FireBird 1.2 vs Rybka 3 40/4 100 games

[kranium]

my latest test, finished this am:

1: FireBird 1.2 w32 62.5/100
2: Rybka 3 w32 37.5/100 1===000===0110=0=10=0110=1000=010====011==0===000=01=01=1=000===0=000===000010=0=1=00=0===0000===1==

100 games played / Tournament is finished
Level: Tournament 40/4
2 threads each
ponder on
FB used both TotalBases and TripleBases
Dell Intel Xeon 2 CPU server
Windows Server 2003

[Vinvin]

Thanks for the match !

Could we get the games somewhere, please ?

[beram]

Does Triplebase/ Totalbase influences the results ?

[kranium]

Does Triplebase/ Totalbase influences the results ?

yes...of course, they help greatly.
i think using both triplebases and totalbases in combination is best.

the triplebases only have info about whether or not a 3, 4, or 5 man endgame position is won, lost, or drawn.
this info is very important and can be used when (for ex in the late middlegame) the engine is deciding if a particular (simplifying) continuation results in a won endgame or not...

the totalbases however, act more like Nalimov EGTBs...they contain exact info about distance-to-mate, etc...
the 3, 4, or 5 piece endgame will normally be played instantly and with exact precision if totalbases are used...

does this make the engine stronger...?
i would have to argue 'yes'...absolutely of course (approx +20 ELO in my estimation)
the more info the engine has available...the better

----
the only disadvantage of the totalbases are their size...
it's quite a significant download (3.5 Gb?)

[benstoker]

Does Triplebase/ Totalbase influences the results ?

yes...of course, they help greatly.
i think using both triplebases and totalbases in combination is best.

the triplebases only have info about whether or not a 3, 4, or 5 man endgame position is won, lost, or drawn.
this info is very important and can be used when (for ex in the late middlegame) the engine is deciding if a particular (simplifying) continuation results in a won endgame or not...

the totalbases however, act more like Nalimov EGTBs...they contain exact info about distance-to-mate, etc...
the 3, 4, or 5 piece endgame will normally be played instantly and with exact precision if totalbases are used...

does this make the engine stronger...?
i would have to argue 'yes'...absolutely of course (approx +20 ELO in my estimation)
the more info the engine has available...the better

----
the only disadvantage of the totalbases are their size...
it's quite a significant download (3.5 Gb?)

EGTBs can actually impair an engine's strength.

[kranium]

Thanks for the match !

Could we get the games somewhere, please ?

of course:
http://www.chesslogik.com/downloads/FB1 ... -4-100.rar

forgot to mention:
Noomen Test Suite 2008
alternating colors
the 1st 100 games in this PGN
everything is the same except TC =1/1

[frcha]

Arena Guantlet, Ponder off, 4cpu, tc=+12s per move similar to 40mov/8min

no opening book, 12 preselected 1 or 2 ply openings . 128hash 32 pawn hash.

Code: Select all

FireBird_12_w32_new_SMP - Rybka 3_(cont 0) : 17.0/24 13-3-8 (=110=111=1=111==1=01101=)  71%  +156
-----------------Igorrit_0086v9_w32-----------------
Igorrit_0086v9_w32 - Rybka 3_(cont 0)      : 15.5/24 9-2-13 (=1=01=1=0==111==1===1=1=)  65%  +108
-----------------IvanHoe999963-----------------
IvanHoe999963 - Rybka 3_(cont 0)           : 13.5/24 6-3-15 (=0=1=1===11====01===0=1=)  56%   +42

And similar gauntlet to above, but with first 9 positions only.
6versions vs Rybka 3 contemp=0


tc= +6s similar to 40/4

Code: Select all

FireBird_12_w32_new_SMP  13.0 / 18 
IvanHoe999963             11.5 / 18   
FireBird_101_w32        11.5 / 18  
Igorrit_0086v9_w32       11.0 / 18 
Igorrit_0086v7_w32       11.0 / 18  
 FireBird_11_w32_in       8.0 / 18 

Looks to me that FireBird_12_w32_is improved on medium time controls.

I used firebird with 3-4 man triple/total bases only. Rybka had 3-4 namilov too.


The weird increment time controls allows igorrit to play without problems

[Vinvin]

Game #157 : FB 1.2 lost on time

Thanks for the match !

Could we get the games somewhere, please ?

of course:
http://www.chesslogik.com/downloads/FB1 ... -4-100.rar

forgot to mention:
Noomen Test Suite 2008
alternating colors
the 1st 100 games in this PGN
everything is the same except TC =1/1

[ernest]

FB used both TotalBases and TripleBases

Hi Norman,

With me, FB1.2 uses TotalBases and TripleBases OK when I enter the path in the Engine parameters (I use the Fritz 11 GUI), and register/load them.
But if I count on the auto-load of the firebird.cfg file to do the job (of course giving the correct path), that does NOT work...

Any help?

And by the way, what does the Verbose (TotalBases/TripleBases) Info (in the Engine parameters) do?

Addition:
I just ran it in console mode. Apparently, despite the path I wrote in config.cfg
TripleBase_Path C:\Documents and Settings\Myname\Mes documents\Ztemp_unzip\RobboBases\RobboTripleBase
it only uses
TripleBase Path: C:\Documents

Bug???

[Marc MP]

FB used both TotalBases and TripleBases

Hi Norman,

With me, FB1.2 uses TotalBases and TripleBases OK when I enter the path in the Engine parameters (I use the Fritz 11 GUI), and register/load them.
But if I count on the auto-load of the firebird.cfg file to do the job (of course giving the correct path), that does NOT work...

Any help?

And by the way, what does the Verbose (TotalBases/TripleBases) Info (in the Engine parameters) do?

Addition:
I just ran it in console mode. Apparently, despite the path I wrote in config.cfg
TripleBase_Path C:\Documents and Settings\Myname\Mes documents\Ztemp_unzip\RobboBases\RobboTripleBase
it only uses
TripleBase Path: C:\Documents

Bug???

Hi Enerst,

Fron my experience, the .cfg file for FB does the job. I do not need to thick the option in my Fritz 8 GUI and I get technical message like: 384 BB loaded.

From there I tried to unthick them in the GUI option, but the .cfg still prevailed.

Now what you're saying is that Fritz GUI settles them right even if youdon't touch to the FB .cfg file.

I'm concluding there is some king of a bug, as you can set them on in .cfg file and off from the GUI option (or vice-versa). One may prevail from the other one.

In practice I suggest you to use the FB .cfg file, and don't touch to the Fritz GUI engine option. You'll get a technical that they are loaded and that will show in the analysis pane when you will have hits on them.

You'll see they help greatly for reduced material postions.