Results of Pro Deo [Time Control] parameter

[Albert Silver]

Geschrieben von:/Posted by: Albert Silver at 15 September 2004 04:57:19:

I ran a 3rd Nunn 2 match against List 5.12 in order to get more results on the
[Time Control = 125] compared to Pro Deo's default settings.
Conditions: Athlon 2400XP+, Nunn 2 set, all learning off, ponder off, 5min + 3s:
List 5.12
---------
Default (without TC=125) : 22 - 18 (+4)
With [Time Control = 125]: 21.5 - 18.5 (+3)
Ruffian 1.05
------------
Default (without TC=125) : 16 - 24 (-8)
With [Time Control = 125]: 20.5 - 19.5 (+1)
Aristarch 4.5
-------------
Default (without TC=125) : 21.5 - 18.5 (+3)
With [Time Control = 125]: 22.5 - 17.5 (+5)
It is still impossible to say whether the overall better results would extend
with further testing, but it does suggest that the function should be thoroughly
tested.
One curiosity appeared in the match forcing me to manually edit the result: the
Arena GUI was set to resign if both opponents reach a score of -9.0, well due to
a certain blindness by BOTH opponents (though no one resigned) the following
position came up:
8/6kP/8/8/8/2p1K3/2B5/8 w - - 0 81
And both List 5.12 and Pro Deo (as white) thought white was dead won, and gave
+10 scores for white. The GUI kicked in and declared Pro Deo the winner. I set
this to a draw of course.
Albert

[Gábor Szots]

Geschrieben von:/Posted by: Gábor Szots at 15 September 2004 08:09:13:
Als Antwort auf:/In reply to: Results of Pro Deo [Time Control] parameter geschrieben von:/posted by: Albert Silver at 15 September 2004 04:57:19:

Hi Albert,

I ran a 3rd Nunn 2 match against List 5.12 in order to get more results on the
[Time Control = 125] compared to Pro Deo's default settings.
Conditions: Athlon 2400XP+, Nunn 2 set, all learning off, ponder off, 5min + 3s:
List 5.12
---------
Default (without TC=125) : 22 - 18 (+4)
With [Time Control = 125]: 21.5 - 18.5 (+3)
Ruffian 1.05
------------
Default (without TC=125) : 16 - 24 (-8)
With [Time Control = 125]: 20.5 - 19.5 (+1)
Aristarch 4.5
-------------
Default (without TC=125) : 21.5 - 18.5 (+3)
With [Time Control = 125]: 22.5 - 17.5 (+5)
It is still impossible to say whether the overall better results would extend
with further testing, but it does suggest that the function should be thoroughly
tested.
One curiosity appeared in the match forcing me to manually edit the result: the
Arena GUI was set to resign if both opponents reach a score of -9.0, well due to
a certain blindness by BOTH opponents (though no one resigned) the following
position came up:
8/6kP/8/8/8/2p1K3/2B5/8 w - - 0 81
And both List 5.12 and Pro Deo (as white) thought white was dead won, and gave
+10 scores for white. The GUI kicked in and declared Pro Deo the winner. I set
this to a draw of course.
Albert

It may well be that the TC parameter is advantageous at incremental clock but not at x moves in y minutes. I will try this in my tournament when I also noticed that ProDeo/Rebel moves too fast.
I think you made a mistake here, because this is a theoretical draw but black can lose the game by a blunder. Therefore the game should be continued at least up to the point where it is clear that black has no intention to leave the corner.
Regards,
Gábor

[Gábor Szots]

Geschrieben von:/Posted by: Gábor Szots at 15 September 2004 08:11:11:
Als Antwort auf:/In reply to: Re: Results of Pro Deo [Time Control] parameter geschrieben von:/posted by: Gábor Szots at 15 September 2004 08:09:13:

I forgot to tell you that I highly appreciate your efforts relating this TC issue. Thank you.
Gábor

[Kurt Utzinger]

Geschrieben von:/Posted by: Kurt Utzinger at 15 September 2004 09:53:10:
Als Antwort auf:/In reply to: Results of Pro Deo [Time Control] parameter geschrieben von:/posted by: Albert Silver at 15 September 2004 04:57:19:

Hi Albert
I am repeating the matches Pro Deo 1.0 has played
with the Nunn2-positions vs some other engines
http://www.utzingerk.com/pro_deo_htm
and this time with own books, learning on an
[Time Control = 125] * 125
[Time Control Status = ALWAYS] * ALWAYS|PB ONLY
and so far Pro Deo 1.0 is doing much better
at 60m+10s (in comparison with the Nunn2-matches)
8.5-7.5 Pro Deo vs Fritz8
4.5-3.5 Pro Deo vs Ruffian 2.1.0
and I can gladly confirm that Pro Deo's time
management is excellent with the paramaters
mentioned above.
Kurt


Kurt & Rolf Chess

[Kurt Utzinger]

Geschrieben von:/Posted by: Kurt Utzinger at 15 September 2004 09:55:54:
Als Antwort auf:/In reply to: Re: Results of Pro Deo [Time Control] parameter geschrieben von:/posted by: Kurt Utzinger at 15 September 2004 09:53:10:

Hi Albert
I am repeating the matches Pro Deo 1.0 has played
with the Nunn2-positions vs some other engines
http://www.utzingerk.com/pro_deo.htm
and this time with own books, learning on an
[Time Control = 125] * 125
[Time Control Status = ALWAYS] * ALWAYS|PB ONLY
and so far Pro Deo 1.0 is doing much better
at 60m+10s (in comparison with the Nunn2-matches)
8.5-7.5 Pro Deo vs Fritz8
4.5-3.5 Pro Deo vs Ruffian 2.1.0
and I can gladly confirm that Pro Deo's time
management is excellent with the paramaters
mentioned above.
Kurt


Kurt & Rolf Chess

[Albert Silver]

Geschrieben von:/Posted by: Albert Silver at 15 September 2004 16:11:20:
Als Antwort auf:/In reply to: Re: Results of Pro Deo [Time Control] parameter geschrieben von:/posted by: Gábor Szots at 15 September 2004 08:09:13:

One curiosity appeared in the match forcing me to manually edit the result: the
Arena GUI was set to resign if both opponents reach a score of -9.0, well due to
a certain blindness by BOTH opponents (though no one resigned) the following
position came up:
8/6kP/8/8/8/2p1K3/2B5/8 w - - 0 81
And both List 5.12 and Pro Deo (as white) thought white was dead won, and gave
+10 scores for white. The GUI kicked in and declared Pro Deo the winner. I set
this to a draw of course.

I think you made a mistake here, because this is a theoretical draw but black can lose the game by a blunder. Therefore the game should be continued at least up to the point where it is clear that black has no intention to leave the corner.
Regards,
Gábor

I'm afraid you didn't quite understand. The GUI already stopped the game and declared white the winner. I changed the result only. As to blundering, I don't see how that is possible. Black can only move to two squares with the King: h8 and g7, without an *immediate* queen promotion (the pawn is on h7 after all). That requires all of 2 plies at most to see it, if that, and I don't think it will EVER miss that no matter how little time is given. It had a 3 second increment per move in the game as it was.
Albert