Dr. Deeb´s FireBird settings

[Sentinel]

Believe it or not, but, MATERIAL VALUES ARE INSENSITIVE TO TIME CONTROL.
So please, stop this BS about correct and incorrect TC.
If one set of material values is better than the default set at 1''/game it will certainly be better at 10h/game. This is a fact. Period.

[Adam Hair]

Believe it or not, but, MATERIAL VALUES ARE INSENSITIVE TO TIME CONTROL.
So please, stop this BS about correct and incorrect TC.
If one set of material values is better than the default set at 1''/game it will certainly be better at 10h/game. This is a fact. Period.

I am currently in the middle of a series of tests to see if material values affect engine strength at different time
controls, using Fire 1.1 x64 and Wael Deeb's settings as the guinea pig. I am testing at three different incremental
time controls:
6 sec + .1 sec
60 sec + 1 sec
600 sec + 10 sec

I could not see how material values would have an effect at different time controls, but Dr. Deeb is so adamant that he
has killer settings for Fire 1.1 that I thought I would test them. If Milos is correct, and I have no reason to doubt that
he has superior knowledge concerning this, then any difference we see in Fire 1.1 64-bit WD's results should be
attributed to other factors. A large difference in the results may point to Dr. Deeb being correct.

Here are some preliminary results:( I hope this is legible for everybody )

CPU : E8400 @ 2.95 GHz
Windows XP 64-bit
Little Blitzer 2.3 tournament manager

Time Control = 6" + .1"
Hash = 16MB ( plus 4 MB pawn hash for Fire 1.1 WD)
358 start positions
Games Completed = 3581 of 3580 (Avg game length = 10.950 sec)
Time = 20697 sec elapsed, -6 sec remaining
                                                                                                                      Avg      Avg       Avg
    Engine                              Score      Win-Loss-Draw  Reason for Loss    Reason for Draw                  Time/Move  Depth  Nodes/Second 
 1.FireBird 1.1 x64 WD               1703.5/3581  512-686-2383   (L: m=685 t=1 i=0) (D: r=540 i=109 f=1724 s=10)     (tpm=161.4 d=12.7 nps=1853239)
 2.Tankist 1.7 32-bit      	336.5/717 	 96-140-481   (L: m=140 t=0 i=0) (D: r=112 i=15 f=353 s=1) 	  (tpm=171.6 d=12.6 nps=1395444)
 3.IvanHoe-T63Mini7        	395.0/717 	 144-71-502   (L: m=71 t=0 i=0)  (D: r=119 i=29 f=352 s=2) 	  (tpm=171.9 d=13.1 nps=1833019)
 4.RobboLito 0.085g3 x64   	385.0/716 	 138-84-494   (L: m=84 t=0 i=0)  (D: r=133 i=25 f=336 s=0) 	  (tpm=170.6 d=13.3 nps=2140830)
 5.Stockfish 1.8 JA 64bit  	348.0/716 	107-127-482  (L: m=127 t=0 i=0) (D: r=54 i=10 f=416 s=2) 	  (tpm=168.2 d=13.8 nps=1685808)
 6.Houdini 1.03a x64 1_CPU 	413.0/715 	 201-90-424   (L: m=89 t=0 i=1)  (D: r=122 i=30 f=267 s=5) 	  (tpm=157.4 d=13.7 nps=2360345)

m = material loss, t = time loss, i = illegal move
r = repetition, i = insufficent material, f = fifty moves, s = stalemate





Time Control = 1' + 1"
Hash = 64MB ( plus 16 MB pawn hash for Fire 1.1 WD)
358 start positions 
Games Completed = 1614 of 3580 (Avg game length = 126.211 sec)
Time = 102509 sec elapsed, 124866 sec remaining
 1.  FireBird 1.1 x64 WD       	806.5/1614 	206-207-1201 (L: m=207 t=0 i=0)     (D: r=333 i=88 f=771 s=9) 	(tpm=1701.1 d=17.7 nps=1829954)
 2.  Tankist 1.7 32-bit        	146.0/324 	27-59-238 	 (L: m=59 t=0 i=0) 	(D: r=68 i=15 f=154 s=1) 	(tpm=1768.1 d=17.1 nps=1380313)
 3.  IvanHoe-T63Mini7          	165.5/323 	36-28-259 	 (L: m=28 t=0 i=0) 	(D: r=78 i=20 f=159 s=2) 	(tpm=1618.0 d=17.9 nps=1950501)
 4.  RobboLito 0.085g3 x64     	167.5/323 	44-32-247 	 (L: m=32 t=0 i=0) 	(D: r=76 i=18 f=152 s=1) 	(tpm=1752.9 d=17.9 nps=2106672)
 5.  Stockfish 1.8 JA 64bit    	158.0/322 	44-50-228 	 (L: m=50 t=0 i=0) 	(D: r=27 i=7 f=191 s=3) 	(tpm=1585.1 d=20.3 nps=1918949)
 6.  Houdini 1.03a x64 1_CPU   	170.5/322 	56-37-229 	 (L: m=35 t=0 i=2) 	(D: r=84 i=28 f=115 s=2) 	(tpm=1601.2 d=18.5 nps=2367761)

[Dr.Wael Deeb]

Believe it or not, but, MATERIAL VALUES ARE INSENSITIVE TO TIME CONTROL.
So please, stop this BS about correct and incorrect TC.
If one set of material values is better than the default set at 1''/game it will certainly be better at 10h/game. This is a fact. Period.

I didn't say that the material values are sensetive to time control,don't put words in my mouth....
It's reaching a certain enough depth with certain time control what is matters....now that makes sense....
Dr.D

[DrChess]

Hello everybody,

normally I always use the engines as they come without getting into settings or configuration files but this thread makes me want to try the settings.
My question is, how exactly do you do that? can I just edit the firebird.cfg file and paste Dr. Deeb's lines?

Thanks for your help.

[Uly]

I'd do it this way:

Change some settings and save, see how the text file is changed, then, edit the text file and change it to Deeb's settings. Usually, the GUI allows you to just change the settings and save them as a new personality.

[stvs]

can anyone explain what the "move on ponderhit" parameter means?

[Uly]

When the opponent plays the expected move the engine doesn't think at all and play the already thought reply immediately, this tends to give the engine a time advantage at blitz and bullet time controls.

[Eduard Nemeth]

BTW: My newest and strong settings for only FIRE 1.1 you can find here:

http://www.open-chess.org/viewtopic.php?f=3&t=642

ED.

[ernest]

Well, in the German CSS Forum, Ernst Happe just tried your settings ("FireBird_11_w32_ Dr. Wael Deep settings"),
with not much success...

http://forum.computerschach.de/cgi-bin/ ... 7#pid30287

[Dr.Wael Deeb]

Well, in the German CSS Forum, Ernst Happe just tried your settings ("FireBird_11_w32_ Dr. Wael Deep settings"),
with not much success...

http://forum.computerschach.de/cgi-bin/ ... 7#pid30287

Well,anything goes when testing the 32w version

Seriously though,these settings perform much better at longer time controls and the reason is that it must reach deeper depth as it plays a very aggressive chess style which can harm the performance when testing at blitz time controls....
Dr.D