Twisted Logic new version!

[Edsel Apostol]

Hi Edsel..it looks good with this engine!

Blitz 5min Core i7 920 @3.87Ghz 2009 both engines 1core!

1 Cyclone xTreme II 2800 1½½½½101½0 5.5/10
2 T20090922 2800 0½½½½010½1 4.5/10

I hope we can see soon the smp version!!

JP.

I have found a serious bug in the search. I fixed it and have done some search changes and parameter adjustments.

Here's the preliminary results (too few games, 128Mb Hash, single core):

Engine Score To
1: T20090924 7.5/12 00111=====11
2: Toga II 2.0 SE 4.5/12 ····················

12 of 120 games played
Name of the tournament: Arena tournament
Site/ Country: ************
Level: Blitz 1/1
Hardware: Intel(R) Core(TM)2 Quad CPU Q8200 @ 2.33GHz4x 2333 MHz with 2,048 MB Memory
Operating system: Microsoft Windows XP Professional Service Pack 2 (Build 2600)
PGN-File: D:\chess\tests\arena201\Tournaments\Arena tournament.pgn
Website:
E-Mail Address:

The first win by Toga is not legitimate as Twisted Logic is running on a PGO compile.

By the way, Toga just won the game 13.

[jpqy]

Hi Edsel,can we download somewhere T20090924 ,or you wait till the tournament is finised!

Regards,
Jean-Paul.

[Werner]

I have found a serious bug in the search. I fixed it and have done some search changes and parameter adjustments..

do we have to start the tests with new version?

best wishes
Werner

[mclane]

[Edsel Apostol]

Hi Edsel..it looks good with this engine!

Blitz 5min Core i7 920 @3.87Ghz 2009 both engines 1core!

1 Cyclone xTreme II 2800 1½½½½101½0 5.5/10
2 T20090922 2800 0½½½½010½1 4.5/10

I hope we can see soon the smp version!!

JP.

I have found a serious bug in the search. I fixed it and have done some search changes and parameter adjustments.

Here's the preliminary results (too few games, 128Mb Hash, single core):

Engine Score To
1: T20090924 7.5/12 00111=====11
2: Toga II 2.0 SE 4.5/12 ····················

12 of 120 games played
Name of the tournament: Arena tournament
Site/ Country: ************
Level: Blitz 1/1
Hardware: Intel(R) Core(TM)2 Quad CPU Q8200 @ 2.33GHz4x 2333 MHz with 2,048 MB Memory
Operating system: Microsoft Windows XP Professional Service Pack 2 (Build 2600)
PGN-File: D:\chess\tests\arena201\Tournaments\Arena tournament.pgn
Website:
E-Mail Address:

The first win by Toga is not legitimate as Twisted Logic is running on a PGO compile.

By the way, Toga just won the game 13.

After 98 games, Toga leads. Twisted Logic just got lucky in the first few games. It's only scoring 40% now.

I need to tune the parameters though, the current values I've set for some parameters after the search changes are just a handpicked value.

Back to parameter tuning now. I might release a new version this weekend depending on results of further tests here.

[Edsel Apostol]

I have found a serious bug in the search. I fixed it and have done some search changes and parameter adjustments..

do we have to start the tests with new version?

best wishes
Werner

Hi Werner,

Please continue to test the current version for now. I couldn't promise to send /release a new version until I'm quite sure that this one has improved upon the public version and has been tuned.

[Edsel Apostol]

Hi Edsel..it looks good with this engine!

Blitz 5min Core i7 920 @3.87Ghz 2009 both engines 1core!

1 Cyclone xTreme II 2800 1½½½½101½0 5.5/10
2 T20090922 2800 0½½½½010½1 4.5/10

I hope we can see soon the smp version!!

JP.

I have found a serious bug in the search. I fixed it and have done some search changes and parameter adjustments.

Here's the preliminary results (too few games, 128Mb Hash, single core):

Engine Score To
1: T20090924 7.5/12 00111=====11
2: Toga II 2.0 SE 4.5/12 ····················

12 of 120 games played
Name of the tournament: Arena tournament
Site/ Country: ************
Level: Blitz 1/1
Hardware: Intel(R) Core(TM)2 Quad CPU Q8200 @ 2.33GHz4x 2333 MHz with 2,048 MB Memory
Operating system: Microsoft Windows XP Professional Service Pack 2 (Build 2600)
PGN-File: D:\chess\tests\arena201\Tournaments\Arena tournament.pgn
Website:
E-Mail Address:

The first win by Toga is not legitimate as Twisted Logic is running on a PGO compile.

By the way, Toga just won the game 13.

After 98 games, Toga leads. Twisted Logic just got lucky in the first few games. It's only scoring 40% now.

I need to tune the parameters though, the current values I've set for some parameters after the search changes are just a handpicked value.

Back to parameter tuning now. I might release a new version this weekend depending on results of further tests here.

I've just discovered that this version (not 20090922, newer) is doing some illegal moves internally after the changes in the search. It might be the cause of a lot of loses due to incorrect score returned. It's debugging time again.

[mcostalba]

It's debugging time again.

I know what it means !

Congratulations for your new version Edsel, I think TL is one of the engines with the biggest potential out there...

[Denis P. Mendoza]

Hi Edsel..it looks good with this engine!

Blitz 5min Core i7 920 @3.87Ghz 2009 both engines 1core!

1 Cyclone xTreme II 2800 1½½½½101½0 5.5/10
2 T20090922 2800 0½½½½010½1 4.5/10

I hope we can see soon the smp version!!

JP.

I have found a serious bug in the search. I fixed it and have done some search changes and parameter adjustments.

Here's the preliminary results (too few games, 128Mb Hash, single core):

Engine Score To
1: T20090924 7.5/12 00111=====11
2: Toga II 2.0 SE 4.5/12 ····················

12 of 120 games played
Name of the tournament: Arena tournament
Site/ Country: ************
Level: Blitz 1/1
Hardware: Intel(R) Core(TM)2 Quad CPU Q8200 @ 2.33GHz4x 2333 MHz with 2,048 MB Memory
Operating system: Microsoft Windows XP Professional Service Pack 2 (Build 2600)
PGN-File: D:\chess\tests\arena201\Tournaments\Arena tournament.pgn
Website:
E-Mail Address:

The first win by Toga is not legitimate as Twisted Logic is running on a PGO compile.

By the way, Toga just won the game 13.

After 98 games, Toga leads. Twisted Logic just got lucky in the first few games. It's only scoring 40% now.

I need to tune the parameters though, the current values I've set for some parameters after the search changes are just a handpicked value.

Back to parameter tuning now. I might release a new version this weekend depending on results of further tests here.

I've just discovered that this version (not 20090922, newer) is doing some illegal moves internally after the changes in the search. It might be the cause of a lot of loses due to incorrect score returned. It's debugging time again.

Congrats pre for your new TL updates . Yes, a little debugging will make a difference.

And just use no more than /arch:SSE in your 32-bit compiles - everybody happy kabayan!

Denis

[bob]

There's a new Twisted Logic version here:

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

Just take a look at the readme file for more information.

I am not quite sure what to make of this program, but I can tell you one thing for certain. There is a certain level of dishonesty surrounding the thing.

For example, I watched Crafty lose 3 of 4 games to it last weekend. It was kibitzing (TwistedLogicCCT) and claimed to be searching 80K nodes per second. I'd be willing to play _any_ program really searching 80K, when the box I was using was searching 20M nodes per second. And I would not expect to lose 3 of every 4 games. Period. That is a speed difference of 250X. Or with an EBF of 2, about 8 plies. The kibitzes are pure crap.

I don't believe the kibitzing info. Nor the depth info. In short, it sounds so much like the original Rybka fiasco I am not sure what to think. One thing is for absolute certain, however. It is absolutely _not_ what it claims to be, in terms of the info it is kibitzing.

I find it both amusing and disappointing that this kind of stuff continually comes up. And then people get upset when someone uses the "clone/illegal-copy" tag. As if they can't understand why such a claim would be made.

I have suspicions about _several_ current programs that are actively being tested and even participating in CCT-type events. My intent is to start naming names before the next event, to get some minimal level of honesty back into the process. Right now, it is at rock-bottom, IMHO.

While it is hard to get a handle on the "is this a clone" question, clearly when a program kibitzes nonsense, suspicion is more than justified. What's to hide? Why hide it? <sigh>