Page 1 of 1

Corchess tests

Posted: Mon Aug 07, 2023 10:21 pm
by Ghppn
Cool Iris 10.50_bmi2 NOU-Anton 101 - CorChess 310723 sl-Eduard Nemeth

With learning disabled
I only used one core.
Opening: Sedat Cambaz Unique_ 50
Processor: Intel ® Core ™ i7– 4790 CPU 3.60 GHz (4 Cores, 8 Threads)
Memoriy (RM): 32 GB, System type: 64-bit, Windows 10.
Graphic board: NVIDIA Geforce GTX 760
Image
Games: https://mega.nz/file/HtARSZIK#I8xWF3SYf ... u1A4sr0I4g

Re: Corchess tests

Posted: Mon Aug 07, 2023 11:01 pm
by Homayoun
Thanks.

Re: Corchess tests

Posted: Tue Aug 08, 2023 9:15 am
by Sedat Canbaz
Homayoun wrote:
Mon Aug 07, 2023 11:01 pm
Thanks.
+1

Re: Corchess tests

Posted: Tue Aug 08, 2023 9:41 am
by Homayoun
I personally think this time control (5+2) is a good time control for corchess and the identical engines. The corchess based engines should never been participated in bullet time control matches against stockfish and also engines identical to stockfish (sfpb, polyfish ).on blitz time controls with increments and also rapid time controls, corchess based engines can show their capabilities well.
Best regards

Re: Corchess tests

Posted: Tue Aug 08, 2023 10:12 am
by Sedat Canbaz
Homayoun wrote:
Tue Aug 08, 2023 9:41 am
I personally think this time control (5+2) is a good time control for corchess and the identical engines. The corchess based engines should never been participated in bullet time control matches against stockfish and also engines identical to stockfish (sfpb, polyfish ).on blitz time controls with increments and also rapid time controls, corchess based engines can show their capabilities well.
Best regards
Well, let me say my 2 cents over these issues

I think that Corchess is very strong engine but
Very likely it requires slower time controls...

For example, in Ghppn test,
We see as Cool Iris as winner...

Actually I ran a serious tour some time ago: 5m+1s, MP, Ponder etc.
https://sites.google.com/site/computers ... nnue-cs-27

Btw,
Sure my test was with older versions...not sure about latest...

And here are the played results:

Code: Select all

    The Winner is again Stockfish dev - Congratulations to Stockfish team!

 Rank Name               Elo    +    -  games score oppo. draws 
   1 Stockfish 240223   3815    2    3   216   51%  3812   99% 
   2 SF PB Skynet       3813    2    3   216   50%  3812   99% 
   3 Kookaburra v1.03a  3812    3    3   216   50%  3812   99% 
   4 Dark Sister 1.9a   3812    3    3   216   50%  3812   99% 
   5 Kayra 1.7          3812    3    3   216   50%  3812  100% 
   6 SF PB ChessMan     3812    3    3   216   50%  3812  100% 
   7 CorChess 240223    3812    3    3   216   50%  3812   99% 
   8 Eman 8.70          3811    3    2   216   50%  3812  100% 
   9 PREDATOR X5        3811    3    2   216   50%  3812   99% 
  10 Genko 1.1          3810    3    2   216   50%  3812   98% 
 
• Calculated by Bayeselo / Draw Ratio (based on 1080 games) is very high: 99%

Conditions:
2x Epyc 7B12, Cutechess, 5m+1s, Ponder ON, 6 Cores/Game, 512 MB hash...
Best,
Sedat

Re: Corchess tests

Posted: Tue Aug 08, 2023 10:47 am
by Homayoun
Sedat Canbaz wrote:
Tue Aug 08, 2023 10:12 am
Homayoun wrote:
Tue Aug 08, 2023 9:41 am
I personally think this time control (5+2) is a good time control for corchess and the identical engines. The corchess based engines should never been participated in bullet time control matches against stockfish and also engines identical to stockfish (sfpb, polyfish ).on blitz time controls with increments and also rapid time controls, corchess based engines can show their capabilities well.
Best regards
Well, let me say my 2 cents over these issues

I think that Corchess is very strong engine but
Very likely it requires slower time controls...

For example, in Ghppn test,
We see as Cool Iris as winner...

Actually I ran a serious tour some time ago: 5m+1s, MP, Ponder etc.
https://sites.google.com/site/computers ... nnue-cs-27

Btw,
Sure my test was with older versions...not sure about latest...

And here are the played results:

Code: Select all

    The Winner is again Stockfish dev - Congratulations to Stockfish team!

 Rank Name               Elo    +    -  games score oppo. draws 
   1 Stockfish 240223   3815    2    3   216   51%  3812   99% 
   2 SF PB Skynet       3813    2    3   216   50%  3812   99% 
   3 Kookaburra v1.03a  3812    3    3   216   50%  3812   99% 
   4 Dark Sister 1.9a   3812    3    3   216   50%  3812   99% 
   5 Kayra 1.7          3812    3    3   216   50%  3812  100% 
   6 SF PB ChessMan     3812    3    3   216   50%  3812  100% 
   7 CorChess 240223    3812    3    3   216   50%  3812   99% 
   8 Eman 8.70          3811    3    2   216   50%  3812  100% 
   9 PREDATOR X5        3811    3    2   216   50%  3812   99% 
  10 Genko 1.1          3810    3    2   216   50%  3812   98% 
 
• Calculated by Bayeselo / Draw Ratio (based on 1080 games) is very high: 99%

Conditions:
2x Epyc 7B12, Cutechess, 5m+1s, Ponder ON, 6 Cores/Game, 512 MB hash...
Best,
Sedat
Thanks. It seems that recent versions of corchess show great progress. When I study the tests on different forums by different testers I see great results by crochess and it’s identical engines like cool iris or incognito even against eman and other engines.
Best regards

Re: Corchess tests

Posted: Tue Aug 08, 2023 10:49 am
by Homayoun
Excuse, instead of sfpb I typed eman.

Re: Corchess tests

Posted: Tue Aug 08, 2023 11:21 am
by Sedat Canbaz
Homayoun wrote:
Tue Aug 08, 2023 10:47 am
Sedat Canbaz wrote:
Tue Aug 08, 2023 10:12 am
Homayoun wrote:
Tue Aug 08, 2023 9:41 am
I personally think this time control (5+2) is a good time control for corchess and the identical engines. The corchess based engines should never been participated in bullet time control matches against stockfish and also engines identical to stockfish (sfpb, polyfish ).on blitz time controls with increments and also rapid time controls, corchess based engines can show their capabilities well.
Best regards
Well, let me say my 2 cents over these issues

I think that Corchess is very strong engine but
Very likely it requires slower time controls...

For example, in Ghppn test,
We see as Cool Iris as winner...

Actually I ran a serious tour some time ago: 5m+1s, MP, Ponder etc.
https://sites.google.com/site/computers ... nnue-cs-27

Btw,
Sure my test was with older versions...not sure about latest...

And here are the played results:

Code: Select all

    The Winner is again Stockfish dev - Congratulations to Stockfish team!

 Rank Name               Elo    +    -  games score oppo. draws 
   1 Stockfish 240223   3815    2    3   216   51%  3812   99% 
   2 SF PB Skynet       3813    2    3   216   50%  3812   99% 
   3 Kookaburra v1.03a  3812    3    3   216   50%  3812   99% 
   4 Dark Sister 1.9a   3812    3    3   216   50%  3812   99% 
   5 Kayra 1.7          3812    3    3   216   50%  3812  100% 
   6 SF PB ChessMan     3812    3    3   216   50%  3812  100% 
   7 CorChess 240223    3812    3    3   216   50%  3812   99% 
   8 Eman 8.70          3811    3    2   216   50%  3812  100% 
   9 PREDATOR X5        3811    3    2   216   50%  3812   99% 
  10 Genko 1.1          3810    3    2   216   50%  3812   98% 
 
• Calculated by Bayeselo / Draw Ratio (based on 1080 games) is very high: 99%

Conditions:
2x Epyc 7B12, Cutechess, 5m+1s, Ponder ON, 6 Cores/Game, 512 MB hash...
Best,
Sedat
Thanks. It seems that recent versions of corchess show great progress. When I study the tests on different forums by different testers I see great results by crochess and it’s identical engines like cool iris or incognito even against eman and other engines.
Best regards
Thanks to you too

Well, actually for a better conclusion,
We need many many games (per player)

E.g according to my experience,
Via strong openings, about 500-1000 games is enough
But If we use more critical, weak openings etc. then
It is required more than 2000- games (per player)

In other words, about Error margin:
It all depends on our used conditions...

And to be more clear, (sure if close to 3750 Elo engines)
If we re-run the same test: 100 games or 200 games (per player)
Very likely, then we may see not same/close standings..

But if we re-run
2000+ games (per player) then we can expect previous played results

Greetings

Re: Corchess tests

Posted: Tue Aug 08, 2023 11:32 am
by Homayoun
8-) thanks