isn't time to sign in finished since yesterday?

I saw a post that said 8 hours were left. Maybe I did not look carefully at when it was posted.
The interesting thing is that Olithink had about 10 lines changed or added at most and it added perhaps 250 Elo.
I think with a really fast machine it would have even some chances to win the tournament.

250 elo?
It seems too big improvement for 10 lines change in olithink because olithink is already a good program relative to its size.
Is it based on games against itself or games against other opponents?
I believe that 100 or 150 elo improvement may be possible for it with 10 lines change but not 250 elo improvement.
Olithink has not a lot of knowledge with its small evaluation and by 10 lines you cannot change it.
I guess that there are improvements in the search but you also cannot have complicated search rules with changing only 10 lines.
Uri

The improvement was in the search. The hash table was underutilized and IID was missing. The latest 4.12 was already pretty strong. It could be considered as bug fixes for the hash, because deep hash searches were ignored and exact hash results were never used.

I have seen it beat Ruffian and SmarThink. Of course, it was a very tiny sample of games.

If the change is only in the search then
I am interested to know how much it is faster in test suites so we can use it to get an estimate for the improvement.
In order to make things clear being 4 times faster is not enough for 250 elo improvement.
Uri

[Dann Corbit]

Geschrieben von: / Posted by: Dann Corbit at 26 January 2004 23:28:27:
Als Antwort auf: / In reply to: Re: Looking for a last moment operator for Olithink 4.1.3 for CCT-6 geschrieben von: / posted by: Uri Blass at 26 January 2004 23:21:44:

There have been some recent changes to Olithink (mostly suggestions from Michel Langeveld due to his work on TSCPGothic).
Due to these changes, Olithink is playing surprisingly well.
Is there someone with a fast machine who would be available to operate Olithink?
Dr. Brausch will be avaialable sometimes to kibits along with the games, but he would not be able to attend all of the games.

Hi Dann, I have only one PC to run Averno, but anyway, isn't time to sign in finished since yesterday?

I saw a post that said 8 hours were left. Maybe I did not look carefully at when it was posted.
The interesting thing is that Olithink had about 10 lines changed or added at most and it added perhaps 250 Elo.
I think with a really fast machine it would have even some chances to win the tournament.

250 elo?
It seems too big improvement for 10 lines change in olithink because olithink is already a good program relative to its size.
Is it based on games against itself or games against other opponents?
I believe that 100 or 150 elo improvement may be possible for it with 10 lines change but not 250 elo improvement.
Olithink has not a lot of knowledge with its small evaluation and by 10 lines you cannot change it.
I guess that there are improvements in the search but you also cannot have complicated search rules with changing only 10 lines.
Uri

The improvement was in the search. The hash table was underutilized and IID was missing. The latest 4.12 was already pretty strong. It could be considered as bug fixes for the hash, because deep hash searches were ignored and exact hash results were never used.

I have seen it beat Ruffian and SmarThink. Of course, it was a very tiny sample of games.

If the change is only in the search then
I am interested to know how much it is faster in test suites so we can use it to get an estimate for the improvement.
In order to make things clear being 4 times faster is not enough for 250 elo improvement.
Uri

The binary is here:
ftp://cap.connx.com/chess-engines/new-a ... ink412.exe
The source code is here:
ftp://cap.connx.com/chess-engines/new-a ... think412.c
The source code I put there has problems (it sometimes makes bad resignation results).
But it should be fine to test against EPD sets using Bruce Moreland's tool.


my ftp site {remove http:// unless you like error messages}

[Thomas Mayer]

Geschrieben von: / Posted by: Thomas Mayer at 27 January 2004 00:32:07:
Als Antwort auf: / In reply to: Re: Looking for a last moment operator for Olithink 4.1.3 for CCT-6 geschrieben von: / posted by: Dann Corbit at 26 January 2004 23:28:27:

Hi Dann,
at the moment I let it run through WACNEW - but it seems not to make such a big difference - it is usually about 10-20% faster then the older version, in some positions it is slower... 250 ELO ?
Greets, Thomas
P.S.: Anyway, if you do not find anyone I might operate it, I think I have an Athlon XP1800+ here... (sorry, the XP2700+ will be user for Quark)

[Dann Corbit]

Geschrieben von: / Posted by: Dann Corbit at 27 January 2004 01:00:48:
Als Antwort auf: / In reply to: Re: Looking for a last moment operator for Olithink 4.1.3 for CCT-6 geschrieben von: / posted by: Thomas Mayer at 27 January 2004 00:32:07:

Hi Dann,
at the moment I let it run through WACNEW - but it seems not to make such a big difference - it is usually about 10-20% faster then the older version, in some positions it is slower... 250 ELO ?
Greets, Thomas
P.S.: Anyway, if you do not find anyone I might operate it, I think I have an Athlon XP1800+ here... (sorry, the XP2700+ will be user for Quark)

My estimate may be a fluke. I ran a bunch of test games against strong programs and Olithink did well. Even the games where it lost, it usually had a lead.
I don't know if it plays tactics well or not.


my ftp site {remove http:// unless you like error messages}

[Lance Perkins]

Geschrieben von: / Posted by: Lance Perkins at 27 January 2004 01:43:07:
Als Antwort auf: / In reply to: Re: Looking for a last moment operator for Olithink 4.1.3 for CCT-6 geschrieben von: / posted by: Thomas Mayer at 27 January 2004 00:32:07:

If the bug fix is with hash storage/look-up, then the improvement might not show-up in single-search position tests. In actual games, I can immagine that a lot savings can happen between a series of searches as most previously-searched trees can be skipped if the hash is efficient enough.
Cheers...

Hi Dann,
at the moment I let it run through WACNEW - but it seems not to make such a big difference - it is usually about 10-20% faster then the older version, in some positions it is slower... 250 ELO ?
Greets, Thomas
P.S.: Anyway, if you do not find anyone I might operate it, I think I have an Athlon XP1800+ here... (sorry, the XP2700+ will be user for Quark)

[Jose Carlos]

Geschrieben von: / Posted by: Jose Carlos at 27 January 2004 08:47:19:
Als Antwort auf: / In reply to: Re: Looking for a last moment operator for Olithink 4.1.3 for CCT-6 geschrieben von: / posted by: Dann Corbit at 26 January 2004 23:28:27:

There have been some recent changes to Olithink (mostly suggestions from Michel Langeveld due to his work on TSCPGothic).
Due to these changes, Olithink is playing surprisingly well.
Is there someone with a fast machine who would be available to operate Olithink?
Dr. Brausch will be avaialable sometimes to kibits along with the games, but he would not be able to attend all of the games.

Hi Dann, I have only one PC to run Averno, but anyway, isn't time to sign in finished since yesterday?

I saw a post that said 8 hours were left. Maybe I did not look carefully at when it was posted.
The interesting thing is that Olithink had about 10 lines changed or added at most and it added perhaps 250 Elo.
I think with a really fast machine it would have even some chances to win the tournament.

250 elo?
It seems too big improvement for 10 lines change in olithink because olithink is already a good program relative to its size.
Is it based on games against itself or games against other opponents?
I believe that 100 or 150 elo improvement may be possible for it with 10 lines change but not 250 elo improvement.
Olithink has not a lot of knowledge with its small evaluation and by 10 lines you cannot change it.
I guess that there are improvements in the search but you also cannot have complicated search rules with changing only 10 lines.
Uri

The improvement was in the search. The hash table was underutilized and IID was missing. The latest 4.12 was already pretty strong. It could be considered as bug fixes for the hash, because deep hash searches were ignored and exact hash results were never used.

I have seen it beat Ruffian and SmarThink. Of course, it was a very tiny sample of games.

If the change is only in the search then
I am interested to know how much it is faster in test suites so we can use it to get an estimate for the improvement.
In order to make things clear being 4 times faster is not enough for 250 elo improvement.
Uri

The binary is here:
ftp://cap.connx.com/chess-engines/new-a ... ink412.exe
The source code is here:
ftp://cap.connx.com/chess-engines/new-a ... think412.c
The source code I put there has problems (it sometimes makes bad resignation results).
But it should be fine to test against EPD sets using Bruce Moreland's tool.

Hi Dann, I cannot access your ftp for some reason. Can you (or anyone) please send me the source file by email (smyslov@hotmail.com)?
Thanks,
José C.

[Uri Blass]

Geschrieben von: / Posted by: Uri Blass at 27 January 2004 10:06:47:
Als Antwort auf: / In reply to: Re: Looking for a last moment operator for Olithink 4.1.3 for CCT-6 geschrieben von: / posted by: Lance Perkins at 27 January 2004 01:43:07:

If the bug fix is with hash storage/look-up, then the improvement might not show-up in single-search position tests. In actual games, I can immagine that a lot savings can happen between a series of searches as most previously-searched trees can be skipped if the hash is efficient enough.
Cheers...

Still it does not seem something close to 250 elo improvement even if you combine it with 10-20% speed improvement.
I do not think that a lot of saving can happen between series of search
because even if you predict the opponent move correctly you still most of the tree that you should search later is different.
It may give you something equivalent to another 20% speed improvement but it is still less than 2 times faster.
The speed improvement can be bigger in the endgame but the middle game is the most important positions.
I will be surprised if it is more than 100 elo better than previous version based on the data that I read.
Uri

[Thomas Mayer]

Geschrieben von: / Posted by: Thomas Mayer at 27 January 2004 16:32:38:
Als Antwort auf: / In reply to: Re: Looking for a last moment operator for Olithink 4.1.3 for CCT-6 geschrieben von: / posted by: Lance Perkins at 27 January 2004 01:43:07:

Hi Lance,

If the bug fix is with hash storage/look-up, then the improvement might not
show-up in single-search position tests. In actual games, I can immagine
that a lot savings can happen between a series of searches as most
previously-searched trees can be skipped if the hash is efficient enough.

that's correct - but hashing improves an engine usually about 100 Elo... IID itself is maybe worth another 20-30 - it depends on the engine... maybe the way he has implemented hashing so far did even harm the engine - e.g. when you only calculate hash, store positions and and look in the table but get no results this would slow you down about 30-50%... so another 30-50 Elo... so we might get near to 200... so Dann could be right, of course...
Greets, Thomas

[Dann Corbit]

Geschrieben von: / Posted by: Dann Corbit at 27 January 2004 20:04:50:
Als Antwort auf: / In reply to: Re: Looking for a last moment operator for Olithink 4.1.3 for CCT-6 geschrieben von: / posted by: Thomas Mayer at 27 January 2004 16:32:38:

Hi Lance,

If the bug fix is with hash storage/look-up, then the improvement might not
show-up in single-search position tests. In actual games, I can immagine
that a lot savings can happen between a series of searches as most
previously-searched trees can be skipped if the hash is efficient enough.

that's correct - but hashing improves an engine usually about 100 Elo... IID itself is maybe worth another 20-30 - it depends on the engine... maybe the way he has implemented hashing so far did even harm the engine - e.g. when you only calculate hash, store positions and and look in the table but get no results this would slow you down about 30-50%... so another 30-50 Elo... so we might get near to 200... so Dann could be right, of course...
Greets, Thomas

It was just a wild guess, not a measurement.
But I saw it paste some top-of-the-line engines in the chops.
It wasn't doing that before. That's all.


my ftp site {remove http:// unless you like error messages}

[Dann Corbit]

Geschrieben von: / Posted by: Dann Corbit at 27 January 2004 20:17:01:
Als Antwort auf: / In reply to: Re: Looking for a last moment operator for Olithink 4.1.3 for CCT-6 geschrieben von: / posted by: Uri Blass at 27 January 2004 10:06:47:

If the bug fix is with hash storage/look-up, then the improvement might not show-up in single-search position tests. In actual games, I can immagine that a lot savings can happen between a series of searches as most previously-searched trees can be skipped if the hash is efficient enough.
Cheers...

Still it does not seem something close to 250 elo improvement even if you combine it with 10-20% speed improvement.
I do not think that a lot of saving can happen between series of search
because even if you predict the opponent move correctly you still most of the tree that you should search later is different.
It may give you something equivalent to another 20% speed improvement but it is still less than 2 times faster.
The speed improvement can be bigger in the endgame but the middle game is the most important positions.
I will be surprised if it is more than 100 elo better than previous version based on the data that I read.

I don't know how much better it is. 250 was a wild guess. But it definitely is much stronger.
How much realistically will require measurement.


my ftp site {remove http:// unless you like error messages}

[Dr. Oliver Brausch]

Geschrieben von: / Posted by: Dr. Oliver Brausch at 03 February 2004 23:38:11:
Als Antwort auf: / In reply to: Re: Looking for a last moment operator for Olithink 4.1.3 for CCT-6 geschrieben von: / posted by: Dann Corbit at 26 January 2004 23:28:27:

Unfortunately we could not use the improvements on CCT6 because
of stability problems and so OliThink only got a place in the
midfield.

The binary is here:
ftp://cap.connx.com/chess-engines/new-a ... ink412.exe
The source code is here:
ftp://cap.connx.com/chess-engines/new-a ... think412.c
The source code I put there has problems (it sometimes makes bad resignation results).
But it should be fine to test against EPD sets using Bruce Moreland's tool.

Could you please make the files again available. especially
the exe file.
Thank you.
Oliver

[Dann Corbit]

Geschrieben von: / Posted by: Dann Corbit at 04 February 2004 00:13:41:
Als Antwort auf: / In reply to: Re: Looking for a last moment operator for Olithink 4.1.3 for CCT-6 geschrieben von: / posted by: Dr. Oliver Brausch at 03 February 2004 23:38:11:

Unfortunately we could not use the improvements on CCT6 because
of stability problems and so OliThink only got a place in the
midfield.

The binary is here:
ftp://cap.connx.com/chess-engines/new-a ... ink412.exe
The source code is here:
ftp://cap.connx.com/chess-engines/new-a ... think412.c
The source code I put there has problems (it sometimes makes bad resignation results).
But it should be fine to test against EPD sets using Bruce Moreland's tool.

Could you please make the files again available. especially
the exe file.
Thank you.

Try now.


my ftp site {remove http:// unless you like error messages}