Komodo? Nein.... Doch!

[Spacious_Mind]

Was taking a break from my Tourney when I decided to test some engines. But, keeping with my 5 ply theme, I figured why not try a little engine Test Tourney between engine versions of the same author. Therefore I decided to test Komodo and Doch.

So here is a first test of 300 games played in Arena with 5 PLY settings:

Code: Select all

    Program                          Elo    +   -   Games   Score   Av.Op.  Draws

  1 Doch32-133-ja                  : 2035   58  58   100    56.0 %   1993   30.0 %
  2 Doch12-ja                      : 2023   61  61   100    54.0 %   1995   22.0 %
  3 Doch13-ja                      : 2023   57  57   100    54.0 %   1995   32.0 %
  4 Doch_09.980_32_ja              : 1986   61  61   100    47.5 %   2003   21.0 %
  5 Doch32-134-ja                  : 1974   57  57   100    45.5 %   2005   31.0 %
  6 Komodo-10-32-ja                : 1959   59  60   100    43.0 %   2008   26.0 %

I used 2000 as a start rating. Individually the performances were like this.

Code: Select all

Individual statistics:

1 Doch32-133-ja             : 2035  100 (+ 41,= 30,- 29), 56.0 %

Doch_09.980_32_ja             :  20 (+ 10,=  5,-  5), 62.5 %
Komodo-10-32-ja               :  20 (+  7,=  4,-  9), 45.0 %
Doch32-134-ja                 :  20 (+  9,=  6,-  5), 60.0 %
Doch12-ja                     :  20 (+  7,=  5,-  8), 47.5 %
Doch13-ja                     :  20 (+  8,= 10,-  2), 65.0 %

2 Doch12-ja                 : 2023  100 (+ 43,= 22,- 35), 54.0 %

Doch_09.980_32_ja             :  20 (+ 11,=  0,-  9), 55.0 %
Komodo-10-32-ja               :  20 (+  5,=  8,-  7), 45.0 %
Doch32-134-ja                 :  20 (+ 13,=  2,-  5), 70.0 %
Doch13-ja                     :  20 (+  6,=  7,-  7), 47.5 %
Doch32-133-ja                 :  20 (+  8,=  5,-  7), 52.5 %

3 Doch13-ja                 : 2023  100 (+ 38,= 32,- 30), 54.0 %

Doch_09.980_32_ja             :  20 (+  8,=  3,-  9), 47.5 %
Komodo-10-32-ja               :  20 (+ 12,=  3,-  5), 67.5 %
Doch32-134-ja                 :  20 (+  9,=  9,-  2), 67.5 %
Doch12-ja                     :  20 (+  7,=  7,-  6), 52.5 %
Doch32-133-ja                 :  20 (+  2,= 10,-  8), 35.0 %

4 Doch_09.980_32_ja         : 1986  100 (+ 37,= 21,- 42), 47.5 %

Komodo-10-32-ja               :  20 (+ 10,=  5,-  5), 62.5 %
Doch32-134-ja                 :  20 (+  4,=  8,-  8), 40.0 %
Doch12-ja                     :  20 (+  9,=  0,- 11), 45.0 %
Doch13-ja                     :  20 (+  9,=  3,-  8), 52.5 %
Doch32-133-ja                 :  20 (+  5,=  5,- 10), 37.5 %

5 Doch32-134-ja             : 1974  100 (+ 30,= 31,- 39), 45.5 %

Doch_09.980_32_ja             :  20 (+  8,=  8,-  4), 60.0 %
Komodo-10-32-ja               :  20 (+ 10,=  6,-  4), 65.0 %
Doch12-ja                     :  20 (+  5,=  2,- 13), 30.0 %
Doch13-ja                     :  20 (+  2,=  9,-  9), 32.5 %
Doch32-133-ja                 :  20 (+  5,=  6,-  9), 40.0 %

6 Komodo-10-32-ja           : 1959  100 (+ 30,= 26,- 44), 43.0 %

Doch_09.980_32_ja             :  20 (+  5,=  5,- 10), 37.5 %
Doch32-134-ja                 :  20 (+  4,=  6,- 10), 35.0 %
Doch12-ja                     :  20 (+  7,=  8,-  5), 55.0 %
Doch13-ja                     :  20 (+  5,=  3,- 12), 32.5 %
Doch32-133-ja                 :  20 (+  9,=  4,-  7), 55.0 %

Really interesting that against the top 2 Komodo 1.0 had better performances.

Might run this Test again to see how this would look after a second set of 300 games.

Best regards

Nick

[gerold]

Guess l will run some games with the same two and see
how they do. I like to use game in 15/0.

Best,
Gerold.

[Spacious_Mind]

Guess l will run some games with the same two and see
how they do. I like to use game in 15/0.

Best,
Gerold.

Hi Gerold,

Go ahead, I am sure the results will be totally upside down in your test

rgds

Nick

[gerold]

Guess l will run some games with the same two and see
how they do. I like to use game in 15/0.

Best,
Gerold.

Hi Gerold,

Go ahead, I am sure the results will be totally upside down in your test

rgds

Nick

I couldn't fine Doch on JA page. I have Komodo. Guess i
will skip this one for now.
Best,
Gerold.

P.S. I will continue testing Stockfish.

[Spacious_Mind]

I couldn't fine Doch on JA page. I have Komodo. Guess i
will skip this one for now.
Best,
Gerold.

P.S. I will continue testing Stockfish.

If you want the Doch versions let me know and I will send them to you.

rgds

Nick

[lkaufman]

"

Was taking a break from my Tourney when I decided to test some engines. But, keeping with my 5 ply theme, I figured why not try a little engine Test Tourney between engine versions of the same author. Therefore I decided to test Komodo and Doch.

So here is a first test of 300 games played in Arena with 5 PLY settings:"

Your result is no surprise, for the simple reason that Komodo is a bit more selective than Doch, so it will test a bit worse on fixed depth but should normally get to a given ply faster and is clearly stronger in timed play.

[Spacious_Mind]

"

Was taking a break from my Tourney when I decided to test some engines. But, keeping with my 5 ply theme, I figured why not try a little engine Test Tourney between engine versions of the same author. Therefore I decided to test Komodo and Doch.

So here is a first test of 300 games played in Arena with 5 PLY settings:"

Your result is no surprise, for the simple reason that Komodo is a bit more selective than Doch, so it will test a bit worse on fixed depth but should normally get to a given ply faster and is clearly stronger in timed play.

Hi Larry,

Yes it was no surprise to me either... I expected that you would be performing differently at normal time settings. Going to run the test again though just for my curiosity to see how they end up a second time.

Still looking for that optimal engine that is great at low setting as well as top setting regards

Nick

[lkaufman]

"Still looking for that optimal engine that is great at low setting as well as top setting regards"
The problem is not whether the settings are low or not; Komodo will outperform Doch easily at one second per move or even at a much faster speed if your interface allows it. It's just that fixed depth settings, high or low, don't have the same meaning in different programs or even different versions of the same program. Five ply in Komodo might correspond to 4 1/2 ply in Doch for example.

[Spacious_Mind]

"Still looking for that optimal engine that is great at low setting as well as top setting regards"
The problem is not whether the settings are low or not; Komodo will outperform Doch easily at one second per move or even at a much faster speed if your interface allows it. It's just that fixed depth settings, high or low, don't have the same meaning in different programs or even different versions of the same program. Five ply in Komodo might correspond to 4 1/2 ply in Doch for example.

I am still looking for the ultimate game experience.... oldies against new.... dedicateds against engines. 1 MHz Colossus 4 or Mychess II, or 5 MHz Mephisto Polgar for example against engines. But the current platforms still don't quite allow it. Ideally to do that I would need a stable Arena/Chessbase/Winboard where I can play not 1 second per move but select setting from 1/100 second per move to be able to play manually a dedicated against an engine game. For example 30 seconds on a dedicated (depending on the MHz) is on my hardware 0.01 seconds for the engine in some cases.

That is why I am playing a 5 ply tournament currently, it is as close as I can get to equivalent speed or search depth, but it's not perfect, it handicaps the engines to some degree, but still it is a lot of fun to see the oldies play against the newbies. To see the different styles of play.

I want to see the improvement from where I can measure and enjoy it from the bottom.

regards

Nick

[yanquis1972]

isn't there some interface that allows this? i know people have mentioned it.