Twisted Logic 0.085 released!

[Edsel Apostol]

Hi Everyone,

Here is the version I promised on my last release that will support some UCI configurable options. You can now enable/disable its nullmove pruning, history pruning, late move reductions, iid, futility prunings, quiescent checks, etc. More UCI options will be supported in the later versions.

There are a lot of minor changes compared to the last released version. This version has more aggressive prunings and reductions resulting to lower branching factor. This is so far the strongest on my bullet tests. It scored 2483 on the IQ6.epd test suite while running on a P4 1.7GHz 256Mb computer. The last version scored 2461. The minimum score on this test suite is 1995 and the maximum is 2665 I think.

One more thing, it doesn't support its own opening book anymore. I am not confident with its opening book since it was only hacked from TSCP before. I will try to write my own opening book code with learning on later releases. For tournaments, please just use the tournament book.

I also tried to fix the bug on ponderhit from Polyglot/Winboard where it caused my program to loss a lot of games on time. I have not tested this though but I hope that the bug is fixed.

Any feedback is welcome. Just send me an e-mail and I will try to answer them.

Here is the link to my web page:

http://www.geocities.com/ed_apostol/index.htm

Until next release. Enjoy.

Edsel Apostol

[Tony Thomas]

It isnt working properly here. I tried to play against it, it made one move, on the second move the engine keeps thinking and it loses on time.

Time control 1 minute + 1sec
hash 32
Gui Arena
processor Celeron D 2800Mhz

[Edsel Apostol]

Hi Tony,

I don't know what's the cause of the problem there that you have. I am using Arena 1.1 for testing it. I know that it has a bug in pondering when used using Polyglot/Winboard but it doesn't have a problem here when I used it under Arena. Are you using the latest Arena beta?

Maybe just try to uninstall it and re-install again. By the way, I will be releasing a slightly improved version within a week. Just some bugfixes and a little speed-up. I'm still giving it some final touches as of now before I release it.

Edsel Apostol

[Tony Thomas]

Ponder is turned off for all engines, I am using Arena 1.1. If the install/reinstall doesnt work, I will wait for the next release. I didnt have any problems until 0.45g, then when you switched to UCI, that's when all the problems started, such as engine playing dumb moves when the hash window shows 99%, engine exiting after around 70 moves etc..

[Edsel Apostol]

Hi Everyone,

I have released a new version (085x3x1). This version is the strongest on my blitz testing. There are some changes in the time management and bug fixes from the previous version especially the drawn endgame eval symmetry and in pondering. Please read the readme.txt included in the distribution for more details. I have put there the UCI options supported by the program.

I will also be releasing in a month or two a new engine. Watch out for it.

Special thanks to Chess War, WBEC, CEGT, CCRL and other testers for testing my program. It helps me to become motivated in continuing the development of this program. Enjoy testing.

Until the next release.

http://www.geocities.com/ed_apostol/index.htm

Edsel Apostol

[Daniel Mehrmann]

Hi !

Looks like, after reading your readme.txt, you're using a variant of my history reduce idea, called "(P)eak (H)istory (R)eduction" ?

Best,
Daniel

[Olivier Deville]

Hi Everyone,

I have released a new version (085x3x1). This version is the strongest on my blitz testing. There are some changes in the time management and bug fixes from the previous version especially the drawn endgame eval symmetry and in pondering. Please read the readme.txt included in the distribution for more details. I have put there the UCI options supported by the program.

I will also be releasing in a month or two a new engine. Watch out for it.

Special thanks to Chess War, WBEC, CEGT, CCRL and other testers for testing my program. It helps me to become motivated in continuing the development of this program. Enjoy testing.

Until the next release.

http://www.geocities.com/ed_apostol/index.htm

Edsel Apostol

Hi Edsel

After a very quick test it seems that the pondering problems with winboard+polyglot are gone. Congratulations if this proves to be true after more testing

Olivier

[Tony Thomas]

Hi !

Looks like, after reading your readme.txt, you're using a variant of my history reduce idea, called "(P)eak (H)istory (R)eduction" ?

Best,
Daniel

Probably just the names are similar as your source isnt available.

[Daniel Mehrmann]

Hi !

Looks like, after reading your readme.txt, you're using a variant of my history reduce idea, called "(P)eak (H)istory (R)eduction" ?

Best,
Daniel

Probably just the names are similar as your source isnt available.

Hi Tony !

Sorry, but i guess you have no clue what i'm talking about.

Sometimes its been better post nothing if you don't know the background, history or facts.

In this case my PHR idea was published on CCC->programmer forum....

Best,
Daniel

[Edsel Apostol]

Hi !

Looks like, after reading your readme.txt, you're using a variant of my history reduce idea, called "(P)eak (H)istory (R)eduction" ?

Best,
Daniel

Hi Daniel,

I don't know about your PHR. I may have read about it in the CCC forum but I can't remember it.

I don't do history reductions anymore. I only do history prunings. It is only a very simple implementation and it prunes a lot of subtrees but I thought that it is a dangerous feature. It may weaken my program tactically. The reason why I'm doing it is that my program's NPS is very slow, so in order for it to catch up with the stronger programs I must lower the branching factor. This way it would search deeper but it is prone to overlooking tactics on boring lines. I just make a compromise, hoping that the extra ply or two of search will compensate with the tactical weaknesses.

Would you mind posting here your idea of PHR?

Edsel

[Tony Thomas]

Hi !

Looks like, after reading your readme.txt, you're using a variant of my history reduce idea, called "(P)eak (H)istory (R)eduction" ?

Best,
Daniel

Probably just the names are similar as your source isnt available.

Hi Tony !

Sorry, but i guess you have no clue what i'm talking about.

Sometimes its been better post nothing if you don't know the background, history or facts.

In this case my PHR idea was published on CCC->programmer forum....

Best,
Daniel

My fault, I didnt know that you had published your source code at the CCC. I came out being right since Edsel dont seem to have any idea about your code.