To Ed: Hannibal 1.0a ... Twisted Logic!

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

Moderator: Ras

Frank Quisinsky
Posts: 7105
Joined: Wed Nov 18, 2009 7:16 pm
Location: Gutweiler, Germany
Full name: Frank Quisinsky

To Ed: Hannibal 1.0a ... Twisted Logic!

Post by Frank Quisinsky »

Hello Ed,

in SWCR-32, Twisted Logic 20100131x played so far 1.600 games (w32) and around 900 games (x64).

Both versions have a little problem. Games are played up to mate (without resign factor) with ponder "on" and 40 moves in 10 minutes with Q9550 quad core systems. In around 100 games Twisted Logic lost on time in clear remis end positions. I don't have collect the games, but can send different examples if you like.

Do you know about these problem?

The games I edit all to "remis" ... for the rating list calculations. Is a bit work to check that.

Now I read in the morning that Hannibal 1.0a is availabe. Thanks again for Hannibal. Like Twisted Logic (different playing style to the others in the list). Very fine engine ...

I haven't checked the problem with Hannibal so far. Hannibal will start end of month in SWCR.

Have thanks again to your partner too ...
Good luck with the further development.

Best
Frank
BubbaTough
Posts: 1154
Joined: Fri Jun 23, 2006 5:18 am

Re: To Ed: Hannibal 1.0a ... Twisted Logic!

Post by BubbaTough »

Thanks for your report. There have been some changes to time management, but I don't think many are relevant to time controls like 40/10, so its quite possible Hannibal the same issues. I have noticed when looking at a couple OpenWar games that Hannibal sometimes does very fast bad moves at the end of this type of time control (though I have not seen time losses yet); also there was a comment from CEGT testing regarding some similar behavior on a private version they were kind enough to look at I think.

I plan to spend some time testing and tuning for this time control for the next release. It seems most testers use this form of time control, so I guess its a little silly for us not to give it any attention.

-Sam
Frank Quisinsky
Posts: 7105
Joined: Wed Nov 18, 2009 7:16 pm
Location: Gutweiler, Germany
Full name: Frank Quisinsky

Re: To Ed: Hannibal 1.0a ... Twisted Logic!

Post by Frank Quisinsky »

Hello Sam,

thanks for answering in detail (very interesting).

OK, end of the month ... Hannibal 1.0a will start ... I will collect such games and send by mail. Possible that the new version will not produce this problem after your comments.

More interesting:
Twisted Logic is in my statistics the number 5 of 21 in aggressiveness :-)
I saw very nice ideas after the opening book moves, very good what you and Ed do here.

Number 1: Spark, Number 2. Hiarcs, Number 3. Stockfish, Number 4. Junior, Number 5. Twisted Logic, Number 6. Junior, Number 7, Thinker ..., Number 12. Rybka ... like that (have two different Junior's in this list).

:-)

Perhaps later I will public all the statistic I generated from SWCR games.

Best
Frank
Last edited by Frank Quisinsky on Tue Jul 20, 2010 5:43 pm, edited 1 time in total.
BubbaTough
Posts: 1154
Joined: Fri Jun 23, 2006 5:18 am

Re: To Ed: Hannibal 1.0a ... Twisted Logic!

Post by BubbaTough »

Frank Quisinsky wrote: More interesting:
Twisted Logic is in my statistics the number 5 of 21 in aggressiveness :-)
I saw very nice ideas after the opening book moves, very good what you and Ed do here.
Just to clarify, I had nothing to do with the creation of TwistedLogic. Edsel gets full credit for its aggressiveness (and everything else about this wonderful program). It will be interesting to see how Hannibal rates. I did a number of changes related to king safety which in theory could effect this, but just from watching the games I suspect Hannibal still rates pretty high on the aggression scale.

-Sam
Edsel Apostol
Posts: 803
Joined: Mon Jul 17, 2006 5:53 am
Full name: Edsel Apostol

Re: To Ed: Hannibal 1.0a ... Twisted Logic!

Post by Edsel Apostol »

Frank Quisinsky wrote:Hello Ed,

in SWCR-32, Twisted Logic 20100131x played so far 1.600 games (w32) and around 900 games (x64).

Both versions have a little problem. Games are played up to mate (without resign factor) with ponder "on" and 40 moves in 10 minutes with Q9550 quad core systems. In around 100 games Twisted Logic lost on time in clear remis end positions. I don't have collect the games, but can send different examples if you like.

Do you know about these problem?

The games I edit all to "remis" ... for the rating list calculations. Is a bit work to check that.

Now I read in the morning that Hannibal 1.0a is availabe. Thanks again for Hannibal. Like Twisted Logic (different playing style to the others in the list). Very fine engine ...

I haven't checked the problem with Hannibal so far. Hannibal will start end of month in SWCR.

Have thanks again to your partner too ...
Good luck with the further development.

Best
Frank
Hi Frank,

Thanks for the feedback. Twisted Logic has a default time buffer of one second so it should not easily lose on time. Does TL send some output like NPS when it lost on time or it is simply idle and doesn't respond? This might be a problem with Hannibal also.

We usually test on fisher time controls and I think TL performs better on that time control that the fixed x moves in y minutes time control.

Do you have some games or maybe positions on which it lost on time?
Frank Quisinsky
Posts: 7105
Joined: Wed Nov 18, 2009 7:16 pm
Location: Gutweiler, Germany
Full name: Frank Quisinsky

Re: To Ed: Hannibal 1.0a ... Twisted Logic!

Post by Frank Quisinsky »

Hi Ed,

I will search different examples in the very late evening.
No problem ...

Later ...

Best
Frank
Frank Quisinsky
Posts: 7105
Joined: Wed Nov 18, 2009 7:16 pm
Location: Gutweiler, Germany
Full name: Frank Quisinsky

Re: Bug report only ... for the programmer!

Post by Frank Quisinsky »

Hi Ed,

I collect ten PGN games from the still running x64 tournament. But the w32 version have the same problem.

Download:
http://www.amateurschach.de/download/tw ... 00131x.zip

BTW: After my list 92 games from the still played 2.500 games (1.600 with w32, around 910 with the x64) I had. But I delete after a new test the game comments, so I have only this ten for the moment.

31x x64
61x w32

All games I edit to remis, because all in clear remis positions, have a look in the games.

No other technical programs I saw, this one only!

Best
Frank
Frank Quisinsky
Posts: 7105
Joined: Wed Nov 18, 2009 7:16 pm
Location: Gutweiler, Germany
Full name: Frank Quisinsky

Re: To Ed: Hannibal 1.0a ... Twisted Logic!

Post by Frank Quisinsky »

Ed,

to your question:

Unfortunately, if I look if a Twisted game is still running I saw nothing. The "Lost on Time" games comes if I don't looked live :-)

Waiting of such a situation a long time to write it by mail.

What I can say it ...
The engine isn't in RAM, next game under Shredder Classic 4.0 GUI started absolutley normaly. Twisted made here no problems, all the times I used your engine.

Perhaps ...
End of max. ply in remis position and engine lost on time. Such a problem has the actual Komodo 1.2 version too. Could be the reason, I don't know.

Best
Frank
Frank Quisinsky
Posts: 7105
Joined: Wed Nov 18, 2009 7:16 pm
Location: Gutweiler, Germany
Full name: Frank Quisinsky

Re: Bug report only ... for the programmer!

Post by Frank Quisinsky »

No other technical problems (not programs) :-)
Little mistake, sorry!

Best
Frank
Edsel Apostol
Posts: 803
Joined: Mon Jul 17, 2006 5:53 am
Full name: Edsel Apostol

Re: To Ed: Hannibal 1.0a ... Twisted Logic!

Post by Edsel Apostol »

Frank Quisinsky wrote:Ed,

to your question:

Unfortunately, if I look if a Twisted game is still running I saw nothing. The "Lost on Time" games comes if I don't looked live :-)

Waiting of such a situation a long time to write it by mail.

What I can say it ...
The engine isn't in RAM, next game under Shredder Classic 4.0 GUI started absolutley normaly. Twisted made here no problems, all the times I used your engine.

Perhaps ...
End of max. ply in remis position and engine lost on time. Such a problem has the actual Komodo 1.2 version too. Could be the reason, I don't know.

Best
Frank
Thanks for the games Frank. May I know what GUI you are using? Is this played on a "ponder on" tournament?

I have a suspicion that this might be due to the End of max ply as you've mentioned.