SCCT Rating List-Houdini 2.0t3 with Perfect 2012c book

By Sedat Canbaz Date 2012-08-11 11:19

Hello Chess Friends,

Last Engine Uptates:
Houdini 2.0t3 Pc x64 6c
Rybka 4.1 79DT v1 x64 6c
Houdini 2.0Bar2 x64 6c
Bouquet 1.4 x64 6c

SCCT Rating List - All Versions:
http://www.sedatcanbaz.com/chess/scct-rating/

SCCT Rating List - Top 20:
http://www.sedatcanbaz.com/chess/sccttop20/

Conditions:
http://www.sedatcanbaz.com/chess/2838-2/

Games:
http://www.sedatcanbaz.com/chess/games/scct_3m2s.rar

Some Notes:
-Houdini 2.0t3 Pc x64 6c is same version as Houdini 2.0t3 Pro x64 6c,just this time is played with another new book:Perfect 2012c
*Actually during the current Houdini 2.0t3 Pc x64 6c test,all its opponents used same Perfect 2012c book too
-Looking at the current early results,it seems the various openings and shorter lines did not make a BIG negative influence over its performance
*But anyway,for a better conclusion: we need more games,probably this number of games between 1000-2000 games will be enough data
-Rybka 4.1 79DT v1 x64 6c is performed slightly better than default,but of course to be sure more games are required...
-I noticed also that in other testings (Non-SCCT results),there are some different standings (e.g i mean especially for Houdini settings)
-From my experience i can say that such different performance is appearing is not due to 95% confidence of error margin or we need at least 5.000 games and so on
*Sure more games is better,but i highly doubt that SCCT standings will be changed a lot,e.g i mean about after 5.000 games player

In other words,
*Of course I can be wrong...but however i have a little bit experience in Computerchess (10 years as Book Maker and 15 years as Tester)
And i strongly believe in that:
-Some SCCT Participants performed higher or lower than others,due to my used conditions are completely different than all available testings

Kind Regards,
Sedat

By Sedat Canbaz Date 2012-08-12 17:45

New Update:
-More games have been added,played mainly by Houdini 2.0t3 and Rybka 4.1 79DT v1

SCCT Rating List - All Versions:
http://www.sedatcanbaz.com/chess/scct-rating/

Best,
Sedat

By Kurt Utzinger Date 2012-08-12 17:56

Hi Sedat

You have - in my opinion - taken a far
too high basic value to that the Elo ratings
of the engines look immeasurably exaggerated.
LOL : Fruit 090705 x64 6c 2969

Greetings
Kurt

By Ernest Bonnem Date 2012-08-12 19:19

[quote="Kurt Utzinger"]You have - in my opinion - taken a far too high basic value to that the Elo ratings of the engines look immeasurably exaggerated.
LOL : Fruit 090705 x64 6c 2969[/quote]
Hi Kurt,

Is that not just a matter of Elo translation?...
And besides, who knows that (very private!) 64-bit SMP Fruit 090705 x64 ???

By Kurt Utzinger Date 2012-08-12 21:41

[quote="Ernest Bonnem"]
[quote="Kurt Utzinger"]You have - in my opinion - taken a far too high basic value to that the Elo ratings of the engines look immeasurably exaggerated.
LOL : Fruit 090705 x64 6c 2969[/quote]
Hi Kurt,

Is that not just a matter of Elo translation?...
And besides, who knows that (very private!) 64-bit SMP Fruit 090705 x64 ???
[/quote]

Hi Ernest
Of course it's just a matter of translation, but I hate to read
such high Elo values.
Kurt

By Sedat Canbaz Date 2012-08-12 20:09

[quote="Kurt Utzinger"]
Hi Sedat

You have - in my opinion - taken a far
too high basic value to that the Elo ratings
of the engines look immeasurably exaggerated.
LOL : Fruit 090705 x64 6c 2969

Greetings
Kurt
[/quote]

Why not ?

As far as i know,still there is no any proof or data that:
- Fruit 090705 x64 using 6 cores (on i7 980X@4.0GHz) should be calculated less or higher

But however,in my opinion the current SCCT rating of Fruit suits more to human Elo ratings

For 100 % confirmation of course we need serious engine testings against humans...
But anyway, the available data (man vs machine matches) can be a good indication of right Elo measuring

Just i'd like to point out once more that even Pocket Hiarcs performed 2938 Elo in Mescosur Cup 2009:
http://www.hiarcs.com/Games/Mercosur2009/mercosur09.htm

And as we see...SCCT calculated Hiarcs 13.2 6 cores approx.130 Elo higher than Pocket Hiarcs
35 Hiarcs 13.2 w32 6c 3055 Elo

*Note also that in Mescosur Cup 2009:
- Hiracs is played on much slower PC than i7 980X @4.0GHz
- Plus, Hiarcs was older (weaker) version than the current Hiarcs 13.2

For more details::
http://www.talkchess.com/forum/viewtopic.php?t=43711&start=20

Hope this helps,
Sedat

By Kurt Utzinger Date 2012-08-12 21:20

Hi Sedat

I do not know what bullshit people play against Hiarcs pocket 13.2. For me
it is rather easy to get draws every third game against Hiarcs on my iPhone.

Kurt

By Kurt Utzinger Date 2012-08-12 21:39

By Sedat Canbaz Date 2012-08-12 22:57

[quote="Kurt Utzinger"]
Hi Sedat

I do not know what bullshit people play against Hiarcs pocket 13.2. For me
it is rather easy to get draws every third game against Hiarcs on my iPhone.

Kurt
[/quote]

Hello Kurt,

Just i'd like to correct you that the played version was Hiarcs 13,not Hiarcs 13.2

Plus,the opponents of Pocket Hiarcs were not bullshit people
They are respected strong GMs from Argentina,average 2407 Elo

I have no idea about your Human Elo performance and your iPhone processor speed details...
So it seems you are very strong human player...if you get very easy draws in every 4 games - really well done !!!

Note also that according to Hiarcs web site, Pocket Hiarcs is played on a HTC Touch HD (ARM/528Mhz) mobile device

For more details :
http://www.chessbase.com/newsprint.asp?newsid=5719

Regards,
Sedat

By Martin Müller Date 2012-08-13 00:05

Hi Sedat,

Kurt did not say by any means that these are "bullshit people", he mentioned that their playing
was inappropriate to get a satisfactoy result against such a type of computers respectively against computers in general.
In previous postings he mentioned that he is pretty successful by applying his very "dry" and unspectacular strategy which he would
never play against humans.

Just to sort out a misunderstanding.

Kind regards
Martin

By Kurt Utzinger Date 2012-08-13 14:10

[quote="Martin Müller"]
Hi Sedat,

Kurt did not say by any means that these are "bullshit people", he mentioned that their playing
was inappropriate to get a satisfactoy result against such a type of computers respectively against computers in general.
In previous postings he mentioned that he is pretty successful by applying his very "dry" and unspectacular strategy which he would
never play against humans.

Just to sort out a misunderstanding.

Kind regards
Martin
[/quote]

Hi Martin
Many thanks for this clarification.
I am tired to repeat again and again
the way that humans should play
against today's computer monsters.
If I can - with only 1952 Elo - reach
draws at every fourth game, there
is no doubt that all (let's say) 2400
Elo players should easily be able
to get draws in almost every game,
thus giving the chess monsters not
more than 2500 Elo. The problem is
that it's very dangerous to apply
such a boring style, a style which
is totally against the way chess is
being played today at IM and GM level.
Much exercise and a complete adaptation
of thinking in chess is necessary to do so
and this may have a very negative influence
on the general playing strength of a good
chess player.
Regards
Kurt

By Sedat Canbaz Date 2012-08-13 20:04

Just my two cents more over this issue,

Actually the above table does not need any comments,but however once more i will try to mention:
-Its clear proof that even more than 10 years ago the older engines started to play on GMs levels,average 2750 Elo

Now we are in 2012 (more than 10 years is past),
We have much faster processors and much stronger engines than the past
Simply just is needed to be checked the ratings,e.g the played older versions of 2000 years (Rebel,Fritz,Junior)
vs with the current Top chess engines of 2012 year (e.g Houdini,Critter,Rybka...)
But anyway,i will try to help:
-The Elo difference between 2000 years engines and 2012 years engines is approx.400-500 Elo
-The hardware Elo difference between 2000 years processors and 2012 years processor is approx.200-300 Elo
*In other words,there will be a BIG mistake,if we rate Houdini 2.0c x64 6 core (on i7 980X 4.0GHz) less than 3300 Elo points

For more details:
http://www.talkchess.com/forum/viewtopic.php?t=42553&postdays=0&postorder=asc&topic_view=flat&start=30

Btw,with respect to all:
-I prefer to check Man vs Machine results,played by trusted official sources
*To be honest,i don't care much on 'human against machine' results,played at home

Greetings,
Sedat

By Sedat Canbaz Date 2012-08-13 00:38

[quote="Kurt Utzinger"]
[quote="Ernest Bonnem"]
[quote="Kurt Utzinger"]You have - in my opinion - taken a far too high basic value to that the Elo ratings of the engines look immeasurably exaggerated.
LOL : Fruit 090705 x64 6c 2969[/quote]
Hi Kurt,

Is that not just a matter of Elo translation?...
And besides, who knows that (very private!) 64-bit SMP Fruit 090705 x64 ???
[/quote]

Hi Ernest
Of course it's just a matter of translation, but I hate to read
such high Elo values.
Kurt
[/quote]

Hello again Kurt,

Actually its a strange and funny to read such comments especially by you

Btw,I dont see nothing wrong,when we compare SCCT Elo results with other major rating lists:

SCCT: Naum 4.2 x64 6c   3170  Elo

SSDF: Naum 4.2 MP x64 2GB Q6600 2,4 GHz  3150 Elo

CCRL: Naum 4.2 64-bit 4CPU   3152 Elo

Btw,did you hate only SCCT Elo high measurements,what about rest engine rating lists ?
And one question more:
-What is your opinion about Naum 4.2 x64 6c's real Elo rating should be on i7 980X 4.0GHz ?

Thanks in advance,
Sedat

By Ernest Bonnem Date 2012-08-13 10:58

[quote="Sedat Canbaz"][quote="Ernest Bonnem"]that (very private!) 64-bit SMP Fruit 090705 x64 ???
[/quote][/quote]
Hi Sedat,

Whom should I ask, in order to "try" to obtain that (very private!) 64-bit SMP Fruit 090705 x64 ?

By Sedat Canbaz Date 2012-08-13 22:55

[quote="Ernest Bonnem"]
[quote="Sedat Canbaz"][quote="Ernest Bonnem"]that (very private!) 64-bit SMP Fruit 090705 x64 ???
[/quote][/quote]
Hi Sedat,

Whom should I ask, in order to "try" to obtain that (very private!) 64-bit SMP Fruit 090705 x64 ?
[/quote]

Hello Ernest,

I think you need to ask Fruit team

Best,
Sedat

By Sedat Canbaz Date 2012-08-14 22:24

The current standings:

Code:

Rank Name                         Elo     +    - games score oppo. draws 
   1 Houdini 2.0t3 Pro x64 6c     3365   12   12  1588   70%  3210   39%
   3 Houdini 2.0t3 Pc x64 6c      3359   18   18   701   74%  3180   39%
  11 Rybka 4.1 79DT v1 x64 6c     3292   17   17   753   66%  3168   37%
  15 Deep Rybka 4.1 x64 6c        3277   11   11  1735   60%  3205   48%

SCCT Rating List - All Versions:
http://www.sedatcanbaz.com/chess/scct-rating/

Greetings,
Sedat

By Sedat Canbaz Date 2012-08-17 06:23

New Standings:

Houdini 2.0t3/Rybka 4.1 79DT v1's standings after 1000 games per player:
http://www.sedatcanbaz.com/chess/scct-rating/

SCCT-BayesElo's calculation notes:
offset 3300 Critter 1.6 x64 6c
mm 1 1

More Details:
-Surprisingly, Houdini 2.0t3 using a shorter book+more various openings performed identically so far
-Well done to Mindbreaker, Rybka 4.1 79DT v1 x64 6c performed 15 Elo better than default version
-Next SCCT Engine,which will be tested:Stockfish VE09 x64 6c

Best Wishes,
Sedat