Elo differences between: Human vs Computer

[carldaman]

Hi Sedat,
Deep Fritz 10.1 is really a multi-processor bugfix of DF10. It might make sense to use that as the standard. DF 10.1 equal to 2900 on 4 cores sounds about right.

Regards,
CL

[Sedat Canbaz]

Hi Sedat,
Deep Fritz 10.1 is really a multi-processor bugfix of DF10. It might make sense to use that as the standard. DF 10.1 equal to 2900 on 4 cores sounds about right.

Regards,
CL

Hello Carl,

Agreed...but I started already (with your recommendation) testing Fritz 10 in my scratch competition and currently I have no plans to test DF 10.1 as mp, maybe later I can test DF 10.1, I mean with 6 cores as reference. And in case of such test (on my i7 980X or i7 970) , I think to fix DF 10.1 min at 3000 Elo

For example, fixing Elo calculation of DF 10.1 (on similar hardwares as AMD 4600) at 2900 Elo sounds good...

And for the current scratch conditions, I plan to set Fritz 8 at 2850 Elo or maybe 2900 Elo

Best,
Sedat

[mehmet karaman]

Dear Sedat thanks for your efforts about computer chess.

For my opinion the elo grade of Fritz 10 is 2900+ for SCCT hardwares.

Maybe some people think 2900 elo for Fritz 10 is too high. I ask them do you have any proof about that

And i ask them what is the elo performance of Pocket Fritz 4 based on Hiarcs 13 engine on 20 Kn/S system at Mercosur Cup 2009. (Samsung Galaxy S3 is ~30 times stronger than this system)

http://en.chessbase.com/post/breakthrou ... enos-aires

[Sedat Canbaz]

Dear Sedat thanks for your efforts about computer chess.

For my opinion the elo grade of Fritz 10 is 2900+ for SCCT hardwares.

Maybe some people think 2900 elo for Fritz 10 is too high. I ask them do you have any proof about that

And i ask them what is the elo performance of Pocket Fritz 4 based on Hiarcs 13 engine on 20 Kn/S system at Mercosur Cup 2009. (Samsung Galaxy S3 is ~30 times stronger than this system)

http://en.chessbase.com/post/breakthrou ... enos-aires

Hello dear Mehmet,

You are welcome, and thanks for your confirmation

Yes, fixing Fritz 10 at 2900 Elo sounds good for me too
So... the new Elo engine calculations (for SCCT Scratch Rating) will be appear in the next update

Agreed... Hiarcs 13 won Mercosur Cup 2009 with outstanding performance, even at much slower hardware than we use nowadays

I am quite sure some of the chess friends (who does not believe that there is at least 500 Elo difference between Man vs Machine)
will change their minds about the performance of the latest Top engines, but I think they still need a little bit time more

Best,
Sedat

[Sedat Canbaz]

My old chess site is online again
It seems there was a server problem during the last days ...

And now about the current issue,
Oh sorry... (before I remembered 300 Elo) and now as we see:
- Deep Fritz 11 has approx. 400 Elo difference between QX6700@3.0GHz and Celeron 1.70GHz
- Rybka 3 has approx. 500 Elo difference between QX6700@3.0GHz and Celeron 1.70GHz
- Crafty 22.8 has approx. 400 Elo difference between QX6700@3.0GHz and Celeron 1.70GHz

So...Blitz's overall Elo difference can be counted as average: 450 Elo
I expect on Slow time controls to see Elo difference approx. 350 Elo


Probably on 2x Xeon E5-2697 v2 2700 Mhz 24 cores:
-There will be at least 600 Elo difference

I hope the below values and data will be helpful too

And here is the full table:

SCCT Hardware Tournament



http://sedatchess.110mb.com/index.php?p=1_31

Code: Select all

Rank  Hardware          Program               Elo   Games   Kn/s    Hardware User

01 2xQX9775@4.0GHz      Rybka 3 x64 t8        3301   200    432     Majd Ansari
02 Core i7 920@4.1GHz   Rybka 3 x64 t4        3246   200    414     Phil Harris
03 QX9650@3.8GHz        Rybka 3 x64 t4        3238   250    320     Sedat Canbaz
04 QX6700@3.0GHz        Rybka 3 x64 t4        3217   600    230     Sedat Canbaz
05 AMD 4600 2.40GHz     Rybka 3 x64 t2        3114   550    95      Sedat Canbaz
06 AMD 4600 2.40GHz     Rybka 3 x64 t1        3067   200    55      Sedat Canbaz
07 AMD 4600 2.40GHz     Rybka 3 w32 t2        3066   200    65      Sedat Canbaz
08 QX6700@3.0GHz        Deep Fritz 11 t4      3056   550    8080    Sedat Canbaz
09 AMD Turion 2.20GHz   Rybka 3 w32 t2        3047   350    50      Sedat Canbaz
10 AMD 3400 2.40GHz     Rybka 3 w32 t1        3027   850    35      Sedat Canbaz
11 AMD 4600 2.40GHz     Deep Fritz 11 t2      2963   500    2695    Sedat Canbaz
12 QX6700@3.0GHz        Toga II 1.4.2JD t4    2918   400    4143    Sedat Canbaz
13 AMD Turion 2.20GHz   Deep Fritz 11 t2      2916   200    2519    Sedat Canbaz
14 AMD 3400 2.40GHz     Deep Fritz 11 t1      2907   300    1252    Sedat Canbaz
15 AMD 4600 2.40GHz     Toga II 1.4.2JD t2    2890   450    1808    Sedat Canbaz
16 AMD 4600 2.40GHz     Toga II 1.4.2JD t1    2854   100    915     Sedat Canbaz
17 AMD Turion 2.20GHz   Toga II 1.4.2JD t2    2848   300    1457    Sedat Canbaz
18 AMD 3400 2.40GHz     Toga II 1.4.2JD t1    2841   300    902     Sedat Canbaz
19 QX6700@3.0GHz        Crafty 22.8 x64 t4    2813   650    11796   Sedat Canbaz
20 Celeron 1.70GHz      Rybka 3 w32 t1        2722   450    7       Sedat Canbaz
21 AMD 4600 2.40GHz     Crafty 22.8 x64 t2    2715   700    4020    Sedat Canbaz
22 AMD 4600 2.40GHz     Crafty 22.8 x64 t1    2668   100    2067    Sedat Canbaz
23 AMD 3400 2.40GHz     Crafty 22.8 w32 t1    2657   400    1431    Sedat Canbaz
24 Celeron 1.70GHz      Deep Fritz 11 t1      2653   350    409     Sedat Canbaz
25 AMD Turion 2.20GHz   Crafty 2.28 w32 t2    2616   200    2389    Sedat Canbaz
26 Celeron 1.70GHz      Toga II 1.4.2JD t1    2532   200    251     Sedat Canbaz
27 Celeron 1.70GHz      Crafty 22.8 w32 t1    2410   350    423     Sedat Canbaz

*Rybka 3 x64 t4 is tested without using any opening book:
1.QX6700@3.0GHz(Rybka 3 x64 t4+No Book) = 57.5/100
2.AMD46002.4GHz(Rybka 3 x64 t2+Storm 3.1) = 42.5/100

*Default conditions (for both engines:Perfect 2009 book):
1.QX6700@3.0GHz(Rybka 3 x64 t4+Perfect2009) = 71.0/100
2.AMD46002.4GHz(Rybka 3 x64 t2+Perfect2009) = 29.0/100

Tournament Details:
Chess GUI:ChessProgram11(Nov 25 2008)
O/S:Windows XP Professional x64
Test Type:Playchess Server/Auto 232,Engine-Matches
Time Control: 3 Min Game
HashTable Size:128 MB
Opening Book for all:Perfect 2009 beta (up to 12 moves)
Book Learning:OFF
Permanent Brain:ON
Tablebases:3-4-5-MEN EGTB
TB Cache:16 MB
Tournament Director:Sedat Canbaz

More details:All games are played on equal conditions,exception the hardwares.
The tournament is running on separate computers by using Auto232 player.
(some of the games are played on Playchess server).For all games is used same neutral book:
Perfect 2009 beta (a new book which is based on only GrandMaster games-average 2600 elo)
The Rating List is calculated by Bayeselo
t8 = Engine is using 4 threads (8 CPU)
t4 = Engine is using 4 threads (4 CPU)
t2 = Engine is using 2 threads (2 CPU)
t1 = Engine is using 1 thread (1 CPU)
x64 = Engine is optimized for Windows x64 bit Operating System
w32 = Engine is optimized for Windows x32 bit Operating System

*Special thanks to:
Majd Ansari
Phil Harris

[Father]

Sedat,
Thank you very much for these análisis. I am not sure if my coments Would be usefull for you. In my lást 100 Games where i have uses new Línes my performance is just of 4% but the day before testerday my performances against top chess sowfwares was 25% in the lást 10 Games. 2.5 on 10!!!
I dont have any elo And my style have. Been crítizied but ti is a reality ...
I repet, i hope Its coments Will be susefull to you
Best regards,
Father And grandfather play chess

[Sedat Canbaz]

Sedat,
Thank you very much for these análisis. I am not sure if my coments Would be usefull for you. In my lást 100 Games where i have uses new Línes my performance is just of 4% but the day before testerday my performances against top chess sowfwares was 25% in the lást 10 Games. 2.5 on 10!!!
I dont have any elo And my style have. Been crítizied but ti is a reality ...
I repet, i hope Its coments Will be susefull to you
Best regards,
Father And grandfather play chess

Hello Father!

Not at all...its my pleasure

Yes...my congratulations! there are not many as you..!!

It seems you are computer killer...well-done !

Is there any possibility to download your games vs machines

Thanks,
Sedat

[Father]

Hello Sedat,
I don’t have in a file or book the games than I have played in my life against chess computers. I have played a lot of games don’t know exactly how many games have I played but there are thousands, and many of these has been eraser from different hard disk for few decades.
More than a book I have tried to have a system of plays. A strategic plan. On 2014 I started again, because of I was retired of chess more or less for 3 years at least, and there are many new ideas in Queen inside, King inside and center.
Here in play chess there is a history in my posts, playing against different engines and different hardware since 2006 at least. In rybka forum there are some games too, and in internet various places some games and a basic explanation about the chess- system that I have called "Antichess.
In the last 3 years chess engines have jumped a lot. They are very stronger day per day. Since Rybka were invented by Vasik chess engines are really different, they are much better than before.
Today to search a score better than 5% is a hard work for everybody. But in new days I believe, than it is possible to search a score of 20% in man versus machine.
Best regards,
Father playchess.com user

[mehmet karaman]

2700+ elo players can’t beat top programs but some 1500-2000 elo players claim that they beat these programs by anti-computer tactics.

There a lot of ways …

To use another computer and chess program, to look at the program analyze, to change the program configuration, to disable the opening book, to give program very little time to play…..
( I have a lot of pgn files in my database they show that i beat or drew these top chess programs, but in fair games i can’t beat them never )

Do you know any person to beat one of the top program (Stockfish, Komodo,Houdini, Rybka,Fritz …) with a strong opening book at tournament time controls under the supervision of referee in last five years.

[Father]

Hello Mehmet,

The art and science in man against machine, in chess, there is not a wealth of elites, a wealth of the owners of Medias, or a wealth of the owners of titles, or owner of honors .... Knowledge in man against machine belongs to those persons who work hard playing computers.
So then, if a world chess champion works hard, science would be reached in man against machine for him-her.
So then if a simple amateur work hard, this person will be reaching science on man against machine. T
Thanks to computer chess, you not need even leave the house to play against players of higher level, silicone or silicone player’s monsters without needing the approval of the media, federations, and owners of world power.