Chessmaster blitz - CM11 Default 2 cpu

[Chris Tatham]

Hi All,

CM11 Default 2CPU has completed the gauntlet and unsurprisingly is the new No1 finishing on 644/1000.

I said I would retire the gauntlet after the 60% mark was breached and certainly CM on 2cpu has 'outgrown' the original set of opponents which were designed for the defaults of earlier CM versions.

As there have been 25 cycles, there is also a nice symmetry as all the opponents have also played 1000 games - so this time, I have included the final integrated table of CM and gauntlet opponents:

Code: Select all

 Program                          Elo    +   -   Games   Score   Av.Op.  Draws

  1 Rybka 1.0 Beta 32-bit          : 2752   19  19  1000    68.4 %   2618   27.6 %
  2 Toga II 1.1a                   : 2724   19  19  1000    64.8 %   2618   27.4 %
  3 Fruit 2.2	                   : 2720   19  19  1000    64.2 %   2618   27.0 %
  4 Shredder 9                     : 2709   19  19  1000    62.8 %   2618   27.5 %
  5 CM11 Default 2cpu              : 2686   19  19  1000    64.4 %   2583   28.4 %
  6 Fritz 7                        : 2683   19  19  1000    59.3 %   2618   27.2 %
  7 CM9000 Tribute                 : 2651   19  19  1000    59.7 %   2583   27.7 %
  8 CM10th Paralyse                : 2648   19  18  1000    59.4 %   2583   27.9 %
  9 CM9000 Enforcer                : 2647   18  18  1000    59.2 %   2583   30.8 %
 10 CM9000 Wizard                  : 2642   18  18  1000    58.5 %   2583   29.4 %
 11 CM9000 Hero                    : 2639   19  19  1000    58.1 %   2583   26.6 %
 12 CM10th Master Yoda+            : 2634   18  18  1000    57.4 %   2583   29.3 %
 13 CM10th Thorondor               : 2634   18  18  1000    57.4 %   2583   28.7 %
 14 CM9000 Apex                    : 2627   18  18  1000    56.4 %   2583   30.5 %
 15 CM10th D2Alos                  : 2625   18  18  1000    56.1 %   2583   34.2 %
 16 CM10th Berean 5.54             : 2624   18  18  1000    56.0 %   2583   29.3 %
 17 CM9000 Attakiewicz             : 2624   18  18  1000    56.0 %   2583   33.1 %
 18 CM11 Default                   : 2624   18  18  1000    55.9 %   2583   29.5 %
 19 CM10th Ogre                    : 2619   18  18  1000    55.2 %   2583   31.1 %
 20 CM10th MrSpock                 : 2617   18  18  1000    54.9 %   2583   29.5 %
 21 CM9000 Grizzly2                : 2617   19  19  1000    54.9 %   2583   26.0 %
 22 CM10th Milan 2.3               : 2607   18  18  1000    53.5 %   2583   32.8 %
 23 List 512                       : 2607   18  18  1000    48.4 %   2618   30.7 %
 24 Spike 1.0 Mainz                : 2603   18  18  1000    47.9 %   2618   32.1 %
 25 CM10th SoFar2                  : 2602   17  17  1000    52.8 %   2583   36.3 %
 26 CM10th WoDra+                  : 2600   18  18  1000    52.5 %   2583   26.7 %
 27 CM9000 Default                 : 2600   18  18  1000    52.5 %   2583   27.8 %
 28 Pro Deo 1.1                    : 2596   18  18  1000    46.9 %   2618   30.7 %
 29 Hiarcs 8                       : 2595   18  18  1000    46.8 %   2618   32.7 %
 30 Ktulu 7.0                      : 2594   18  18  1000    46.6 %   2618   28.2 %
 31 CM10th Schumacher              : 2591   19  19  1000    51.2 %   2583   23.5 %
 32 CM10th ChesGek                 : 2591   19  19  1000    51.2 %   2583   23.5 %
 33 CM10th Default                 : 2582   19  19  1000    49.9 %   2583   26.2 %
 34 SlowChess Blitz WV2            : 2581   18  18  1000    44.8 %   2618   30.3 %
 35 Pseudo 0.7c                    : 2571   18  18  1000    43.2 %   2618   28.2 %
 36 Naum 1.82                      : 2561   18  18  1000    41.8 %   2618   31.2 %
 37 CM10th Vegeta 2d               : 2555   19  19  1000    46.1 %   2583   25.7 %
 38 Glaurung 051213                : 2553   18  18  1000    40.8 %   2618   28.4 %
 39 CM8000 Default                 : 2553   18  18  1000    45.8 %   2583   27.4 %
 40 Jonny 2.82                     : 2551   19  19  1000    40.4 %   2618   25.0 %
 41 Pharaon 3.3                    : 2547   18  18  1000    39.9 %   2618   28.9 %
 42 Aristarch 4.50                 : 2545   18  18  1000    39.6 %   2618   30.1 %
 43 ETChess 18.10.05               : 2536   18  18  1000    38.4 %   2618   32.6 %
 44 Thinker 4.7a                   : 2536   18  18  1000    38.4 %   2618   31.1 %
 45 Delfi 4.6                      : 2532   19  19  1000    37.8 %   2618   26.6 %
 46 Little Goliath Evolution       : 2529   18  18  1000    37.5 %   2618   31.4 %
 47 Zappa 1.1                      : 2515   19  19  1000    35.5 %   2618   28.1 %
 48 Anaconda 2.0.1                 : 2513   18  18  1000    35.3 %   2618   30.2 %
 49 Ufim 7.01                      : 2484   19  19  1000    31.6 %   2618   27.1 %
 50 The Baron 1.7.0                : 2420   21  21  1000    24.2 %   2618   21.6 %

For CM11 Default 2cpu at least under my test conditions I make it around 100 elo improvement over CM10th Default (c40 elo - new engine ; c60 elo extra processor).

I guess if someone can replicate the improvement of the best setting I've tested (as in greatest improvement over the respective default - CM10th Paralyse) then CM11 could just about finally challenge the original Rybka!

Best wishes


Chris

[Tony Thomas]

I am glad to see that the King still knows how to play chess, with all these young boys running around playing chess. If you stop your testing do you think all these people are going to invent settings that will enable it to gain about 50 points? I wonder if any of the old settings are good for the new version..If there arent any updates for another two years, you should continue your list with all the settings available at the time.

[ozziejoe]

It is possible that the CM team saw the testing of CM10, and then made the best settings in CM 10 the default settings in CM11. If so, it might have made little if any improvement on a single processor (with the dual processor presumably leading to 40 or 50 pnts). Or, to put this another way, maybe the correct comparison is cm10 paralysis versus cm11 default? (if not paralysis, maybe some other good configuration of cm10 or cm9)

best
J

[Chris Tatham]

Hi Tony / Joseph

Thanks for your interest!

It is possible that the CM team saw the testing of CM10, and then made the best settings in CM 10 the default settings in CM11

I wondered the same, but I understand from Cock that this wasn't the case.

If you stop your testing do you think all these people are going to invent settings that will enable it to gain about 50 points? I wonder if any of the old settings are good for the new version..

There was an active community of tweakers long before I started these specific tests and I am sure it will continue. I've always been impressed that almost every setting I've tested has been better than the default!

I'm no expert but I don't think any of the old settings are directly transferable with any degree of reliability.

If there arent any updates for another two years, you should continue your list with all the settings available at the time.

thanks for your vote of support but I think this would probably finish me!

Best wishes


Chris

[Graham Banks]

I'm no expert but I don't think any of the old settings are directly transferable with any degree of reliability.

Each Chessmaster engine is different, so I certainly wouldn't start by copying the same parameters to a different engine.

The best way to start is to watch or play through many Chessmaster 11 games to get a feel for its style.
You will then get an impression as to how well it handles the various aspects of play that you are able to adjust.
Then the experimenting begins in earnest.

You also need look at what other "tweakers" are doing!

Regards, Graham.

[ozziejoe]

great. I will receive chessmaster fromt he U.K. soon, and look forward to joing the CM tweaker community.

[Chris Tatham]

The best way to start is to watch or play through many Chessmaster 11 games to get a feel for its style.
You will then get an impression as to how well it handles the various aspects of play that you are able to adjust.
Then the experimenting begins in earnest.

Hi Graham

This is good advice - are you planning any settings for the new CM or is the champion tweaker now an ex-tweaker!?

Best wishes


Chris

[ozziejoe]

Here is an idea for those of you who can run computer versus computer tourni (i.e., an engine gets its own computer). Run cm11 default against the various cm10 personalities and see which personality cm11 correlates most highly with (maybe in terms of ponder hit).

If i was a lazy programer, I would simply make the discovered best parameters in cm10 the default parameters in cm11, and pretend i had made an actual improvement. why not?

did the cm programer take the lazy option? Perhaps he would rather be chasing women then working on that one last tweak in the search?

best
J

[Graham Banks]

The best way to start is to watch or play through many Chessmaster 11 games to get a feel for its style.
You will then get an impression as to how well it handles the various aspects of play that you are able to adjust.
Then the experimenting begins in earnest.

Hi Graham

This is good advice - are you planning any settings for the new CM or is the champion tweaker now an ex-tweaker!?

Best wishes


Chris

Eventually I'll give it a go Chris!

Cock de Gorter, Wilhelm Hudetz and others are just as good at "tweaking" the settings.

Regards, Graham.