SCCT Rating List - Calculation by EloStat 1.3

Discussion of computer chess matches and engine tournaments.

Moderator: Ras

Sedat Canbaz
Posts: 3018
Joined: Thu Mar 09, 2006 11:58 am
Location: Antalya/Turkey

Re: SCCT Rating List - Calculation by EloStat 1.3

Post by Sedat Canbaz »

And here is another proof:

Actually its not necessary to use especially the same database up to 29750 games (games up to 27.08.2012)

Even the latest SCCT database (30408 games) produce same BayesElo results with:
http://www.sedatcanbaz.com/chess/games/scct_3m2s_2.rar


readpgn scct.pgn

elo

mm

exactdist

offset 3300 Critter 1.6 x64 6c

ratings >18.txt

Code: Select all

Rank Name                          Elo    +    - games score oppo. draws 
   1 Houdini 2.0t3 Pro x64 6c     3359   14   14  1735   70%  3215   39% 
   2 Houdini 2.0t3* Pro x64 6c    3358   19   19  1000   75%  3184   37% 
   3 Houdini 2.0z Pro x64 6c      3355   15   15  1600   71%  3201   36% 
   4 Houdini 2.0s2 Pro x64 6c     3354   19   19  1000   74%  3178   34% 
   5 Houdini 1.5a x64 6c          3342   17   17  1100   68%  3217   41% 
   6 Houdini 2.0Bar2 x64 6c       3341   18   18  1050   72%  3189   44% 
   7 Houdini 2.0c Pro x64 6c      3340   15   15  1500   71%  3192   39% 
   8 Houdini 2.0Higgs Pro x64 6c  3337   18   18  1050   70%  3197   42% 
   9 Houdini2Bar1 Pro x64 6c      3328   17   17  1100   69%  3199   46% 
  10 Critter 1.6 x64 6c           3300   13   13  1935   63%  3213   53% 
  11 Critter 1.4 x64 6c           3289   16   16  1200   66%  3176   47% 
  12 Rybka 4.1 79DT v1 x64 6c     3287   17   17  1134   66%  3175   38% 
  13 Stockfish 120430P x64 6c     3282   13   13  1884   60%  3212   50% 
  14 Ivanhoe B46fE.02 x64 6c      3276   13   13  1935   59%  3214   52% 
  15 Rybka 4.1 SSE42 x64 6c       3276   13   13  1800   59%  3212   49% 
  16 Ivanhoe B46fC x64 6c         3275   16   16  1250   63%  3184   48% 
  17 Stockfish 2.2.2 JA x64 6c    3275   16   16  1200   62%  3191   47% 
  18 Rybka 4.1 NO-SSE x64 6c      3274   14   14  1500   60%  3203   49% 
  19 Fire 2.2 xTreme x64 6c       3263   13   12  1935   57%  3214   52% 
  20 Stockfish VE09 x64 6c        3262   17   17  1000   63%  3177   48% 
  21 Vitruvius 1.11C x64 6c       3261   12   12  1934   57%  3214   51% 
  22 Gull II beta2 x64 6c         3214   14   14  1435   51%  3208   51% 
  23 Strelka 5.5 x64 1c           3197   13   14  1684   45%  3227   48% 
  24 Bouquet 1.4 x64 6c           3184   15   15  1285   46%  3205   47% 
  25 Naum 4.2 x64 6c              3177   13   13  1935   44%  3216   44% 
  26 Komodo 4.0 x64 1c            3159   13   13  1935   41%  3217   42% 
  27 Deep Hiarcs 14 WCSC w32 6c   3156   21   21   658   45%  3192   44% 
  28 Deep Fritz 13 w32 6c         3128   13   13  1935   37%  3218   44% 
  29 Equinox 1.35 x64 6c          3128   14   15  1550   40%  3193   40% 
  30 Spike 1.4 Leiden w32 6c      3109   13   13  1934   34%  3218   38% 
  31 Chiron 1.1a x64 6c           3107   13   13  1935   34%  3218   39% 
  32 Deep Fritz 12 w32 6c         3092   16   16  1200   36%  3184   42% 
  33 Deep Junior 13.3 x64 6c      3091   14   14  1735   31%  3225   36% 
  34 Protector 1.4.0 x64 6c       3086   13   14  1934   31%  3219   37% 
  35 Spark 1.0 x64 6c             3083   14   14  1884   31%  3216   39% 
  36 Deep Junior 13 x64 6c        3081   16   16  1300   35%  3188   36% 
  37 Deep Shredder 12 x64 6c      3080   13   14  1935   30%  3219   37% 
  38 Hiarcs 13.2 w32 6c           3063   14   14  1900   29%  3220   32% 
  39 Zappa Mexico II x64 6c       3052   15   15  1600   29%  3205   34% 
  40 Fruit 090705 x64 6c          2980   18   19  1200   23%  3189   29%
Best,
Sedat
Top
Sedat Canbaz
Posts: 3018
Joined: Thu Mar 09, 2006 11:58 am
Location: Antalya/Turkey

Re: SCCT Rating List - Calculation by EloStat 1.3

Post by Sedat Canbaz »

My final words over this issue

Just to make it more clear,
BayesElo seems to be an accurate great tool as Ordo and Elostat

After all,i noticed that:
-The BayesElo Elo difference is appearing in case of changing the default values from mm to mm 1 1 or mm 0 1
*Thats why i noticed Fruit +16 Elo higher Elo performance than the previous BayesElo calculations
*Frankly:i did not expect to see +16 Elo BayesElo difference,just with changing from mm to mm 1 1

So...here is another BayesElo calculation with default mm
*Note:the below list is created without Rybka 4.1 NO-SSE and Hiarcs 14 games
Note also that Fruit's BayesElo Elo is falling down too, approx. 3 Elo after 50 games (against Rybka)

Code: Select all

Rank Name                          Elo    +    - games score oppo. draws 
   1 Houdini 2.0t3 Pro x64 6c     3360   14   14  1650   70%  3217   39% 
   2 Houdini 2.0t3* Pro x64 6c    3360   19   19  1000   75%  3186   37% 
   3 Houdini 2.0z Pro x64 6c      3359   15   15  1550   71%  3201   36% 
   4 Houdini 2.0s2 Pro x64 6c     3356   19   19  1000   74%  3180   34% 
   5 Houdini 1.5a x64 6c          3344   18   18  1050   68%  3217   40% 
   6 Houdini 2.0c Pro x64 6c      3344   16   15  1450   71%  3191   39% 
   7 Houdini 2.0Bar2 x64 6c       3342   18   18  1000   73%  3186   43% 
   8 Houdini 2.0Higgs Pro x64 6c  3338   18   18  1000   71%  3195   42% 
   9 Houdini2Bar1 Pro x64 6c      3329   17   17  1100   69%  3201   46% 
  10 Critter 1.6 x64 6c           3300   13   13  1850   63%  3215   53% 
  11 Critter 1.4 x64 6c           3291   17   16  1150   67%  3174   47% 
  12 Rybka 4.1 79DT v1 x64 6c     3287   17   17  1100   66%  3177   38% 
  13 Stockfish 120430P x64 6c     3285   13   13  1800   60%  3214   50% 
  14 Ivanhoe B46fE.02 x64 6c      3278   13   13  1850   59%  3215   52% 
  15 Rybka 4.1 SSE42 x64 6c       3278   13   13  1750   60%  3212   48% 
  16 Ivanhoe B46fC x64 6c         3277   16   16  1200   64%  3182   47% 
  17 Stockfish 2.2.2 JA x64 6c    3276   16   16  1200   62%  3193   47% 
  18 Fire 2.2 xTreme x64 6c       3264   13   13  1850   57%  3216   51% 
  19 Stockfish VE09 x64 6c        3264   17   17  1000   63%  3179   48% 
  20 Vitruvius 1.11C x64 6c       3263   13   13  1850   57%  3216   51% 
  21 Gull II beta2 x64 6c         3217   15   15  1350   51%  3209   51% 
  22 Strelka 5.5 x64 1c           3199   14   14  1600   45%  3228   48% 
  23 Bouquet 1.4 x64 6c           3186   16   16  1200   47%  3205   47% 
  24 Naum 4.2 x64 6c              3179   13   13  1850   44%  3218   44% 
  25 Komodo 4.0 x64 1c            3161   13   13  1850   41%  3218   42% 
  26 Equinox 1.35 x64 6c          3130   14   15  1550   40%  3195   40% 
  27 Deep Fritz 13 w32 6c         3129   13   13  1850   37%  3219   43% 
  28 Spike 1.4 Leiden w32 6c      3111   14   14  1850   34%  3220   39% 
  29 Chiron 1.1a x64 6c           3109   13   14  1850   34%  3220   39% 
  30 Deep Fritz 12 w32 6c         3095   17   17  1150   37%  3182   42% 
  31 Deep Junior 13.3 x64 6c      3094   14   15  1650   31%  3227   36% 
  32 Protector 1.4.0 x64 6c       3085   14   14  1850   31%  3220   36% 
  33 Spark 1.0 x64 6c             3085   14   14  1800   31%  3217   39% 
  34 Deep Junior 13 x64 6c        3083   16   16  1300   35%  3190   36% 
  35 Deep Shredder 12 x64 6c      3081   14   14  1850   30%  3221   36% 
  36 Hiarcs 13.2 w32 6c           3066   14   14  1850   29%  3221   32% 
  37 Zappa Mexico II x64 6c       3052   15   15  1550   29%  3205   34% 
  38 Fruit 090705 x64 6c          2984   18   19  1150   23%  3187   29%

Best,
Sedat
Top
Daniel Shawul
Posts: 4186
Joined: Tue Mar 14, 2006 11:34 am
Location: Ethiopia

Re: SCCT Rating List - Calculation by EloStat 1.3

Post by Daniel Shawul »

Sedat,
Guilt is a terrible feeling. You should have checked what you did before you made such ridiclous claims as to even announce a change of rating tools. It is your choice to use what you want, but you don't have to downplay a very good tool that does far better than elostat and ordo to justify it. You failed miserably at that.
Just to make it more clear,
BayesElo seems to be an accurate great tool as Ordo and Elostat
In time you will see it is much better. But atleast it is progress now ..
After all,i noticed that:
-The BayesElo Elo difference is appearing in case of changing the default values from mm to mm 1 1 or mm 0 1
*Thats why i noticed Fruit +16 Elo higher Elo performance than the previous BayesElo calculations
*Frankly:i did not expect to see +16 Elo BayesElo difference,just with changing from mm to mm 1 1
Again you are wrong. With or without I still get small values for bayeselo. My guess is you mixed up somewhere, using scale for one and not for the other, using mm for one and mm 1 1 for the other, and many other reasons I can't think of. You should have been consistent when you comparing results with and without the fruit games. Otherwise it is comparing apples and oranges. It is not an unforgivable fault but admit that you made a mistake and move on.
So...here is another BayesElo calculation with default mm
*Note:the below list is created without Rybka 4.1 NO-SSE and Hiarcs 14 games
Note also that Fruit's BayesElo Elo is falling down too, approx. 3 Elo after 50 games (against Rybka)
Doing that doesn't make sense at all. You want to compare only changes due to addition of fruit games.

Here is the result with mm i.e using default values of white advantage and draw elo. Similar result now difference of +3 elo from +1elo

Code: Select all

Diff1 = 56 - -235 =291
Diff2 = 59 - -235 =294
Increment = 3 elo
Detail

Code: Select all

version 0056, Copyright (C) 1997-2007 Remi Coulom.
compiled Jan 30 2007 20:30:07.
This program comes with ABSOLUTELY NO WARRANTY.
This is free software, and you are welcome to redistribute it
under the terms and conditions of the GNU General Public License.
See http://www.gnu.org/copyleft/gpl.html for details.
ResultSet>read scct1.pgn
Unknown command: read
type '?' for help
ResultSet>readpgn scct1.pgn
29250 game(s) loaded, 0 game(s) with unknown result ignored.
ResultSet>mm
Unknown command: mm
type '?' for help
ResultSet>elo
ResultSet-EloRating>mm
00:00:00,00
ResultSet-EloRating>ratings
Rank Name                          Elo    +    - games score oppo. draws
   1 Houdini 2.0t3 Pro x64 6c      142   14   14  1700   70%     0   39%
   2 Houdini 2.0t3* Pro x64 6c     142   19   19  1000   75%   -32   37%
   3 Houdini 2.0z Pro x64 6c       140   15   15  1550   71%   -17   36%
   4 Houdini 2.0s2 Pro x64 6c      138   19   19  1000   74%   -39   34%
   5 Houdini 2.0c Pro x64 6c       125   16   16  1450   71%   -27   39%
   6 Houdini 1.5a x64 6c           125   17   17  1100   68%     1   41%
   7 Houdini 2.0Bar2 x64 6c        124   18   18  1000   73%   -32   43%
   8 Houdini 2.0Higgs Pro x64 6c   119   18   18  1000   71%   -23   42%
   9 Houdini2Bar1 Pro x64 6c       111   17   17  1100   69%   -17   46%
  10 Critter 1.6 x64 6c             83   13   13  1900   63%    -2   53%
  11 Critter 1.4 x64 6c             73   17   17  1150   67%   -44   47%
  12 Rybka 4.1 79DT v1 x64 6c       69   17   17  1100   66%   -41   38%
  13 Stockfish 120430P x64 6c       67   13   13  1850   60%    -3   50%
  14 Deep Rybka 4.1 x64 6c          59   13   13  1750   60%    -7   48%
  15 Ivanhoe B46fE.02 x64 6c        59   13   13  1900   59%    -2   53%
  16 Ivanhoe B46fC x64 6c           59   16   16  1200   64%   -36   47%
  17 Stockfish 2.2.2 JA x64 6c      58   16   16  1200   62%   -25   47%
  18 Rybka 4.1 NO-SSE x64 6c        56   18   18  1000   63%   -29   49%
  19 Fire 2.2 xTreme x64 6c         46   13   13  1900   57%    -1   52%
  20 Stockfish VE09 x64 6c          45   18   18  1000   63%   -39   48%
  21 Vitruvius 1.11C x64 6c         44   13   13  1900   56%    -1   51%
  22 Gull II beta2 x64 6c           -3   15   15  1400   50%    -7   51%
  23 Strelka 5.5 x64 1c            -20   14   14  1650   45%    12   48%
  24 Bouquet 1.4 x64 6c            -32   16   16  1250   46%   -10   47%
  25 Naum 4.2 x64 6c               -39   13   13  1900   44%     1   44%
  26 Komodo 4.0 x64 1c             -58   13   13  1900   41%     2   42%
  27 Equinox 1.35 x64 6c           -88   14   14  1550   40%   -23   40%
  28 Deep Fritz 13 w32 6c          -88   13   13  1900   36%     2   43%
  29 Spike 1.4 Leiden w32 6c      -107   13   13  1900   34%     3   38%
  30 Chiron 1.1a x64 6c           -109   13   13  1900   34%     3   39%
  31 Deep Fritz 12 w32 6c         -123   17   17  1150   37%   -36   42%
  32 Deep Junior 13.3 x64 6c      -126   15   15  1700   31%    10   36%
  33 Protector 1.4.0 x64 6c       -130   14   14  1900   31%     3   36%
  34 Spark 1.0 x64 6c             -133   14   14  1850   31%     0   39%
  35 Deep Junior 13 x64 6c        -135   16   16  1300   35%   -28   36%
  36 Deep Shredder 12 x64 6c      -137   14   14  1900   30%     4   37%
  37 Hiarcs 13.2 w32 6c           -154   14   14  1900   29%     4   32%
  38 Zappa Mexico II x64 6c       -166   15   15  1550   29%   -13   34%
  39 Fruit 090705 x64 6c          -235   19   19  1150   23%   -31   29%
ResultSet-EloRating>

version 0056, Copyright (C) 1997-2007 Remi Coulom.
compiled Jan 30 2007 20:30:07.
This program comes with ABSOLUTELY NO WARRANTY.
This is free software, and you are welcome to redistribute it
under the terms and conditions of the GNU General Public License.
See http://www.gnu.org/copyleft/gpl.html for details.
ResultSet>readpgn scct2.pgn
30408 game(s) loaded, 0 game(s) with unknown result ignored.
ResultSet>elo
ResultSet-EloRating>mm
00:00:00,01
ResultSet-EloRating>ratings
Rank Name                          Elo    +    - games score oppo. draws
   1 Houdini 2.0t3 Pro x64 6c      144   14   14  1735   70%     0   39%
   2 Houdini 2.0t3* Pro x64 6c     143   19   19  1000   75%   -31   37%
   3 Houdini 2.0z Pro x64 6c       140   15   15  1600   71%   -14   36%
   4 Houdini 2.0s2 Pro x64 6c      139   19   19  1000   74%   -37   34%
   5 Houdini 1.5a x64 6c           127   17   17  1100   68%     2   41%
   6 Houdini 2.0Bar2 x64 6c        126   18   18  1050   72%   -26   44%
   7 Houdini 2.0c Pro x64 6c       125   15   15  1500   71%   -23   39%
   8 Houdini 2.0Higgs Pro x64 6c   122   18   18  1050   70%   -18   42%
   9 Houdini2Bar1 Pro x64 6c       113   17   17  1100   69%   -16   46%
  10 Critter 1.6 x64 6c             85   13   13  1935   63%    -2   53%
  11 Critter 1.4 x64 6c             74   16   16  1200   66%   -39   47%
  12 Rybka 4.1 79DT v1 x64 6c       72   17   17  1134   66%   -40   38%
  13 Stockfish 120430P x64 6c       67   13   13  1884   60%    -3   50%
  14 Ivanhoe B46fE.02 x64 6c        61   12   12  1935   59%    -1   52%
  15 Rybka 4.1 SSE42 x64 6c         61   13   13  1800   59%    -3   49%
  16 Ivanhoe B46fC x64 6c           60   16   16  1250   63%   -31   48%
  17 Stockfish 2.2.2 JA x64 6c      60   16   16  1200   62%   -24   47%
  18 Rybka 4.1 NO-SSE x64 6c        59   14   14  1500   60%   -12   49%
  19 Fire 2.2 xTreme x64 6c         48   12   12  1935   57%    -1   52%
  20 Stockfish VE09 x64 6c          47   18   18  1000   63%   -37   48%
  21 Vitruvius 1.11C x64 6c         46   12   12  1934   57%    -1   51%
  22 Gull II beta2 x64 6c           -1   14   14  1435   51%    -7   51%
  23 Strelka 5.5 x64 1c            -18   13   13  1684   45%    12   48%
  24 Bouquet 1.4 x64 6c            -31   15   15  1285   46%   -10   47%
  25 Naum 4.2 x64 6c               -38   13   13  1935   44%     1   44%
  26 Komodo 4.0 x64 1c             -56   13   13  1935   41%     2   42%
  27 Deep Hiarcs 14 WCSC w32 6c    -59   22   22   658   45%   -23   44%
  28 Deep Fritz 13 w32 6c          -87   13   13  1935   37%     3   44%
  29 Equinox 1.35 x64 6c           -87   14   14  1550   40%   -22   40%
  30 Spike 1.4 Leiden w32 6c      -106   13   13  1934   34%     3   38%
  31 Chiron 1.1a x64 6c           -108   13   13  1935   34%     3   39%
  32 Deep Fritz 12 w32 6c         -123   17   17  1200   36%   -31   42%
  33 Deep Junior 13.3 x64 6c      -124   14   14  1735   31%    11   36%
  34 Protector 1.4.0 x64 6c       -129   14   14  1934   31%     4   37%
  35 Spark 1.0 x64 6c             -132   14   14  1884   31%     1   39%
  36 Deep Junior 13 x64 6c        -134   16   16  1300   35%   -27   36%
  37 Deep Shredder 12 x64 6c      -135   14   14  1935   30%     4   37%
  38 Hiarcs 13.2 w32 6c           -152   14   14  1900   29%     6   32%
  39 Zappa Mexico II x64 6c       -163   15   15  1600   29%   -10   34%
  40 Fruit 090705 x64 6c          -235   19   19  1200   23%   -26   29%
ResultSet-EloRating>
Top
Sedat Canbaz
Posts: 3018
Joined: Thu Mar 09, 2006 11:58 am
Location: Antalya/Turkey

Re: SCCT Rating List - Calculation by EloStat 1.3

Post by Sedat Canbaz »

Daniel Shawul wrote: Guilt is a terrible feeling.
Hello Daniel,

What are you talking about ?
I dont feel any regret or guilt for nothing...
Of course i respect all programers,that's why i decided to clear some things...

There were just some misunderstandings before
Once more i'd like to point that i did not know before that:
- There can be a such BIG Fruit Elo difference (+16 Elo ) between mm and mm 1 1
*Note also i could not find much difference between mm 1 1 and mm 0 1
That's why i did not expect to appear a such + 16 Elo difference between mm and mm 1 1 or mm 0 1
To be honest:i did not remember exactly (on 27.08.2012) which values i used for BayesElo calculations
But later i found the reason...

I asked you before why a such Elo difference is appeared and you could not replay to my question !!
Just because you (like me :) ) are far from to be an BayesElo expert !

So if you still believe that you are BayesElo expert...
I have a question to you:
The BayesElo ratings should be counted with mm or mm 0 1 or mm 1 1 ?

And i still repeat:
-In case of changing from mm 1 1 to default mm,we see Fruit with 16 Elo better performance

And now what ? maybe i should be go in prison ?
Just because i said that how we can trust to BayesElo in the next calculations ?
or
Maybe because i switched from BayesElo to use Ordo ?

Believe me i am very happy in using Ordo calculation program

So tell me please,where i am wrong ???

Code: Select all

1) BayesElo mm 1 1 calculation:Rybka 4.1 NO-SSE are included

Rank Name                          Elo    +    - games score oppo. draws 
   1 Houdini 2.0t3 Pro x64 6c     3363   12   12  1700   70%  3212   39% 
   2 Houdini 2.0t3* Pro x64 6c    3362   15   15  1000   75%  3177   37% 
   3 Houdini 2.0z Pro x64 6c      3357   12   12  1600   71%  3196   36% 
   4 Houdini 2.0s2 Pro x64 6c     3356   15   15  1000   74%  3171   34% 
   5 Houdini 2.0Bar2 x64 6c       3345   14   14  1050   72%  3182   44% 
   6 Houdini 1.5a x64 6c          3345   14   14  1100   68%  3212   41% 
   7 Houdini 2.0c Pro x64 6c      3341   12   12  1500   71%  3186   39% 
   8 Houdini 2.0Higgs Pro x64 6c  3340   14   14  1050   70%  3191   42% 
   9 Houdini2Bar1 Pro x64 6c      3329   14   14  1100   69%  3193   46% 
  10 Critter 1.6 x64 6c           3300   11   11  1900   63%  3209   53% 
  11 Critter 1.4 x64 6c           3289   13   13  1200   66%  3169   47% 
  12 Rybka 4.1 79DT v1 x64 6c     3287   14   14  1100   66%  3168   38% 
  13 Stockfish 120430P x64 6c     3282   11   10  1850   60%  3208   50% 
  14 Rybka 4.1 SSE42 x64 6c       3275   11   11  1800   59%  3206   49% 
  15 Stockfish 2.2.2 JA x64 6c    3274   13   13  1200   62%  3185   47% 
  16 Ivanhoe B46fE.02 x64 6c      3273   10   10  1900   59%  3210   53% 
  17 Ivanhoe B46fC x64 6c         3273   13   13  1250   63%  3177   48% 
  18 Rybka 4.1 NO-SSE x64 6c      3273   12   12  1500   60%  3197   49% 
  19 Stockfish VE09 x64 6c        3263   14   14  1000   63%  3170   48% 
  20 Fire 2.2 xTreme x64 6c       3260   10   10  1900   57%  3210   52% 
  21 Vitruvius 1.11C x64 6c       3258   10   10  1900   56%  3210   51% 
  22 Gull II beta2 x64 6c         3209   12   12  1400   50%  3204   51% 
  23 Strelka 5.5 x64 1c           3189   11   11  1650   45%  3224   48% 
  24 Bouquet 1.4 x64 6c           3177   13   13  1250   46%  3201   47% 
  25 Naum 4.2 x64 6c              3169   10   10  1900   44%  3213   44% 
  26 Komodo 4.0 x64 1c            3149   11   11  1900   41%  3213   42% 
  27 Equinox 1.35 x64 6c          3117   12   12  1550   40%  3187   40% 
  28 Deep Fritz 13 w32 6c         3117   11   11  1900   36%  3214   43% 
  29 Spike 1.4 Leiden w32 6c      3098   11   11  1900   34%  3215   38% 
  30 Chiron 1.1a x64 6c           3095   11   11  1900   34%  3215   39% 
  31 Deep Fritz 12 w32 6c         3079   13   14  1200   36%  3177   42% 
  32 Deep Junior 13.3 x64 6c      3078   12   12  1700   31%  3223   36% 
  33 Protector 1.4.0 x64 6c       3073   11   11  1900   31%  3215   36% 
  34 Spark 1.0 x64 6c             3070   11   11  1850   31%  3212   39% 
  35 Deep Junior 13 x64 6c        3068   13   13  1300   35%  3181   36% 
  36 Deep Shredder 12 x64 6c      3067   11   11  1900   30%  3215   37% 
  37 Hiarcs 13.2 w32 6c           3051   11   11  1900   29%  3216   32% 
  38 Zappa Mexico II x64 6c       3038   12   12  1600   29%  3199   34% 
  39 Fruit 090705 x64 6c          2964   15   15  1200   23%  3182   29% 


2) BayesElo default mm calculation:Rybka 4.1 NO-SSE and Hiarcs 14 games are included

Rank Name                          Elo    +    - games score oppo. draws 
   1 Houdini 2.0t3 Pro x64 6c     3359   14   14  1735   70%  3215   39% 
   2 Houdini 2.0t3* Pro x64 6c    3358   19   19  1000   75%  3184   37% 
   3 Houdini 2.0z Pro x64 6c      3355   15   15  1600   71%  3201   36% 
   4 Houdini 2.0s2 Pro x64 6c     3354   19   19  1000   74%  3178   34% 
   5 Houdini 1.5a x64 6c          3342   17   17  1100   68%  3217   41% 
   6 Houdini 2.0Bar2 x64 6c       3341   18   18  1050   72%  3189   44% 
   7 Houdini 2.0c Pro x64 6c      3340   15   15  1500   71%  3192   39% 
   8 Houdini 2.0Higgs Pro x64 6c  3337   18   18  1050   70%  3197   42% 
   9 Houdini2Bar1 Pro x64 6c      3328   17   17  1100   69%  3199   46% 
  10 Critter 1.6 x64 6c           3300   13   13  1935   63%  3213   53% 
  11 Critter 1.4 x64 6c           3289   16   16  1200   66%  3176   47% 
  12 Rybka 4.1 79DT v1 x64 6c     3287   17   17  1134   66%  3175   38% 
  13 Stockfish 120430P x64 6c     3282   13   13  1884   60%  3212   50% 
  14 Ivanhoe B46fE.02 x64 6c      3276   13   13  1935   59%  3214   52% 
  15 Rybka 4.1 SSE42 x64 6c       3276   13   13  1800   59%  3212   49% 
  16 Ivanhoe B46fC x64 6c         3275   16   16  1250   63%  3184   48% 
  17 Stockfish 2.2.2 JA x64 6c    3275   16   16  1200   62%  3191   47% 
  18 Rybka 4.1 NO-SSE x64 6c      3274   14   14  1500   60%  3203   49% 
  19 Fire 2.2 xTreme x64 6c       3263   13   12  1935   57%  3214   52% 
  20 Stockfish VE09 x64 6c        3262   17   17  1000   63%  3177   48% 
  21 Vitruvius 1.11C x64 6c       3261   12   12  1934   57%  3214   51% 
  22 Gull II beta2 x64 6c         3214   14   14  1435   51%  3208   51% 
  23 Strelka 5.5 x64 1c           3197   13   14  1684   45%  3227   48% 
  24 Bouquet 1.4 x64 6c           3184   15   15  1285   46%  3205   47% 
  25 Naum 4.2 x64 6c              3177   13   13  1935   44%  3216   44% 
  26 Komodo 4.0 x64 1c            3159   13   13  1935   41%  3217   42% 
  27 Deep Hiarcs 14 WCSC w32 6c   3156   21   21   658   45%  3192   44% 
  28 Deep Fritz 13 w32 6c         3128   13   13  1935   37%  3218   44% 
  29 Equinox 1.35 x64 6c          3128   14   15  1550   40%  3193   40% 
  30 Spike 1.4 Leiden w32 6c      3109   13   13  1934   34%  3218   38% 
  31 Chiron 1.1a x64 6c           3107   13   13  1935   34%  3218   39% 
  32 Deep Fritz 12 w32 6c         3092   16   16  1200   36%  3184   42% 
  33 Deep Junior 13.3 x64 6c      3091   14   14  1735   31%  3225   36% 
  34 Protector 1.4.0 x64 6c       3086   13   14  1934   31%  3219   37% 
  35 Spark 1.0 x64 6c             3083   14   14  1884   31%  3216   39% 
  36 Deep Junior 13 x64 6c        3081   16   16  1300   35%  3188   36% 
  37 Deep Shredder 12 x64 6c      3080   13   14  1935   30%  3219   37% 
  38 Hiarcs 13.2 w32 6c           3063   14   14  1900   29%  3220   32% 
  39 Zappa Mexico II x64 6c       3052   15   15  1600   29%  3205   34% 
  40 Fruit 090705 x64 6c          2980   18   19  1200   23%  3189   29% 


3) BayesElo default mm calculation:Without Rybka 4.1 NO-SSE and Hiarcs 14 games

Rank Name                          Elo    +    - games score oppo. draws
   1 Houdini 2.0t3 Pro x64 6c     3360   14   14  1650   70%  3217   39%
   2 Houdini 2.0t3* Pro x64 6c    3360   19   19  1000   75%  3186   37%
   3 Houdini 2.0z Pro x64 6c      3359   15   15  1550   71%  3201   36%
   4 Houdini 2.0s2 Pro x64 6c     3356   19   19  1000   74%  3180   34%
   5 Houdini 1.5a x64 6c          3344   18   18  1050   68%  3217   40%
   6 Houdini 2.0c Pro x64 6c      3344   16   15  1450   71%  3191   39%
   7 Houdini 2.0Bar2 x64 6c       3342   18   18  1000   73%  3186   43%
   8 Houdini 2.0Higgs Pro x64 6c  3338   18   18  1000   71%  3195   42%
   9 Houdini2Bar1 Pro x64 6c      3329   17   17  1100   69%  3201   46%
  10 Critter 1.6 x64 6c           3300   13   13  1850   63%  3215   53%
  11 Critter 1.4 x64 6c           3291   17   16  1150   67%  3174   47%
  12 Rybka 4.1 79DT v1 x64 6c     3287   17   17  1100   66%  3177   38%
  13 Stockfish 120430P x64 6c     3285   13   13  1800   60%  3214   50%
  14 Ivanhoe B46fE.02 x64 6c      3278   13   13  1850   59%  3215   52%
  15 Rybka 4.1 SSE42 x64 6c       3278   13   13  1750   60%  3212   48%
  16 Ivanhoe B46fC x64 6c         3277   16   16  1200   64%  3182   47%
  17 Stockfish 2.2.2 JA x64 6c    3276   16   16  1200   62%  3193   47%
  18 Fire 2.2 xTreme x64 6c       3264   13   13  1850   57%  3216   51%
  19 Stockfish VE09 x64 6c        3264   17   17  1000   63%  3179   48%
  20 Vitruvius 1.11C x64 6c       3263   13   13  1850   57%  3216   51%
  21 Gull II beta2 x64 6c         3217   15   15  1350   51%  3209   51%
  22 Strelka 5.5 x64 1c           3199   14   14  1600   45%  3228   48%
  23 Bouquet 1.4 x64 6c           3186   16   16  1200   47%  3205   47%
  24 Naum 4.2 x64 6c              3179   13   13  1850   44%  3218   44%
  25 Komodo 4.0 x64 1c            3161   13   13  1850   41%  3218   42%
  26 Equinox 1.35 x64 6c          3130   14   15  1550   40%  3195   40%
  27 Deep Fritz 13 w32 6c         3129   13   13  1850   37%  3219   43%
  28 Spike 1.4 Leiden w32 6c      3111   14   14  1850   34%  3220   39%
  29 Chiron 1.1a x64 6c           3109   13   14  1850   34%  3220   39%
  30 Deep Fritz 12 w32 6c         3095   17   17  1150   37%  3182   42%
  31 Deep Junior 13.3 x64 6c      3094   14   15  1650   31%  3227   36%
  32 Protector 1.4.0 x64 6c       3085   14   14  1850   31%  3220   36%
  33 Spark 1.0 x64 6c             3085   14   14  1800   31%  3217   39%
  34 Deep Junior 13 x64 6c        3083   16   16  1300   35%  3190   36%
  35 Deep Shredder 12 x64 6c      3081   14   14  1850   30%  3221   36%
  36 Hiarcs 13.2 w32 6c           3066   14   14  1850   29%  3221   32%
  37 Zappa Mexico II x64 6c       3052   15   15  1550   29%  3205   34%
  38 Fruit 090705 x64 6c          2984   18   19  1150   23%  3187   29%

Btw,i respect you as Author of Scorpio
But please dont waste my time more over this issue

And instead of spending more time over this thread
I suggest you to work more on improving Scorpio


Thats all...


Best,
Sedat
Top
Daniel Shawul
Posts: 4186
Joined: Tue Mar 14, 2006 11:34 am
Location: Ethiopia

Re: SCCT Rating List - Calculation by EloStat 1.3

Post by Daniel Shawul »

There were just some misunderstandings before
Once more i'd like to point that i did not know before that:
- There can be a such BIG Fruit Elo difference (+16 Elo ) between mm and mm 1 1
Are you daft , because there is no other explanation? I just showed you there is not much difference b/n using mm and mm 1 1. It is a +3elo for mm, a +1 elo for mm 1 1 and I did also for mm 0 1 (that I didn't post here) and the difference is still +1 elo

Code: Select all

Summary
mm  -> +3elo
mm 0 1 -> +1elo
mm 1 1 -> +1elo
You still have to show your results before and after fruit games are added. You keep on spamming with useless results that takes out hiarcs games, rybka games etc. You said when Fruit games are added only. Do that and show us the +16 elo, period Is that too much to ask ?
That's why i did not expect to appear a such + 16 Elo difference between mm and mm 1 1 or mm 0 1
To be honest:i did not remember exactly (on 27.08.2012) which values i used for BayesElo calculations
But later i found the reason...

I asked you before why a such Elo difference is appeared and you could nor replay to my question
Just because you (like me ) are far from to be an BayesElo expert
Only reason I threw you a bone (replied to your spam) is out of coutesy. Do show us the +16 elo now ... still waiting.
And i still repeat:
-In case of changing from mm 1 1 to default mm,we see Fruit with 16 Elo better performance
What an idiotic comment. That is because of the offset, was that your problem all along?? I didn't know you were this dubm not to understand elo offset. Elo is relative and you should compare it to another engine's rating. Bayeselo or elostat or whathave calculate a default offset even if you don't specifity it ... They try to center the elos in general which is why you see negative elos at first.
Maybe because i switched from BayesElo to use Ordo ?

Believe me i am very happy in using Ordo calculation program

So tell me please,where i am wrong ???
yes for the nth time. And i now even suspect you had even missed even more. Such a waste. Calculate elos with and without fruit games like you forced me to do and come bakc..
Top
Sedat Canbaz
Posts: 3018
Joined: Thu Mar 09, 2006 11:58 am
Location: Antalya/Turkey

Re: SCCT Rating List - Calculation by EloStat 1.3

Post by Sedat Canbaz »

Daniel Shawul wrote: Are you daft


Hey...be careful with your words,because i can not promise to be so kind against you

In reality:Who are you ?


Why you dont open your eyes a little bit and see/check the previous my postings

But anyway,as you are a VIP member i will post again:


readpgn scct.pgn

elo

mm

exactdist

offset 3300 Critter 1.6 x64 6c

ratings >18.txt

Code: Select all

Rank Name                          Elo    +    - games score oppo. draws
   1 Houdini 2.0t3 Pro x64 6c     3359   14   14  1735   70%  3215   39%
   2 Houdini 2.0t3* Pro x64 6c    3358   19   19  1000   75%  3184   37%
   3 Houdini 2.0z Pro x64 6c      3355   15   15  1600   71%  3201   36%
   4 Houdini 2.0s2 Pro x64 6c     3354   19   19  1000   74%  3178   34%
   5 Houdini 1.5a x64 6c          3342   17   17  1100   68%  3217   41%
   6 Houdini 2.0Bar2 x64 6c       3341   18   18  1050   72%  3189   44%
   7 Houdini 2.0c Pro x64 6c      3340   15   15  1500   71%  3192   39%
   8 Houdini 2.0Higgs Pro x64 6c  3337   18   18  1050   70%  3197   42%
   9 Houdini2Bar1 Pro x64 6c      3328   17   17  1100   69%  3199   46%
  10 Critter 1.6 x64 6c           3300   13   13  1935   63%  3213   53%
  11 Critter 1.4 x64 6c           3289   16   16  1200   66%  3176   47%
  12 Rybka 4.1 79DT v1 x64 6c     3287   17   17  1134   66%  3175   38%
  13 Stockfish 120430P x64 6c     3282   13   13  1884   60%  3212   50%
  14 Ivanhoe B46fE.02 x64 6c      3276   13   13  1935   59%  3214   52%
  15 Rybka 4.1 SSE42 x64 6c       3276   13   13  1800   59%  3212   49%
  16 Ivanhoe B46fC x64 6c         3275   16   16  1250   63%  3184   48%
  17 Stockfish 2.2.2 JA x64 6c    3275   16   16  1200   62%  3191   47%
  18 Rybka 4.1 NO-SSE x64 6c      3274   14   14  1500   60%  3203   49%
  19 Fire 2.2 xTreme x64 6c       3263   13   12  1935   57%  3214   52%
  20 Stockfish VE09 x64 6c        3262   17   17  1000   63%  3177   48%
  21 Vitruvius 1.11C x64 6c       3261   12   12  1934   57%  3214   51%
  22 Gull II beta2 x64 6c         3214   14   14  1435   51%  3208   51%
  23 Strelka 5.5 x64 1c           3197   13   14  1684   45%  3227   48%
  24 Bouquet 1.4 x64 6c           3184   15   15  1285   46%  3205   47%
  25 Naum 4.2 x64 6c              3177   13   13  1935   44%  3216   44%
  26 Komodo 4.0 x64 1c            3159   13   13  1935   41%  3217   42%
  27 Deep Hiarcs 14 WCSC w32 6c   3156   21   21   658   45%  3192   44%
  28 Deep Fritz 13 w32 6c         3128   13   13  1935   37%  3218   44%
  29 Equinox 1.35 x64 6c          3128   14   15  1550   40%  3193   40%
  30 Spike 1.4 Leiden w32 6c      3109   13   13  1934   34%  3218   38%
  31 Chiron 1.1a x64 6c           3107   13   13  1935   34%  3218   39%
  32 Deep Fritz 12 w32 6c         3092   16   16  1200   36%  3184   42%
  33 Deep Junior 13.3 x64 6c      3091   14   14  1735   31%  3225   36%
  34 Protector 1.4.0 x64 6c       3086   13   14  1934   31%  3219   37%
  35 Spark 1.0 x64 6c             3083   14   14  1884   31%  3216   39%
  36 Deep Junior 13 x64 6c        3081   16   16  1300   35%  3188   36%
  37 Deep Shredder 12 x64 6c      3080   13   14  1935   30%  3219   37%
  38 Hiarcs 13.2 w32 6c           3063   14   14  1900   29%  3220   32%
  39 Zappa Mexico II x64 6c       3052   15   15  1600   29%  3205   34%
  40 Fruit 090705 x64 6c          2980   18   19  1200   23%  3189   29%
Top
Daniel Shawul
Posts: 4186
Joined: Tue Mar 14, 2006 11:34 am
Location: Ethiopia

Re: SCCT Rating List - Calculation by EloStat 1.3

Post by Daniel Shawul »

Hey...be careful with your words,because i can not promise to be so kind against you

In reality:Who are you ?
Save the talk and answer your ridiclous claims. Why do you say mm and mm 1 1 brings a +16 elo difference when I clearly showed you it is a +3elo vs +1elo just in my previous post. You didn't show your result before and after fruit games are added. You still have to answer for that.
Why you dont open your eyes a little bit and see/check the previous my postings
I did and now I see you even have more problems. You simply don't seem to understnad what offset is. Let me ask you a question. In all your test critter is fixed at +3000 elo, do you conclude from that critter's perfromance is the same when games are added? The answer is ofcourse no because if you fix another engine critter's elo will change. That seems to be your problem because you were not comparing Fruit's elo agains Rybka No SE but just its absolute elo value of results of mm and mm 1 1. I told you many times to do the following without mixing:
a) mm without fruit games THEN mm with fruit games
b) mm 1 1 without fruit games THEN mm 1 1 with fruit games
That is exactly what I did and got a +1 vs +3 elo increment before and after additions of fuit games (i.e Fruit's elo relative to Rybka), and NOT mixing mm and mm 1 1 against each other.
Daniel
Top
marijan
Posts: 56
Joined: Mon Jan 16, 2012 1:16 am

Re: SCCT Rating List - Calculation by EloStat 1.3

Post by marijan »

First of all; I think Sedat doing a great job, and he don t need to explain himself... As SCCT director, Sedat got right to choose which program for calculation ELOs he will use...

For smart people thats enough....

Regards
Top
Sedat Canbaz
Posts: 3018
Joined: Thu Mar 09, 2006 11:58 am
Location: Antalya/Turkey

Re: SCCT Rating List - Calculation by EloStat 1.3

Post by Sedat Canbaz »

Hey Daniel,

I see you still use negative words over me,bravo to you !!

Who are you ?

What is your problem with me ??

Really you have a lot of negative complexes..

Do you have a family ? or maybe you have ethnic problems too ?

If you owner of Scorpio chess engine,it does not mean that you are star...
Or who knows... maybe your blood is white ?
And you have no any right to use a such foul language over me

In the first of your postings,
I thought that you came here on my thread to help as a friend
But now its quite clear your main reason about why you are here

And its quite clear too,
You don't like my works and my activities (i knew this from the past)
Plus you have 'zero' understanding to show respect to others work


And definitely you are not a expert (like me :) ) on BayesElo calculations
Simply just because my previous questions remained without answers:
-The BayesElo ratings should be counted with mm or mm 0 1 or mm 1 1 ?
-Can you explain me please,why in the 12.txt and 13.txt we see such strange results ?

A few notes more about BayesElo,
I used BayesElo since several years,that's why i am so thankful to Remi Coulom

But however,
BayesElo is still seems to me strange,just with changing the default values from mm to mm 1 1
And where we get with Fruit +16 Elo difference

Sorry...i dont want to waste more time over this issue

I wish you all the best,
Sedat
Top
Sedat Canbaz
Posts: 3018
Joined: Thu Mar 09, 2006 11:58 am
Location: Antalya/Turkey

Re: SCCT Rating List - Calculation by EloStat 1.3

Post by Sedat Canbaz »

marijan wrote:First of all; I think Sedat doing a great job, and he don t need to explain himself... As SCCT director, Sedat got right to choose which program for calculation ELOs he will use...

For smart people thats enough....

Regards
Dear Marijan,

Thanks a lot or understanding and support

Greetings,
Sedat
Top

Return to “Computer Chess Club: Tournaments and Matches”