ProDeo
Would you like to react to this message? Create an account in a few clicks or log in to continue.
ProDeo

Computer Chess
 
HomeHome  CalendarCalendar  Latest imagesLatest images  FAQFAQ  SearchSearch  MemberlistMemberlist  UsergroupsUsergroups  RegisterRegister  Log in  

 

 I wonder what is the reason that so many cores work for stockfish

Go down 
+4
Dio
Admin
TheSelfImprover
Uri Blass
8 posters
Go to page : Previous  1, 2, 3  Next
AuthorMessage
TheSelfImprover

TheSelfImprover


Posts : 3112
Join date : 2020-11-18

I wonder what is the reason that so many cores work for stockfish - Page 2 Empty
PostSubject: Re: I wonder what is the reason that so many cores work for stockfish   I wonder what is the reason that so many cores work for stockfish - Page 2 EmptyWed Jul 27, 2022 2:30 pm

pohl4711 wrote:
My UHO openings are selected out of the Megabase and all openings were played by strong human players. So, no openings could be more "normal"/human chess than UHO. And I did not change anything in these openings, I just used the KomodoDragon eval for selecting them. Thats why I called them UHO: Unbalanced Human Openings. All Openings were played by strong humans (all players must have 2300+ Elo. All games played after WW2, so no crazy gambits from 19th century or so). All games out of the Megabase.


Do top computers play these openings?

I suspect the answer will be something like, "only when made to" - so not when they're playing their best game.
Back to top Go down
TheSelfImprover

TheSelfImprover


Posts : 3112
Join date : 2020-11-18

I wonder what is the reason that so many cores work for stockfish - Page 2 Empty
PostSubject: Re: I wonder what is the reason that so many cores work for stockfish   I wonder what is the reason that so many cores work for stockfish - Page 2 EmptyWed Jul 27, 2022 2:40 pm

TheSelfImprover wrote:
pohl4711 wrote:
My UHO openings are selected out of the Megabase and all openings were played by strong human players. So, no openings could be more "normal"/human chess than UHO. And I did not change anything in these openings, I just used the KomodoDragon eval for selecting them. Thats why I called them UHO: Unbalanced Human Openings. All Openings were played by strong humans (all players must have 2300+ Elo. All games played after WW2, so no crazy gambits from 19th century or so). All games out of the Megabase.


Do top computers play these openings?

I suspect the answer will be something like, "only when made to" - so not when they're playing their best game.


Maybe an effect of using a "human" opening book is to amplify small differences in standard between different computers. If so, this could be a good thing which is worthy of praise. There would be a risk, though, that doing this is not "scientifically sound". Smile
Back to top Go down
Admin
Admin
Admin


Posts : 2608
Join date : 2020-11-17
Location : Netherlands

I wonder what is the reason that so many cores work for stockfish - Page 2 Empty
PostSubject: Re: I wonder what is the reason that so many cores work for stockfish   I wonder what is the reason that so many cores work for stockfish - Page 2 EmptyWed Jul 27, 2022 9:00 pm

This is the GRL without draws, I like it a lot for the top engines.

Code:

   # PLAYER                        :  RATING  ERROR  POINTS  PLAYED   (%)  CFS(%)     W    D     L  D(%)  OppAvg
   1 Stockfish_15                  :  4369.8   74.2  1315.0    1336    98      88  1315    0    21     0  3560.7
   2 Stockfish_14.1                :  4310.6   56.5  1863.0    1899    98      78  1863    0    36     0  3573.1
   3 Stockfish_21-10-06            :  4279.6   63.7  1212.0    1238    98      98  1212    0    26     0  3524.4
   4 Stockfish_14                  :  4200.2   49.3  2787.0    2890    96      99  2787    0   103     0  3469.1
   5 Stockfish_13                  :  4160.7   49.1  1915.0    2010    95      94  1915    0    95     0  3447.5
   6 Stockfish_21-05-18            :  4110.3   66.0   617.0     652    95      98   617    0    35     0  3392.3
   7 Komodo-Dragon_2.5             :  4047.0   19.1  1720.0    2043    84     100  1720    0   323     0  3653.3
   8 Stockfish_12                  :  3957.6   33.5  1222.0    1438    85     100  1222    0   216     0  3510.1
   9 CSTal_1.35-V20-E520           :  3875.6   34.2   415.0     560    74      60   415    0   145     0  3675.5
  10 Komodo-Dragon_2               :  3870.8   20.8  3337.0    4107    81      65  3337    0   770     0  3530.3
  11 Lc0_v28                       :  3860.0   54.3   335.0     487    69      63   335    0   152     0  3681.8
  12 Komodo-Dragon                 :  3845.8   50.8  1317.0    1637    80      61  1317    0   320     0  3484.8
  13 Komodo_Dragon                 :  3825.2  127.5    56.0      82    68      63    56    0    26     0  3693.8
  14 CSTal-NNUE_134-E195           :  3804.8   39.6   363.0     547    66      99   363    0   184     0  3676.3
  15 Koivisto_8.0                  :  3771.5   32.1   416.0    1010    41      98   416    0   594     0  3889.1
  16 Berserk_8.5.1                 :  3750.0   30.4   463.0     960    48      52   463    0   497     0  3796.4
  17 CSTal-NNUE_133-4              :  3748.9   34.7   324.0     543    60      84   324    0   219     0  3675.8
  18 CSTal-NNUE_1.30               :  3733.2   48.0   561.0     808    69      72   561    0   247     0  3566.6
  19 Lc0-v27                       :  3707.3   58.5   307.0     412    75      59   307    0   105     0  3488.5
  20 Ethereal_13.50-NNUE           :  3699.2   39.3   357.0     877    41      63   357    0   520     0  3815.6
  21 CSTal-NNUE_1.31               :  3693.8   52.1   127.0     310    41      65   127    0   183     0  3761.1
  22 Berserk_8                     :  3680.5   43.1   496.0     822    60      51   496    0   326     0  3604.0
  23 Chess-System-Tal-NNUE-1.26    :  3679.7   20.1   493.0     778    63      69   493    0   285     0  3574.1
  24 Stockfish_11                  :  3667.3   41.5   518.0     736    70      84   518    0   218     0  3494.6
  25 Koivisto_7.9                  :  3642.7   27.5   666.0    1263    53      64   666    0   597     0  3651.7
  26 Koivisto_7.5                  :  3635.5   34.9   575.0    1054    55      83   575    0   479     0  3640.4
  27 Stockfish_10                  :  3617.7   29.1  1233.0    2091    59      77  1233    0   858     0  3552.1
  28 Koivisto_7.2                  :  3607.2   21.4   553.0    1089    51      56   553    0   536     0  3654.7
  29 Koivisto_7.0                  :  3604.4   26.7   702.0    1323    53      91   702    0   621     0  3621.1
  30 SlowChess_2.8                 :  3587.6   14.7   865.0    2066    42      92   865    0  1201     0  3699.6
  31 RubiChess_20220223            :  3559.5   40.7   203.0     861    24      75   203    0   658     0  3846.2
  32 Ethereal_13.25-NNUE           :  3543.4   16.1  1366.0    3021    45      80  1366    0  1655     0  3633.6
  33 Berserk_7                     :  3535.6   22.7   461.0    1044    44      52   461    0   583     0  3586.7
  34 Seer_2.5.0                    :  3534.3   35.0   177.0     827    21      53   177    0   650     0  3846.9
  35 Koivisto_6.16                 :  3532.5   28.4   539.0    1466    37      53   539    0   927     0  3749.3
  36 Koivisto_6.23                 :  3531.2   33.0   511.0    1176    43      79   511    0   665     0  3664.3
  37 Stockfish_9                   :  3521.4   11.7   973.0    1683    58     100   973    0   710     0  3461.7
  38 Ethereal_13-NNUE              :  3495.7   22.6   826.0    1986    42      62   826    0  1160     0  3671.0
  39 SlowChess_2.7                 :  3491.9   24.6   628.0    1936    32      95   628    0  1308     0  3725.4
  40 SlowChess_2.6                 :  3472.9   26.2   817.0    2203    37      53   817    0  1386     0  3694.8
  41 Komodo_14                     :  3471.7   24.1   654.0    1573    42      75   654    0   919     0  3570.5
  42 Berserk_6                     :  3464.3   25.7   861.0    1403    61      91   861    0   542     0  3383.2
  43 Seer_2.4.0                    :  3444.1   39.2   478.0     747    64      55   478    0   269     0  3344.1
  44 RubiChess_2.2                 :  3441.8   18.2  1102.0    3221    34      61  1102    0  2119     0  3637.3
  45 Rebel_E240-T50                :  3436.7   36.1   287.0     508    56      61   287    0   221     0  3391.7
  46 Koivisto_6.0-sse2             :  3431.8   23.5   920.0    1782    52      86   920    0   862     0  3477.3
  47 Toga4_1.1                     :  3417.2   21.1   470.0     650    72      69   470    0   180     0  3248.2
  48 Komodo_12                     :  3407.2   25.7   653.0    1212    54      70   653    0   559     0  3381.8
  49 h6                            :  3400.0    8.5   956.0    2133    45      52   956    0  1177     0  3570.1
  50 Stockfish_8                   :  3399.2   28.5   735.0    1324    56      79   735    0   589     0  3361.4
  51 RubiChess_2.1                 :  3393.9   29.8   671.0    2528    27      64   671    0  1857     0  3727.2
  52 Rebel_E208                    :  3389.8   19.5   257.0     518    50      97   257    0   261     0  3394.3
  53 SlowChess_2.5                 :  3371.0   20.4   646.0    2037    32      81   646    0  1391     0  3600.5
  54 Arasan_23.0.1                 :  3355.8   19.2   411.0     629    65      60   411    0   218     0  3244.4
  55 Berserk_5                     :  3350.8   27.7   655.0    1110    59      58   655    0   455     0  3287.8
  56 Ethereal_12.75                :  3349.1   21.2   839.0    2637    32      53   839    0  1798     0  3578.3
  57 Rebel_14.2                    :  3347.6   35.0   391.0     612    64      53   391    0   221     0  3247.8
  58 Komodo_11                     :  3346.5   13.1   637.0    1294    49      52   637    0   657     0  3355.2
  59 Pedone_3.1                    :  3345.9   22.7  1454.0    4304    34      64  1454    0  2850     0  3542.4
  60 Arasan_23.2                   :  3341.6   19.9   386.0     804    48      88   386    0   418     0  3356.6
  61 Igel_3.0.5                    :  3331.7   18.3   865.0    3571    24      56   865    0  2706     0  3635.4
  62 Ethereal_12.25                :  3329.6   29.3   214.0     700    31      55   214    0   486     0  3638.8
  63 Rebel_15x2                    :  3327.4   24.8   218.0     528    41      53   218    0   310     0  3392.2
  64 Ethereal_12.50                :  3325.8   34.9   295.0     666    44      50   295    0   371     0  3422.9
  65 Rebel_15                      :  3325.7   35.6   223.0     548    41      80   223    0   325     0  3394.3
  66 SlowChess_2.4                 :  3307.0   32.9   267.0     719    37      74   267    0   452     0  3534.9
  67 Marvin_5.2                    :  3298.0   21.7   762.0    1199    64      56   762    0   437     0  3197.3
  68 Zahak_9.0                     :  3296.8   20.3   821.0    1308    63      51   821    0   487     0  3203.1
  69 Ethereal_12.00                :  3296.5   27.4   571.0    1476    39      52   571    0   905     0  3469.3
  70 Igel_3.0.0                    :  3295.5   51.5   111.0     534    21      52   111    0   423     0  3686.5
  71 Seer_2.3.0                    :  3294.2   24.0   558.0    1079    52      62   558    0   521     0  3285.0
  72 Koivisto_4.83                 :  3286.7   27.3   750.0    1059    71      58   750    0   309     0  3126.1
  73 Berserk_4.5.0                 :  3281.9   28.7   527.0    1055    50      57   527    0   528     0  3284.2
  74 Stockfish_7                   :  3279.5   19.1   386.0     726    53      58   386    0   340     0  3259.1
  75 Nemorino_6.00                 :  3276.2   23.8   888.0    2748    32      61   888    0  1860     0  3501.0
  76 Minic_3.17                    :  3270.5   24.2   546.0     895    61      60   546    0   349     0  3192.3
  77 Pedone_3.0                    :  3265.1   24.3   719.0    1439    50      50   719    0   720     0  3265.0
  78 Seer_2.2.0                    :  3265.0   19.8   488.0     900    54      54   488    0   412     0  3235.4
  79 Weiss_2.0                     :  3263.9   12.1   965.0    1813    53      72   965    0   848     0  3242.2
  80 Komodo_10                     :  3255.9   22.7   571.0    1378    41      76   571    0   807     0  3324.3
  81 Berserk_4.4.0                 :  3248.7   25.4   377.0     759    50      57   377    0   382     0  3252.3
  82 Stockfish_5                   :  3245.0   23.7  1006.0    1875    54      75  1006    0   869     0  3220.2
  83 Minic_3.16                    :  3232.0   27.5   451.0     779    58      54   451    0   328     0  3176.9
  84 Berserk_4.3.0                 :  3230.0   19.9   769.0    1496    51      54   769    0   727     0  3221.5
  85 Booot_6.5                     :  3229.4   11.6  1228.0    3951    31      59  1228    0  2723     0  3456.4
  86 rofChade_2.3                  :  3228.3   16.7  1285.0    3941    33      62  1285    0  2656     0  3434.4
  87 Stash_32.0                    :  3222.7   27.6   426.0     894    48      63   426    0   468     0  3243.4
  88 Clover_2.4                    :  3216.9   15.5  1181.0    2443    48      58  1181    0  1262     0  3231.1
  89 Danasah_9.0                   :  3213.0   37.1   617.0    1048    59      51   617    0   431     0  3149.7
  90 Wasp_5.00                     :  3212.8   21.3   362.0     727    50      57   362    0   365     0  3214.4
  91 Stockfish_6                   :  3210.1   24.3   397.0     909    44      72   397    0   512     0  3256.9
  92 Seer_2.1.0                    :  3203.2    8.4  1187.0    2409    49      54  1187    0  1222     0  3208.8
  93 Arasan_23.1                   :  3201.8   29.5   191.0     598    32      90   191    0   407     0  3335.8
  94 Minic_3.14                    :  3178.4   23.6   552.0    1104    50      51   552    0   552     0  3179.7
  95 Lzero_v28                     :  3177.9   13.8   511.0     975    52      54   511    0   464     0  3161.8
  96 Arasan_22.3                   :  3176.0   26.6   668.0    1047    64      55   668    0   379     0  3072.3
  97 Minic_3.13                    :  3174.3   17.0   819.0    1451    56      68   819    0   632     0  3128.4
  98 Weiss_1.4                     :  3167.9   17.2  1167.0    2443    48      67  1167    0  1276     0  3184.7
  99 Stash_31.0                    :  3162.9   19.0  1409.0    2702    52      55  1409    0  1293     0  3146.0
 100 Halogen_10                    :  3161.9   11.1  2043.0    4019    51      58  2043    0  1976     0  3155.2
 101 Toga_III-03.12                :  3158.5   36.4   559.0     968    58      64   559    0   409     0  3103.1
 102 Invictus_r382                 :  3150.2   45.9   272.0     390    70      50   272    0   118     0  3002.2
 103 Berserk_4.2.0                 :  3150.2   26.6   951.0    1661    57      50   951    0   710     0  3095.8
 104 Critter_1.6                   :  3150.0    8.5   907.0    1523    60      56   907    0   616     0  3079.2
 105 Beef_0.3.6                    :  3149.3   12.6   925.0    1786    52      52   925    0   861     0  3135.4
 106 Wasp_4.50                     :  3148.7   17.4  1430.0    3267    44      78  1430    0  1837     0  3195.0
 107 Seer_2.0.1                    :  3143.2   11.5  1350.0    2354    57      63  1350    0  1004     0  3087.8
 108 Demolito_2021-09-30           :  3139.1   21.2   458.0    1062    43      66   458    0   604     0  3189.1
 109 Combusken_1.4.0               :  3133.5   11.8   930.0    1673    56      74   930    0   743     0  3092.4
 110 Counter_4.0                   :  3126.8   17.6   483.0     922    52      55   483    0   439     0  3109.8
 111 Lzero_v27                     :  3125.2   14.1   655.0    1086    60      54   655    0   431     0  3048.2
 112 Winter_0.9                    :  3124.2   22.9   760.0    1399    54      57   760    0   639     0  3091.3
 113 Marvin_5.1                    :  3122.5   21.8  1144.0    2138    54      50  1144    0   994     0  3095.4
 114 Zahak_8.0                     :  3122.5   20.6   608.0    1441    42      77   608    0   833     0  3178.8
 115 Clover_2.3.1                  :  3115.1   15.3  1185.0    2200    54      76  1185    0  1015     0  3087.5
 116 Mr_Bob_1.1.0                  :  3108.1   24.2   751.0    1305    58      56   751    0   554     0  3052.6
 117 Minic_3.04                    :  3105.2   27.8   272.0     482    56      57   272    0   210     0  3058.7
 118 Drofa_3.2.0                   :  3102.9   17.5   548.0     911    60      59   548    0   363     0  3029.3
 119 Danasah_8.8                   :  3100.4   20.2   872.0    1703    51      53   872    0   831     0  3090.7
 120 Clover_2.2                    :  3099.3   23.9   472.0     838    56      67   472    0   366     0  3053.0
 121 Stash_30.0                    :  3092.8   15.4   752.0    1497    50      50   752    0   745     0  3090.4
 122 Seer_2.0.0                    :  3092.7   41.0   425.0     646    66      61   425    0   221     0  2972.6
 123 Berserk_4.1.0                 :  3087.2   16.8   747.0    1363    55      80   747    0   616     0  3052.2
 124 Weiss_1.3                     :  3076.7   18.6   989.0    1819    54      53   989    0   830     0  3043.5
 125 Minic_3.08                    :  3076.0   17.5   480.0     989    49      60   480    0   509     0  3086.1
 126 Velvet_3.1.0                  :  3071.9   26.8   555.0     873    64      50   555    0   318     0  2974.2
 127 Counter_3.9                   :  3071.8   26.0   335.0     623    54      89   335    0   288     0  3044.4
 128 Minic_3.07                    :  3062.1   29.5   403.0     732    55      66   403    0   329     0  3025.5
 129 BitGenie_9                    :  3055.2   20.6   498.0     872    57      69   498    0   374     0  3004.3
 130 Stash_29.0                    :  3048.0   22.6   407.0     889    46      64   407    0   482     0  3078.0
 131 Berserk_3.3.0                 :  3041.2   27.4   407.0     531    77      56   407    0   124     0  2817.3
 132 Orion_0.8                     :  3039.1   18.0   942.0    2318    41      63   942    0  1376     0  3108.3
 133 Bagatur_2.3                   :  3034.4   25.8   369.0     641    58      54   369    0   272     0  2981.2
 134 Minic_3.06                    :  3033.0   22.0   439.0     905    49      56   439    0   466     0  3044.3
 135 Berserk_4.0.0                 :  3031.3    9.2   430.0     831    52      61   430    0   401     0  3018.8
 136 Minic_3.09                    :  3027.6   21.9   431.0    1166    37      55   431    0   735     0  3122.4
 137 Vajolet2_2.8                  :  3025.0   31.5   214.0     488    44      51   214    0   274     0  3070.0
 138 Arasan_22.2                   :  3024.7   29.2   156.0     637    24      53   156    0   481     0  3239.3
 139 Monolith_2                    :  3023.1   26.9   441.0     897    49      53   441    0   456     0  3029.0
 140 Invictus_r340                 :  3021.4   30.6   560.0     843    66      70   560    0   283     0  2897.1
 141 Marvin_5.0                    :  3015.1   17.7   778.0    2109    37      51   778    0  1331     0  3112.9
 142 Minic_3.04x                   :  3014.9   21.2   431.0    1222    35      60   431    0   791     0  3122.9
 143 RubiChess_2.0                 :  3009.7   41.0    75.0     633    12      53    75    0   558     0  3502.5
 144 Velvet_3.0.0                  :  3007.7   11.0   528.0     760    69      52   528    0   232     0  2861.8
 145 Pirarucu_3.5.5                :  3007.0   21.4   310.0     622    50      79   310    0   312     0  3009.2
 146 Rodent_4.022                  :  2989.8   36.4   419.0     925    45      52   419    0   506     0  3023.3
 147 Koivisto_4.0                  :  2988.6   24.5   256.0     691    37      63   256    0   435     0  3085.1
 148 Cheng_4.41                    :  2984.1   19.0   988.0    2219    45      61   988    0  1231     0  3023.3
 149 Wasp_2.01                     :  2981.4   24.7   567.0     926    61      62   567    0   359     0  2899.4
 150 Counter_3.8                   :  2977.4   13.9   675.0    1734    39      53   675    0  1059     0  3058.3
 151 Amoeba_3.3                    :  2976.7   19.8  1138.0    2423    47      50  1138    0  1285     0  2998.1
 152 Wasp_2.00                     :  2976.6   18.6   649.0    1087    60      71   649    0   438     0  2906.0
 153 FabChess_1.16                 :  2972.0   16.7  1173.0    2469    48      96  1173    0  1296     0  2990.2
 154 Tucano_9.0                    :  2953.8    9.5   788.0    1817    43      57   788    0  1029     0  3001.2
 155 Seer_1.2.1                    :  2950.7   29.5   280.0     700    40      58   280    0   420     0  3025.2
 156 Topple_0.8.0                  :  2947.3   22.0   831.0    2449    34      51   831    0  1618     0  3068.6
 157 Mr.Bob_1.0.0                  :  2947.1   12.7   919.0    2083    44      51   919    0  1164     0  2990.7
 158 Drofa_3.1.0                   :  2946.6   27.1   501.0    1084    46      50   501    0   583     0  2974.1
 159 Wasp_1.25                     :  2946.4   21.8   506.0     912    55      61   506    0   406     0  2906.7
 160 Zahak_7.0                     :  2942.5   27.9   393.0     724    54      65   393    0   331     0  2910.6
 161 Counter_3.7                   :  2934.5   28.4   487.0    1401    35      82   487    0   914     0  3049.9
 162 Cheng_4.40                    :  2925.7   19.5   837.0    1885    44      73   837    0  1048     0  2967.0
 163 Komodo_1.0                    :  2915.3   14.4   629.0    1238    51      99   629    0   609     0  2909.7
 164 Asymptote_0.8                 :  2891.0   19.8   787.0    1773    44      70   787    0   986     0  2929.4
 165 Wasp_1.02                     :  2883.5   18.6   457.0     959    48      73   457    0   502     0  2900.0
 166 BitGenie_8                    :  2876.8   20.3   704.0    1339    53      97   704    0   635     0  2857.7
 167 Drofa_3.0.0                   :  2856.7   13.8  1269.0    2535    50      76  1269    0  1266     0  2854.5
 168 Black-Marlin_1.0              :  2850.6   18.9   467.0     843    55      52   467    0   376     0  2811.3
 169 Nemo_1.01                     :  2850.0    8.5  1109.0    2855    39      63  1109    0  1746     0  2902.3
 170 Velvet_2.0.0                  :  2845.7   24.9   803.0    1472    55      60   803    0   669     0  2811.7
 171 Olithink_5.9.5                :  2842.2   25.3   304.0     660    46      71   304    0   356     0  2872.1
 172 Cheese_2.2                    :  2834.0   21.7  1198.0    2462    49      56  1198    0  1264     0  2843.6
 173 Olithink_5.9.8                :  2831.7   21.3   620.0    1562    40      53   620    0   942     0  2906.5
 174 Olithink_5.9.9                :  2830.2   32.0   395.0     926    43      91   395    0   531     0  2884.7
 175 Benjamin_1.1                  :  2805.6   10.2   480.0     883    54      64   480    0   403     0  2771.6
 176 Olithink_5.9.3                :  2801.6   23.3   439.0     964    46      56   439    0   525     0  2832.2
 177 Arasan_16.0                   :  2798.1   32.3   453.0     765    59      58   453    0   312     0  2730.3
 178 Nalwald_1.14                  :  2793.7   22.8   544.0    1204    45      55   544    0   660     0  2826.9
 179 GreKo_2021-08                 :  2791.6   20.0   535.0    1048    51      52   535    0   513     0  2781.3
 180 Shallow_v4                    :  2790.7   21.0   443.0    1024    43      56   443    0   581     0  2837.2
 181 Zahak_6.0                     :  2789.1   17.9   593.0    1194    50      69   593    0   601     0  2790.5
 182 Mantissa_2.5.0                :  2783.2   17.3   521.0     906    58      65   521    0   385     0  2727.6
 183 ProDeo_3.1                    :  2779.2   16.8   686.0    1494    46      87   686    0   808     0  2809.8
 184 BitGenie_7                    :  2764.4   16.5   777.0    1470    53      67   777    0   693     0  2741.0
 185 Nalwald_1.10                  :  2756.3   27.1   487.0     897    54      55   487    0   410     0  2723.7
 186 Colossus_2021b                :  2754.8   21.1   407.0     863    47      90   407    0   456     0  2772.0
 187 Nalwald_1.11                  :  2745.4   11.5   818.0    1937    42      60   818    0  1119     0  2800.2
 188 Benjamin_1.0                  :  2742.8   13.9   931.0    1699    55      50   931    0   768     0  2700.3
 189 Orion_07                      :  2742.7   16.0   806.0    1653    49      51   806    0   847     0  2749.8
 190 Velvet_1.2.0                  :  2742.5   20.3   797.0    1597    50      54   797    0   800     0  2739.0
 191 k2_099                        :  2741.3   19.3   954.0    1760    54      73   954    0   806     0  2705.7
 192 Dumb_1.8                      :  2734.4   21.4   980.0    1854    53      71   980    0   874     0  2708.5
 193 Zahak_5.0                     :  2726.5   20.4  1152.0    2015    57      50  1152    0   863     0  2672.0
 194 ProDeo_2.2                    :  2726.3   23.4   517.0    1085    48      71   517    0   568     0  2741.1
 195 Colossus_2021a                :  2712.7   32.6   327.0     745    44      68   327    0   418     0  2752.7
 196 ProDeo_3.0                    :  2701.5   31.5   245.0     782    31      53   245    0   537     0  2849.7
 197 Fruit_2.1                     :  2700.0    8.5  1029.0    2187    47      66  1029    0  1158     0  2723.5
 198 Supernova_2.3                 :  2695.9   19.7   950.0    1898    50      62   950    0   948     0  2690.2
 199 Fridolin_3.10                 :  2689.9   25.7   495.0     960    52      57   495    0   465     0  2670.3
 200 Supernova_2.4                 :  2686.3   18.1   501.0    1105    45      86   501    0   604     0  2717.0
 201 Nalwald_1.9                   :  2672.7   32.1   402.0     746    54      60   402    0   344     0  2641.2
 202 Orion_06                      :  2667.2   24.6   352.0     669    53      61   352    0   317     0  2646.8
 203 Monolith_0.4                  :  2662.0   19.6   543.0    1036    52      72   543    0   493     0  2642.5
 204 BitGenie_6                    :  2652.0   26.5   637.0     961    66      69   637    0   324     0  2517.7
 205 Knightx_2.8                   :  2643.4   30.4   388.0     961    40      55   388    0   573     0  2709.7
 206 Arasan_14.1                   :  2640.1   30.5   138.0     289    48      65   138    0   151     0  2654.4
 207 Myrddin_0.88                  :  2633.8   31.4   699.0    1052    66      72   699    0   353     0  2506.6
 208 Lozza_2.0a                    :  2624.2   47.4   515.0     857    60      60   515    0   342     0  2546.8
 209 Zevra_2.4-r380                :  2617.3   16.3   363.0     987    37      92   363    0   624     0  2710.2
 210 Galjoen_0.41.2                :  2600.3   25.0   351.0     728    48      54   351    0   377     0  2609.8
 211 MadChess_3.0                  :  2598.6   27.5   296.0     783    38      57   296    0   487     0  2683.5
 212 Zahak_4.0                     :  2595.7   29.0   943.0    1497    63      71   943    0   554     0  2492.9
 213 Jumbo_0.7.0                   :  2590.4   36.8   343.0     769    45      57   343    0   426     0  2627.0
 214 Arasan_12.2                   :  2587.3   34.5   392.0     668    59      50   392    0   276     0  2520.0
 215 Galjoen_0.41.1                :  2587.2   33.1   738.0    1372    54      69   738    0   634     0  2555.2
 216 Orion_05                      :  2579.1   36.3   515.0     967    53      51   515    0   452     0  2551.1
 217 Arasan_13.0                   :  2578.2   29.9   114.0     308    37      79   114    0   194     0  2669.0
 218 Nalwald_1.8.1                 :  2560.6   37.0   494.0     845    58      98   494    0   351     0  2492.4
 219 Monolith_0.3                  :  2543.2   30.9   931.0    1789    52      78   931    0   858     0  2521.2
 220 Jumbo_0.6.10                  :  2535.6   37.7   801.0    1572    51     100   801    0   771     0  2521.3
 221 CT800_1.43                    :  2511.3   32.1   515.0     963    53      62   515    0   448     0  2477.0
 222 Marvin_2.0                    :  2506.0   49.8   241.0     573    42      68   241    0   332     0  2558.3
 223 Loki_3.5.0                    :  2497.5   33.2   508.0    1114    46      55   508    0   606     0  2524.6
 224 Foxsee_7.20.1                 :  2496.3   29.6   768.0    1674    46      64   768    0   906     0  2520.4
 225 Orion_04                      :  2492.6   32.7   648.0    1371    47     100   648    0   723     0  2507.3
 226 CT800_1.42                    :  2461.8   35.1   811.0    2130    38      63   811    0  1319     0  2549.7
 227 BitGenie_5.h64                :  2458.8   34.2   340.0     691    49      84   340    0   351     0  2460.0
 228 Jumbo_0.5.3                   :  2446.8   30.1   578.0    1476    39     100   578    0   898     0  2522.7
 229 MinimalChess_06               :  2410.2   32.6   273.0     767    36      67   273    0   494     0  2510.2
 230 Foxsee_7.8                    :  2404.3   37.7   279.0    1081    26      79   279    0   802     0  2606.3
 231 Zahak_3.0                     :  2391.1   40.9   341.0     858    40      92   341    0   517     0  2463.8
 232 Orion_03                      :  2367.4   35.3   330.0    1050    31     100   330    0   720     0  2507.2
 233 BitGenie_5                    :  2259.4   58.0   203.0     881    23     ---   203    0   678     0  2469.4

Dio likes this post

Back to top Go down
http://rebel13.nl/
TheSelfImprover

TheSelfImprover


Posts : 3112
Join date : 2020-11-18

I wonder what is the reason that so many cores work for stockfish - Page 2 Empty
PostSubject: Re: I wonder what is the reason that so many cores work for stockfish   I wonder what is the reason that so many cores work for stockfish - Page 2 EmptyWed Jul 27, 2022 9:48 pm

Admin wrote:
This is the GRL without draws...


Again, amplifying the differences - you thrill seeker! Smile

I'm with mwyoung on this one:

* draws are part of chess

* draws are required information in making ratings

As a demonstration of how this distorts the information, imagine hearing, "there were 10 draws between engine A and engine B", but not being told that engine A won the other 999,990 games!  Suspect
Back to top Go down
mwyoung

mwyoung


Posts : 880
Join date : 2020-11-25
Location : USA

I wonder what is the reason that so many cores work for stockfish - Page 2 Empty
PostSubject: Re: I wonder what is the reason that so many cores work for stockfish   I wonder what is the reason that so many cores work for stockfish - Page 2 EmptyWed Jul 27, 2022 10:41 pm

pohl4711 wrote:
Uri Blass wrote:

I talk about progress in normal chess at LTC
UHO opening is not normal chess.

And this just false. My UHO openings are selected out of the Megabase and all openings were played by strong human players. So, no openings could be more "normal"/human chess than UHO. And I did not change anything in these openings, I just used the KomodoDragon eval for selecting them. Thats why I called them UHO: Unbalanced Human Openings. All Openings were played by strong humans (all players must have 2300+ Elo. All games played after WW2, so no crazy gambits from 19th century or so). All games out of the Megabase.

Just as example, here the first 5 lines out of the UHO_2022_6mvs_+120_+129 file, which I use for the VLTC Stockfish regression test.
(Shirov, Ivanchuk, Kamsky, Morozevich are the players, which played the lines with black, which ends with a clear white advantage. Are these players not "normal" or not strong enough for playing good chess? I am so tired of reading such nonsense about my work)

[Event "Novgorod"]
[Site "Novgorod"]
[Date "1994.??.??"]
[Round "1"]
[White "Kasparov, Garry"]
[Black "Shirov, Alexei"]
[Result "1/2-1/2"]
[ECO "A21"]
[WhiteElo "2805"]
[BlackElo "2740"]
[Annotator "depth=29 eval=+124"]
[PlyCount "12"]
[EventDate "1994.??.??"]

1. c4 e5 2. Nc3 Bb4 3. Nd5 Be7 4. d4 d6 5. e4 c6 6. Nxe7 Qxe7 1/2-1/2

[Event "Horgen"]
[Site "Horgen"]
[Date "1995.10.26"]
[Round "6"]
[White "Kasparov, Garry"]
[Black "Ivanchuk, Vassily"]
[Result "1/2-1/2"]
[ECO "C16"]
[WhiteElo "2795"]
[BlackElo "2740"]
[Annotator "depth=28 eval=+126"]
[PlyCount "12"]
[EventDate "1995.??.??"]

1. e4 e6 2. d4 d5 3. Nc3 Bb4 4. e5 b6 5. a3 Bf8 6. Nf3 Ne7 1/2-1/2

[Event "Amber-rapid 05th"]
[Site "Monte Carlo"]
[Date "1996.??.??"]
[Round "3"]
[White "Anand, Viswanathan"]
[Black "Ivanchuk, Vassily"]
[Result "1/2-1/2"]
[ECO "C41"]
[WhiteElo "2725"]
[BlackElo "2735"]
[Annotator "depth=30 eval=+126"]
[PlyCount "12"]
[EventDate "1996.??.??"]

1. e4 e5 2. Nf3 d6 3. Bc4 c6 4. d4 Nd7 5. O-O Be7 6. dxe5 dxe5 1/2-1/2

[Event "FIDE-Wch"]
[Site "Elista"]
[Date "1996.07.04"]
[Round "15"]
[White "Karpov, Anatoly"]
[Black "Kamsky, Gata"]
[Result "1/2-1/2"]
[ECO "A70"]
[WhiteElo "2770"]
[BlackElo "2735"]
[Annotator "depth=27 eval=+120"]
[PlyCount "12"]
[EventDate "1996.??.??"]

1. d4 Nf6 2. c4 c5 3. d5 e6 4. Nc3 exd5 5. cxd5 d6 6. e4 g6 1/2-1/2

[Event "Corus"]
[Site "Wijk aan Zee"]
[Date "2001.01.21"]
[Round "7"]
[White "Anand, Viswanathan"]
[Black "Morozevich, Alexander"]
[Result "1/2-1/2"]
[ECO "D07"]
[WhiteElo "2790"]
[BlackElo "2745"]
[Annotator "depth=29 eval=+129"]
[PlyCount "12"]
[EventDate "2001.??.??"]

1. d4 d5 2. c4 Nc6 3. Nc3 dxc4 4. Nf3 Nf6 5. e4 Bg4 6. Be3 Bxf3 1/2-1/2


Lets take a look at your work. And see what the best players feel about your openings today.

Not one top human player plays your garbage openings anymore. As they know the openings are awful even for them. After the first top player blundered the opening.

These openings are in the scrap heap of history.

"And this just false. My UHO openings are selected out of the Megabase and all openings were played by strong human players. So, no openings could be more "normal"/human chess than UHO."  "I am so tired of reading such nonsense about my work." lol!





Opening report



1.c4 e5 2.Nc3 Bb4 3.Nd5 Be7 4.d4 d6 5.e4 c6 6.Nxe7 Qxe7

fen:rnb1k1nr/pp2qppp/2pp4/4p3/2PPP3/8/PP3PPP/R1BQKBNR w KQkq - 0 1;


A21, 6...Qxe7
Reference

1. History


Earliest game: Kasparov,G - Shirov,A 1-0
Latest grandmaster game: Skoberne,J - Tratar,M 1-0





Games played in years (10)


2. Players

a) Strong grandmasters who used this line as Black:

Jure Borisek Result=2/2 2008 Elo-Ø: 2494 Games: 2
Utut Adianto Result=1/1 2008 Elo-Ø: 2554 Games: 1
Marko Tratar Result=0/1 2011 Elo-Ø: 2491 Games: 1
Alexei Shirov Result=0/1 1994 Elo-Ø: 2740 Games: 1
Timofey Iljin Result=0/1 2011 Elo-Ø: 2494 Games: 1


b) Other notable players:

Yaacov Stisis Result=1.5/2 1999-2002 Elo-Ø: 2389 Games: 2
Jean Netzer Result=0.5/2 2005-2006 Elo-Ø: 2373 Games: 2


3. Statistics
Black scores above average (50%).
Black performs Elo 2502 against an opposition of Elo 2502 (0).
White performs Elo 2479 against an opposition of Elo 2479 (0).
White wins: 4 (=40%), Draws: 2 (=20%), Black wins: 4 (=40%)

4. Moves and Plans


5. Opening Keys

Key: 4.... d6 5.---

Opening report



1.e4 e6 2.d4 d5 3.Nc3 Bb4 4.e5 b6 5.a3 Bf8 6.Nf3 Ne7

fen:rnbqkb1r/p1p1nppp/1p2p3/3pP3/3P4/P1N2N2/1PP2PPP/R1BQKB1R w KQkq - 0 1;


C16, 6...Ne7
Reference

1. History


Earliest game: Dolmatov,S - Gulko,B ½-½
Latest grandmaster game: Pichot,A - Roselli Mailhe,B 1-0
Latest game: Alekseev,E - Ovchinnikov,V 1-0




Games played in years (83)


Fashion index - Distribution of games relative to database [%]


2. Players

a) Strong grandmasters who used this line as Black:

Boris F Gulko Result=3/5 1981-2007 Elo-Ø: 2554 Games: 5
Viktor Moskalenko Result=2/5 2003-2008 Elo-Ø: 2558 Games: 5
Lajos Portisch Result=1.5/2 1988-1990 Elo-Ø: 2620 Games: 2
Vassily Ivanchuk Result=1/1 1995 Elo-Ø: 2740 Games: 1
Andrei Istratescu Result=1/1 2015 Elo-Ø: 2628 Games: 1
Artur Jussupow Result=0.5/2 2017-2018 Elo-Ø: 2581 Games: 2
Rafael A Vaganian Result=0.5/1 1999 Elo-Ø: 2605 Games: 1
Dmitry Andreikin Result=0/1 2008 Elo-Ø: 2604 Games: 1
Lev Psakhis Result=0/2 1983-1985 Elo-Ø: 2567 Games: 2


b) Other notable players:

Iulija Osmak Result=2/6 2021 Elo-Ø: 2418 Games: 6
Goran Dizdar Result=1.5/3 1983-2007 Elo-Ø: 2501 Games: 3
Bernardo Roselli Mailhe Result=1.5/3 2009-2019 Elo-Ø: 2424 Games: 3
Nils Gustaf Renman Result=1/2 1996-2001 Elo-Ø: 2381 Games: 2


3. Statistics
Black scores badly (32%).
Black performs Elo 2356 against an opposition of Elo 2487 (-131).
White performs Elo 2600 against an opposition of Elo 2469 (+131).
White wins: 44 (=53%), Draws: 25 (=30%), Black wins: 14 (=17%)
The drawing quote is lower than average. (1% quick draws, < 20 Moves)

White wins are longer than average (43).
Black wins are longer than average (44).
Draws are of average length (41).

4. Moves and Plans

a) 7.h4

30 Games
White scores well (70%).

played by: Kasparov, 2795, 0/1; Sokolov, 2585, 0.5/1; Rogers, 2570, 1/1; Dolmatov, 2555, 2/2; Short, 2535, 1/1

You should play: 7...h5

Sokolov,A - Portisch,L ½-½; Van de Oudeweetering,A - Vogel,J 0-1; Dolmatov,S - Danielian,O 1-0


Alternative: 7...Qd7

Lanka,Z - Mchedlishvili,M ½-½; Short,N - Psakhis,L 1-0; Rogers,I - Sirota,A 1-0


b) 7.Bb5+

14 Games
White scores excellently (82%).

played by: Sax, 2600, 0/1; Georgiev, 2585, 1/1; Velimirovic, 2535, 0.5/1; Kruppa, 2440, 1/1; Grigorov, 2385, 1/1

You should play: 7...c6

Velimirovic,D - Dizdar,G ½-½; Georgiev,K - Conquest,S 1-0; Kruppa,Y - Moroz,A 1-0; Grigorov,J - Mollov,E 1-0

Main line:


8.Ba4 a5  N=4, Last played 2020, 100% =>2839  Reference
8.Ba4 Ba6 9.Ne2 Bb5  N=4, Last played 2000, 88% =>2678  Reference


c) 7.b4

13 Games
White scores badly (46%).

played by: Tal, 2620, 0.5/1; Yurtaev, 2385, 0/1; Pozdniakov, 2360, 0.5/1

You should play: 7...c5

Yurtaev,L - Gulko,B 0-1; Tal,M - Dizdar,G ½-½

Main line:


8.Bg5  N=4, Last played 2021, 50% =>2479  Reference



5. Opening Keys

Key: 6.... Ne7 7.---


Opening report



1.e4 e5 2.Nf3 d6 3.Bc4 c6 4.d4 Nd7 5.0-0 Be7 6.dxe5 dxe5

fen:r1bqk1nr/pp1nbppp/2p5/4p3/2B1P3/5N2/PPP2PPP/RNBQ1RK1 w kq - 0 1;


C41, 6...dxe5
Reference

1. History


Earliest game: Fischer,R - Fine,R 1-0
Latest grandmaster game: Vachier Lagrave,M - Cossin,S 1-0
Latest game: Medarde Santiago,L - Suarez Pousa,D 1-0




Games played in years (95)


Fashion index - Distribution of games relative to database [%]


2. Players

a) Strong grandmasters who used this line as Black:

Vassily Ivanchuk Result=0.5/1 1996 Elo-Ø: 2735 Games: 1
Tonu O Oim Result=0/1 1993 Elo-Ø: 2562 Games: 1
Istvan Csom Result=0/1 1977 Elo-Ø: 2535 Games: 1
Pengxiang Zhang Result=0/1 2008 Elo-Ø: 2640 Games: 1
Ian Rogers Result=0/2 1998-2000 Elo-Ø: 2566 Games: 2


b) Other notable players:

Tamaz Gelashvili Result=2.5/4 1996-2008 Elo-Ø: 2502 Games: 4
Juan Ignacio Alonso Arburu Result=2/3 1992-2007 Elo-Ø: 2337 Games: 3
Zaw Htung Wynn Result=2/2 2004-2005 Elo-Ø: 2504 Games: 2
Zoran Runic Result=1.5/6 2002-2006 Elo-Ø: 2425 Games: 6
Dibyendu Barua Result=1/2 1983-2000 Elo-Ø: 2472 Games: 2
Zenon Franco Ocampos Result=1/2 1994 Elo-Ø: 2492 Games: 2
Ramil Hasangatin Result=1/2 1996 Elo-Ø: 2390 Games: 2


3. Statistics
Black scores badly (28%).
Black performs Elo 2299 against an opposition of Elo 2463 (-164).
White performs Elo 2588 against an opposition of Elo 2424 (+164).
White wins: 51 (=54%), Draws: 34 (=36%), Black wins: 10 (=11%)
The drawing quote is Average. (5% quick draws, < 20 Moves)

White wins are of average length (40).
Black wins are longer than average (43).
Draws are of average length (38).

4. Moves and Plans

a) 7.Ng5

90 Games
White scores well (72%).

played by: Anand, 2725, 0.5/1; Topalov, 2630, 0.5/1; Granda Zuniga, 2630, 1/1; Hracek, 2615, 2/2; Akopian, 2615, 1/1;

You should play: 7...Bxg5 Reference
fen:r1bqk1nr/pp1n1ppp/2p5/4p1b1/2B1P3/8/PPP2PPP/RNBQ1RK1 w kq - 0 1;

Anand,V - Ivanchuk,V ½-½; Matulovic,M - Abramovic,B 0-1; Topalov,V - Franco Ocampos,Z ½-½; Andrijevic,M - Gelashvili,T 0-1; Dizdar,G - Ivanovic,B 0-1; Tseshkovsky,V - Averkin,O ½-½

Main line:


8.Qh5 Qe7 9.Qxg5 Ngf6 10.b3  N=9, Last played 2018, 83% =>2649  Reference
8.Qh5 Qe7 9.Qxg5 Qxg5 10.Bxg5 Ngf6  N=9, Last played 2015, 83% =>2629  Reference
8.Qh5 Qe7 9.Qxg5 Qxg5 10.Bxg5 Nc5  N=5, Last played 2008, 90% =>2782  Reference
8.Qh5 Qe7 9.Bxg5 Ngf6 10.Qh4 Nf8  N=6, Last played 2008, 50% =>2462  Reference
8.Qh5 Qe7 9.Bxg5 Ngf6 10.Qe2 h6  N=9, Last played 2014, 44% =>2441  Reference
8.Qh5 g6 9.Qxg5 Qxg5 10.Bxg5 Nc5  N=9, Last played 2011, 89% =>2716  Reference
8.Qh5 Qf6 9.Bxg5 Qg6 10.Qxg6 hxg6  N=8, Last played 2013, 75% =>2593  Reference


b) 7.Qe2

4 Games
White scores excellently (75%).

played by: Rodriguez Vargas, 2485, 0.5/1; Ricardi, 2335, 0.5/1; Fischer, 1/1

You should play: 7...h6

Rodriguez Vargas,O - Hernando Pertierra,J ½-½; Ricardi,P - Paglilla,C ½-½



5. Opening Keys

Key: 6.de5 de5 7.---

Opening report



1.d4 Nf6 2.c4 c5 3.d5 e6 4.Nc3 exd5 5.cxd5 d6 6.e4 g6

fen:rnbqkb1r/pp3p1p/3p1np1/2pP4/4P3/2N5/PP3PPP/R1BQKBNR w KQkq - 0 1;


A70, 6...g6
3164 Games in 'Big Database 2021'

1. History


Earliest game: Boleslavsky,I - Nezhmetdinov,R ½-½
Latest grandmaster game: Navara,D - Pantsulaia,L ½-½
Latest game: Barp,A - Batsuren,D 1-0




Games played in years (3164)


Fashion index - Distribution of games relative to database [%]


2. Players

a) Strong grandmasters who used this line as Black:

Etienne Bacrot Result=3/4 2009-2011 Elo-Ø: 2719 Games: 4
Veselin Topalov Result=3/5 1994-2004 Elo-Ø: 2713 Games: 5
Vugar Gashimov Result=1.5/3 2008-2011 Elo-Ø: 2719 Games: 3
Vassily Ivanchuk Result=1.5/3 1994-2009 Elo-Ø: 2735 Games: 3
Vladimir Malakhov Result=1/1 2007 Elo-Ø: 2690 Games: 1
Vladimir1 Fedoseev Result=1/1 2021 Elo-Ø: 2704 Games: 1
Judit Polgar Result=1/1 2011 Elo-Ø: 2699 Games: 1
Arkadij Naiditsch Result=1/3 2006-2012 Elo-Ø: 2694 Games: 3
Gata Kamsky Result=1/2 1993-1996 Elo-Ø: 2690 Games: 2
Alexander Grischuk Result=1/1 2011 Elo-Ø: 2746 Games: 1


b) Other notable players:

Aleksandar Indjic Result=28.5/46 2011-2021 Elo-Ø: 2573 Games: 46
Artur Gabrielian Result=28/50 2004-2021 Elo-Ø: 2518 Games: 50
Levan Pantsulaia Result=25.5/50 2001-2021 Elo-Ø: 2575 Games: 50
Pavel Simacek Result=21/52 1999-2020 Elo-Ø: 2454 Games: 52
Dmitry Kononenko Result=15/23 2003-2015 Elo-Ø: 2533 Games: 23
Kjetil A Lie Result=12.5/18 2001-2016 Elo-Ø: 2472 Games: 18
Mikhailo Oleksienko Result=10/17 2004-2014 Elo-Ø: 2505 Games: 17
Aleksandar Kovacevic Result=10/15 1997-2017 Elo-Ø: 2529 Games: 15


3. Statistics
Black scores below average (41%).
Black performs Elo 2396 against an opposition of Elo 2459 (-63).
White performs Elo 2508 against an opposition of Elo 2444 (+64).
White wins: 1387 (=44%), Draws: 947 (=30%), Black wins: 828 (=26%)
The drawing quote is lower than average. (4% quick draws, < 20 Moves)

White wins are of average length (41).
Black wins are of average length (42).
Draws are of average length (40).

4. Moves and Plans

a) 7.f4

1088 Games
White scores above average (60%).

played by: Kasparov, 2675, 1/1; Sokolov, 2665, 1/1; Jussupow, 2640, 1/1; Tal, 2605, 1/1; Van Wely, 2590, 2/2;

You should play: 7...Bg7 Click for games
fen:rnbqk2r/pp3pbp/3p1np1/2pP4/4PP2/2N5/PP4PP/R1BQKBNR w KQkq - 0 1;

Lalic,B - Kotsur,P 0-1; Hort,V - Hulak,K 0-1; Djuric,S - Pavlovic,M 0-1; Cvitan,O - Velimirovic,D 0-1; Farago,I - Suetin,A 0-1; Leitao,R - Ginzburg,M 0-1

Main line:


8.Bb5+ Nfd7 9.a4 0-0 10.Nf3 Na6  N=243, Last played 2021, 62% =>2538  Click for games
8.Bb5+ Nfd7 9.Nf3  N=212, Last played 2021, 63% =>2535  Click for games


b) 7.Nf3

545 Games
White scores averagely (55%).

played by: Karpov, 2750, 1/2; Kamsky, 2695, 0/1; Polugaevsky, 2645, 1/1; Portisch, 2630, 0/1; Granda Zuniga, 2600, 1/1;

You should play: 7...Bg7 Reference
fen:rnbqk2r/pp3pbp/3p1np1/2pP4/4P3/2N2N2/PP3PPP/R1BQKB1R w KQkq - 0 1;

Kamsky,G - Ivanchuk,V 0-1; Karpov,A - Kamsky,G ½-½; Portisch,L - Nunn,J 0-1; Karpov,A - Timman,J ½-½; Alburt,L - Tukmakov,V 0-1; Schmidt,W - Sax,G 0-1

Main line:


8.h3 0-0 9.Bd3 b5  N=242, Last played 2021, 61% =>2527  Click for games
8.Be2 0-0 9.0-0  N=162, Last played 2021, 55% =>2461  Click for games


c) 7.Bd3

450 Games
White scores above average (58%).

played by: Spassky, 2630, 0.5/1; Kortschnoj, 2625, 0.5/1; Chernin, 2617, 1/2; Gheorghiu, 2605, 1/1; Bareev, 2605, 1/1;

You should play: 7...Bg7 Reference
fen:rnbqk2r/pp3pbp/3p1np1/2pP4/4P3/2NB4/PP3PPP/R1BQK1NR w KQkq - 0 1;

Atalik,S - Velimirovic,D 0-1; Ibragimov,I - Pigusov,E 0-1; Kortschnoj,V - Kamsky,G ½-½; Mozetic,D - Pavlovic,M 0-1; Spassky,B - Ljubojevic,L ½-½; Razuvaev,Y - Psakhis,L 0-1

Main line:


8.Nge2 0-0 9.0-0  N=178, Last played 2020, 56% =>2475  Click for games


d) 7.f3

192 Games
White scores above average (60%).

played by: Lautier, 2660, 0.5/1; Kortschnoj, 2650, 1/1; Jussupow, 2645, 1/1; Spassky, 2625, 0.5/1; Polugaevsky, 2615, 1/1;

You should play: 7...Bg7 Reference
fen:rnbqk2r/pp3pbp/3p1np1/2pP4/4P3/2N2P2/PP4PP/R1BQKBNR w KQkq - 0 1;

Seirawan,Y - Nunn,J 0-1; Giorgadze,T - Psakhis,L 0-1; Piket,J - Galego,L 0-1; Damljanovic,B - Velimirovic,D 0-1; Spassky,B - Nunn,J ½-½; Lautier,J - Kovacevic,A ½-½

Main line:


8.Bg5  N=132, Last played 2021, 59% =>2493  Click for games


e) 7.h3

519 Games
White scores well (64%).

played by: Bareev, 2685, 1/1; Van Wely, 2605, 0/1; Krasenkow, 2605, 1/1; Kaidanov, 2600, 0.5/1; Ibragimov, 2580, 1/1;

You should play: 7...Bg7 Reference
fen:rnbqk2r/pp3pbp/3p1np1/2pP4/4P3/2N4P/PP3PP1/R1BQKBNR w KQkq - 0 1;

Kaidanov,G - De Firmian,N ½-½; Miladinovic,I - Graf,A ½-½; Wells,P - De Firmian,N ½-½; Babula,V - Pisk,P 0-1; Flear,G - Kovacevic,S 0-1; Bareev,E - Lobron,E 1-0


f) 7.Bf4

45 Games
White scores above average (61%).

played by: Kortschnoj, 2695, 1/1; Ree, 2470, 2/2; Najdorf, 2465, 0/1; Razuvaev, 2460, 0/1; Reshevsky, 2455, 1/1;

You should play: 7...a6

Rubinetti,J - Najdorf,M 0-1; Schussler,H - Petrosian,T ½-½; Schussler,H - Ermenkov,E 0-1; Najdorf,M - Glavina,P 0-1; Kortschnoj,V - Nunn,J 1-0; Krasenkow,M - Shabalov,A 0-1


Alternative: 7...Bg7

Garcia Martinez,S - Kasparov,G 0-1; Razuvaev,Y - Tseshkovsky,V 0-1; Mohr,S - Pliester,L 0-1; Reshevsky,S - Gruenfeld,Y 1-0; Forintos,G - Plachetka,J 1-0; Starck,B - Goldberg,A ½-½


g) 7.Be2

63 Games
White scores below average (50%).

played by: Larsen, 2610, 0/1; Alburt, 2575, 1/1; Beliavsky, 2555, 1/1; Dorfman, 2505, 0/1; Karpman, 2500, 0.5/1

You should play: 7...Bg7

Larsen,B - Christiansen,L 0-1; Efimov,I - Gavrikov,V 0-1; Dorfman,J - Dolmatov,S 0-1; Gaprindashvili,N - Oll,L 0-1; Gaprindashvili,N - Shcherbakov,R 0-1; Krasilnikov,V - Kotsur,P 0-1


h) 7.Nge2

221 Games
White scores averagely (56%).

played by: Dreev, 2645, 1/1; Gulko, 2525, 1/1; Najdorf, 2465, 0/1; Fioramonti, 2380, 0.5/1; Fokin, 2380, 1/1

You should play: 7...Bg7

Najdorf,M - Szmetan,J 0-1; Gulko,B - Findlay,I 1-0; Fioramonti,H - Goldstern,F ½-½; Fokin,S - Kotsur,P 1-0; Kaposztas,M - Feher,G ½-½; Fernandez,C - Fraguela Gil,J 1-0


i) 7.Bg5

17 Games
White scores below average (50%).

played by: Timman, 2590, 0/1; Rajkovic, 2485, 0.5/1; Chandler, 2450, 1/1; Vladimirov, 2435, 1/1; Fahnenschmidt, 2385, 0.5/1

You should play: 7...Bg7

Timman,J - Enklaar,B 0-1; Neiman,E - Pigusov,E 0-1; Rajkovic,D - Armas,J ½-½; Yuneev,A - Agapov,K 0-1; Fahnenschmidt,G - Kekelidze,M ½-½; Chandler,M - Lobron,E 1-0


j) 7.Bb5+

19 Games
White scores below average (47%).

played by: Burmakin, 2550, 1/1; Tosic, 2445, 0/1; Zvolanek, 2340, 0.5/1; Kovacs, 2335, 0/1; Alber, 2335, 0.5/1

You should play: 7...Bd7

Kovacs,L - Armas,J 0-1; Burmakin,V - Notkin,M 1-0; Alber,H - Arnold,L ½-½


Alternative: 7...Nbd7

Tosic,M - Kuzmin,G 0-1; Zvolanek,L - Poloch,P ½-½; Reshevsky,S - Tal,M 0-1



5. Endgames

Ordinary often reaches an endgame



6. Opening Keys

Key: 6.... g6 7.---

Opening report



1.d4 d5 2.c4 Nc6 3.Nc3 dxc4 4.Nf3 Nf6 5.e4 Bg4 6.Be3 Bxf3

fen:r2qkb1r/ppp1pppp/2n2n2/8/2pPP3/2N1Bb2/PP3PPP/R2QKB1R w KQkq - 0 1;


D07, 6...Bxf3
Reference

1. History


Earliest game: Sorensen,T - Kroll,O 0-1
Latest grandmaster game: Anand,V - Morozevich,A ½-½
Latest game: Mirzoev,A - Rzayev,B ½-½




Games played in years (23)


2. Players

a) Strong grandmasters who used this line as Black:

Alexander Stripunsky Result=0.5/2 2008-2011 Elo-Ø: 2560 Games: 2
Nenad Fercec Result=0.5/2 2004-2005 Elo-Ø: 2490 Games: 2
Alexander Morozevich Result=0.5/1 2001 Elo-Ø: 2745 Games: 1
Aleksander Mista Result=0.5/1 2012 Elo-Ø: 2587 Games: 1
Aleksandr Karpatchev Result=0/1 2005 Elo-Ø: 2461 Games: 1


b) Other notable players:

Gergely Antal Result=0.5/2 2002-2004 Elo-Ø: 2452 Games: 2


3. Statistics
Black scores badly (28%).
Black performs Elo 2269 against an opposition of Elo 2433 (-164).
White performs Elo 2591 against an opposition of Elo 2427 (+164).
White wins: 12 (=52%), Draws: 9 (=39%), Black wins: 2 (=9%)
The drawing quote is higher than average. (4% quick draws, < 20 Moves)

White wins are shorter than average (36).
Black wins are shorter than average (35).
Draws are long (51).

4. Moves and Plans

a) 7.gxf3

23 Games
White scores well (72%).

played by: Kishnev, 2520, 1/1; Flear, 2470, 1/1; Ligterink, 2440, 1/1; Berebora, 2415, 0.5/1; Sorensen, 2360, 0/1

You should play: 7...e5

Sorensen,T - Kroll,O 0-1; Kishnev,S - Baumhus,R 1-0; Berebora,F - Glavas,D ½-½; Flear,G - Bell,S 1-0; Nikolaev,S - Mann,C ½-½; Ligterink,G - Bartels,H 1-0

Main line:


8.d5 Nb8 9.Bxc4 Nbd7 10.Qb3 Bc5  N=6, Last played 2016, 67% =>2598  Reference
8.d5 Nb8 9.Bxc4 Bd6 10.Qb3  N=3, Last played 2011, 83% =>2640  Reference
8.d5 Ne7 9.Bxc4  N=5, Last played 2012, 70% =>2540  Reference
8.d5 Ne7 9.Qa4+ Nd7 10.d6  N=3, Last played 1995, 67% =>2456  Reference



5. Opening Keys

Key: 6.... Bf3
Back to top Go down
Admin
Admin
Admin


Posts : 2608
Join date : 2020-11-17
Location : Netherlands

I wonder what is the reason that so many cores work for stockfish - Page 2 Empty
PostSubject: Re: I wonder what is the reason that so many cores work for stockfish   I wonder what is the reason that so many cores work for stockfish - Page 2 EmptyThu Jul 28, 2022 9:35 am

Take a look a the GRL - http://rebel13.nl/grl-best-40-2.html

And watch the number of lost games for each engine.

Code:
Engine           Games   Loss  Percent
Stockfish 15     2000     21    1.05%
Dragon 2.5       4200    323    7,69%
CSTal-1.35       1400    145   10,35%
Koivisto 8.0     2600    594   22.84%
Berserk 8.5.1    2400    497   20.70%
SlowChess 2.8    4800   1201   25.02%

These numbers matter.

The gap between SF and Komodo is gigantic, SF only losing 1% of the games it plays, Komodo 7%. This somehow should play a role in the elo calculation with the draw percentage that keeps on rising and rising the higher the time control, the higher the number of cores.
Back to top Go down
http://rebel13.nl/
mwyoung

mwyoung


Posts : 880
Join date : 2020-11-25
Location : USA

I wonder what is the reason that so many cores work for stockfish - Page 2 Empty
PostSubject: Re: I wonder what is the reason that so many cores work for stockfish   I wonder what is the reason that so many cores work for stockfish - Page 2 EmptyThu Jul 28, 2022 10:16 am

Admin wrote:
Take a look a the GRL - http://rebel13.nl/grl-best-40-2.html

And watch the number of lost games for each engine.

Code:
Engine           Games   Loss  Percent
Stockfish 15     2000     21    1.05%
Dragon 2.5       4200    323    7,69%
CSTal-1.35       1400    145   10,35%
Koivisto 8.0     2600    594   22.84%
Berserk 8.5.1    2400    497   20.70%
SlowChess 2.8    4800   1201   25.02%

These numbers matter.

The gap between SF and Komodo is gigantic, SF only losing 1% of the games it plays, Komodo 7%. This somehow should play a role in the elo calculation with the draw percentage that keeps on rising and rising the higher the time control, the higher the number of cores.

You are making a very bad assumption. You are assuming the whole truth can be found with only a 1 tier testing standard. In your rating list that is 40/2 and 1 core.

Stockfish NNUE is a very fast engine, but a horrible scaling engine. This allows other engines to catch up.

I always test with multi tier testing. That is why I was the first to find the Stockfish NNUE defect.
Back to top Go down
Admin
Admin
Admin


Posts : 2608
Join date : 2020-11-17
Location : Netherlands

I wonder what is the reason that so many cores work for stockfish - Page 2 Empty
PostSubject: Re: I wonder what is the reason that so many cores work for stockfish   I wonder what is the reason that so many cores work for stockfish - Page 2 EmptyThu Jul 28, 2022 10:37 am

Ha ha, I can show you the results of 20 core testing and the pattern remains, SF hardly losing a game and that is the reason why they almost win every tournament.

Also take into consideration that an elo improvement of (say) 20 elo in the current top-3 in reality is an elo improvement of ~100 elo for a (say) 3200 engine.

Bob Dylan already said it long time ago, the times are changing.
Back to top Go down
http://rebel13.nl/
mwyoung

mwyoung


Posts : 880
Join date : 2020-11-25
Location : USA

I wonder what is the reason that so many cores work for stockfish - Page 2 Empty
PostSubject: Re: I wonder what is the reason that so many cores work for stockfish   I wonder what is the reason that so many cores work for stockfish - Page 2 EmptyThu Jul 28, 2022 10:53 am

Admin wrote:
Ha ha, I can show you the results of 20 core testing and the pattern remains, SF hardly losing a game and that is the reason why they almost win every tournament.

Also take into consideration that an elo improvement of (say) 20 elo in the current top-3 in reality is an elo improvement of ~100 elo for a (say) 3200 engine.

Bob Dylan already said it long time ago, the times are changing.

Ha Ha, then show us the results at 20 cores. And lets compare the win% of Stockfish at each tier. Smile

Because facts are facts and defects are defects.

Test 5.

Bench 1024 16 26.

2950X
Total time (ms) : 129677
Nodes searched  : 3077330368
Nodes/second    : 23730733

Winner 5900HX 8 cores vs 2950X 16 cores.
Even with the 2950X searching many more NPS

Winner 5900HX 4 cores vs 2950X 16 cores.
Even with the 2950X 16 cores searching 23.7 Mnps vs 5900HX 4 cores searching 9.8 Mnps.

https://prodeo.actieforum.com/t872-stockfish-what-is-really-better-faster-cores-more-cores-and-nps-vs-time-to-depth

Code:
Rank Name                          Elo     +/-   Games    Wins  Losses   Draws   Points   Score    Draw
   1 Stockfish 17/09/21 - 8 CPU      11       9     160       5       0     155     82.5   51.6%   96.9%
   2 Stockfish 17/09/21 - 16 CPU       9       8     160       4       0     156     82.0   51.2%   97.5%
   3 Stockfish 17/09/21 - 4 CPU       7       9     160       4       1     155     81.5   50.9%   96.9%
   4 Stockfish 17/09/21 - 2 CPU      -2       9     160       2       3     155     79.5   49.7%   96.9%
   5 Stockfish 17/09/21 - 1 CPU     -24      14     160       0      11     149     74.5   46.6%   93.1%
Back to top Go down
Admin
Admin
Admin


Posts : 2608
Join date : 2020-11-17
Location : Netherlands

I wonder what is the reason that so many cores work for stockfish - Page 2 Empty
PostSubject: Re: I wonder what is the reason that so many cores work for stockfish   I wonder what is the reason that so many cores work for stockfish - Page 2 EmptyThu Jul 28, 2022 12:21 pm

And now you made Ed angry Very Happy

GRL - 20 cores

Code:
   # PLAYER                 :  RATING  ERROR  POINTS  PLAYED   (%)  CFS(%)    W    D    L  D(%)  OppAvg
   1 Stockfish_14           :  3617.5    9.7   408.5     600    68      96  227  363   10    61  3480.4
   2 Komodo-Dragon_2.5      :  3604.6   13.3   338.5     500    68     100  194  289   17    58  3470.6
   3 Komodo-Dragon_2        :  3542.5   12.2   299.5     500    60     100  137  325   38    65  3470.6
   4 Ethereal_13.25-NNUE    :  3446.4    8.2   250.0     600    42      92   60  380  160    63  3508.9
   5 Koivisto_6.16          :  3434.6    9.6   239.0     600    40      70   65  348  187    58  3510.9
   6 SlowChess_2.7          :  3430.2   10.7   235.0     600    39      79   52  366  182    61  3511.6
   7 RubiChess_2.2          :  3424.2   11.1   229.5     600    38     ---   46  367  187    61  3512.6

Code:
Engine           Games   Loss  Percent
Stockfish 14      600     10    1.66%
Dragon 2.5        500     17    3,40%
Dragon 2          500     38    7,60%
Ethereal 13.25    600    160   26.66%
Koivisto 6.16     600    187   31.16%
SlowChess 2.7     600    182   30.33%
RubiChess 2.2     600    187   31.16%

Unfortunately no SF15 and with Komodo could close in but even with Sf14 the pattern remains, SF hardly loses a game.

Secondly, Graham recently played a 80 game match between SF 15 and Dragon 3, 16 cores, time control 2 hours + 30 second increments

The result SF won 3, lost none and 77 draws.

http://talkchess.com/forum3/viewtopic.php?f=6&t=80053&sid=049ed403a77aca07119652211faa29e1

SF is way stronger, in no circumstance this will happen the other way around.

The current elo system has reached its limits, engines have become too strong.

TheSelfImprover likes this post

Back to top Go down
http://rebel13.nl/
mwyoung

mwyoung


Posts : 880
Join date : 2020-11-25
Location : USA

I wonder what is the reason that so many cores work for stockfish - Page 2 Empty
PostSubject: Re: I wonder what is the reason that so many cores work for stockfish   I wonder what is the reason that so many cores work for stockfish - Page 2 EmptyThu Jul 28, 2022 9:56 pm

Admin wrote:
And now you made Ed angry Very Happy

GRL - 20 cores

Code:
   # PLAYER                 :  RATING  ERROR  POINTS  PLAYED   (%)  CFS(%)    W    D    L  D(%)  OppAvg
   1 Stockfish_14           :  3617.5    9.7   408.5     600    68      96  227  363   10    61  3480.4
   2 Komodo-Dragon_2.5      :  3604.6   13.3   338.5     500    68     100  194  289   17    58  3470.6
   3 Komodo-Dragon_2        :  3542.5   12.2   299.5     500    60     100  137  325   38    65  3470.6
   4 Ethereal_13.25-NNUE    :  3446.4    8.2   250.0     600    42      92   60  380  160    63  3508.9
   5 Koivisto_6.16          :  3434.6    9.6   239.0     600    40      70   65  348  187    58  3510.9
   6 SlowChess_2.7          :  3430.2   10.7   235.0     600    39      79   52  366  182    61  3511.6
   7 RubiChess_2.2          :  3424.2   11.1   229.5     600    38     ---   46  367  187    61  3512.6

Code:
Engine           Games   Loss  Percent
Stockfish 14      600     10    1.66%
Dragon 2.5        500     17    3,40%
Dragon 2          500     38    7,60%
Ethereal 13.25    600    160   26.66%
Koivisto 6.16     600    187   31.16%
SlowChess 2.7     600    182   30.33%
RubiChess 2.2     600    187   31.16%

Unfortunately no SF15 and with Komodo could close in but even with Sf14 the pattern remains, SF hardly loses a game.

Secondly, Graham recently played a 80 game match between SF 15 and Dragon 3, 16 cores, time control 2 hours + 30 second increments

The result SF won 3, lost none and 77 draws.

http://talkchess.com/forum3/viewtopic.php?f=6&t=80053&sid=049ed403a77aca07119652211faa29e1

SF is way stronger, in no circumstance this will happen the other way around.

The current elo system has reached its limits, engines have become too strong.

Ed we are comparing ratings gap or also know as performance as core and/or time increase.

As you see I am exactly correct in what I am saying and showing.

For example in your data at 40/2 time controls.

Stockfish 14 at 20 cores Rating 3617.5
Dragon 2.5  at 20 cores Rating 3604.6

Difference at 20 cores 12.9 Elo

Stockfish 14 at 1 core 3689.3
Komodo-Dragon 2.5 at 1 core 3648.1

Difference at 1 core 41.2 Elo

"The current elo system has reached its limits, engines have become too strong."

Ed what are you talking about. That statement makes no sense.

"Secondly, Graham recently played a 80 game match between SF 15 and Dragon 3, 16 cores, time control 2 hours + 30 second increments

The result SF won 3, lost none and 77 draws."

I think we need more games to draw a conclusion.

But we do see also from CCRL.

Time control: From Jan 2020, equivalent to 2'+1" on an Intel i7-4770K

1 Stockfish 15 64-bit 8CPU 3744 +12 −12 66.5% −94.3 66.7% 1975
99.5%
2 Dragon by Komodo 3 64-bit 8CPU 3718 +16 −16 61.2% −61.7 72.5% 996
7


Ratings difference 26 Elo


Time control: Equivalent to 40 moves in 15 minutes on an Intel i7-4770k.

1 Stockfish 14 64-bit 4CPU 3538 +18 −18 66.9% −101.2 65.0% 874
68.6%
2 Dragon by Komodo 3 64-bit 4CPU 3532 +16 −16 63.0% −74.5 73.3% 1012
94.8%

Rating difference 6 Elo And now within the margin of error.

Now lets see how all the best stockfish's did playing Komodo 3 in each match.

I wonder what is the reason that so many cores work for stockfish - Page 2 K_m_r10

Stockfish did not win a single game playing Komodo 3.

I think my statement holds water!

"You are making a very bad assumption. You are assuming the whole truth can be found with only a 1 tier testing standard. In your rating list that is 40/2 and 1 core.

Stockfish NNUE is a very fast engine, but a horrible scaling engine. This allows other engines to catch up."
Back to top Go down
Admin
Admin
Admin


Posts : 2608
Join date : 2020-11-17
Location : Netherlands

I wonder what is the reason that so many cores work for stockfish - Page 2 Empty
PostSubject: Re: I wonder what is the reason that so many cores work for stockfish   I wonder what is the reason that so many cores work for stockfish - Page 2 EmptyFri Jul 29, 2022 12:12 pm

You are totally missing my points.

1. As for Graham test, 16 cores, tc 2 hours + 30 second increments, do you have any idea how strong the top engines play? Beyond our understanding, agree? With such hardware and time control it's only logical other engines catch up, do you understand that? Meaning, your conclusion that SF scales bad is pure speculation. We arrived at a playing strength the elo system (basically 1%=7 elo points) was never build for.

2. We need more data to make better assumptions. Here is one. In ProDeo I kept a statistic at which depth the engine still changed the best move, example from ~2005

Code:
                                    SEARCH OVERVIEW        ProDeo 1.6
                                    ===============        

Depth           Moves                    Moves                   Moves    
               Changed                  Changed                 Changed  
             Middle Game              End Game (1)            End Game (2)

 1      34228 - 29886 = 87.3%     16906 - 10766 = 63.7%     13273 - 7262 = 54.7%
 2      33402 - 16295 = 48.8%     16163 - 3713 = 23.0%     12732 - 3189 = 25.0%
 3      28353 - 11075 = 39.1%     13365 - 2350 = 17.6%      9839 - 2414 = 24.5%
 4      25356 - 7948 = 31.3%     12088 - 1948 = 16.1%      8373 - 1679 = 20.1%
 5      15013 - 4060 = 27.0%     11976 - 1723 = 14.4%      8039 - 1346 = 16.7%
 6      13949 - 3731 = 26.7%     11848 - 1467 = 12.4%      7871 - 1228 = 15.6%
 7      13860 - 3254 = 23.5%     11786 - 1645 = 14.0%      7743 - 1235 = 15.9%
 8      13521 - 2845 = 21.0%     11728 - 1104 =  9.4%      7569 - 1101 = 14.5%
 9      12840 - 2872 = 22.4%     11656 - 1113 =  9.5%      7266 - 1160 = 16.0%
10      11546 - 2473 = 21.4%     11464 - 804 =  7.0%      7118 - 1072 = 15.1%
11       9961 - 2074 = 20.8%     11020 - 978 =  8.9%      6855 - 903 = 13.2%
12       7293 - 1396 = 19.1%      9943 - 543 =  5.5%      6553 - 799 = 12.2%
13       5158 - 922 = 17.9%      9215 - 424 =  4.6%      6126 - 661 = 10.8%
14       3113 - 431 = 13.8%      8388 - 288 =  3.4%      5596 - 689 = 12.3%
15       1541 - 141 =  9.1%      7656 - 218 =  2.8%      4817 - 598 = 12.4%
16        767 - 26 =  3.4%      6960 - 102 =  1.5%      3907 - 386 =  9.9%
17        516 - 4 =  0.8%      6501 - 36 =  0.6%      3145 - 233 =  7.4%
18        461 - 0 =  0.0%      6252 - 21 =  0.3%      2558 - 159 =  6.2%
19        454 - 0 =  0.0%      6131 - 11 =  0.2%      2105 - 80 =  3.8%
20        448 - 0 =  0.0%      6063 - 7 =  0.1%      1819 - 49 =  2.7%
21        444 - 0 =  0.0%      6005 - 2 =  0.0%      1638 - 30 =  1.8%
22        443 - 0 =  0.0%      5965 - 1 =  0.0%      1508 - 18 =  1.2%
23        441 - 0 =  0.0%      5938 - 1 =  0.0%      1417 - 14 =  1.0%
24        437 - 0 =  0.0%      5918 - 1 =  0.0%      1339 - 13 =  1.0%
25        435 - 0 =  0.0%      5896 - 0 =  0.0%      1261 - 6 =  0.5%
26        434 - 0 =  0.0%      5876 - 1 =  0.0%      1208 - 6 =  0.5%
27        434 - 0 =  0.0%      5862 - 0 =  0.0%      1183 - 3 =  0.3%
28        433 - 0 =  0.0%      5850 - 0 =  0.0%      1170 - 0 =  0.0%
29        430 - 0 =  0.0%      5839 - 0 =  0.0%      1161 - 2 =  0.2%
30        430 - 0 =  0.0%      5834 - 0 =  0.0%      1155 - 2 =  0.2%
31        429 - 0 =  0.0%      5826 - 0 =  0.0%      1150 - 3 =  0.3%
32        428 - 0 =  0.0%      5816 - 0 =  0.0%      1139 - 0 =  0.0%
33        427 - 0 =  0.0%      5812 - 0 =  0.0%      1131 - 0 =  0.0%
34        426 - 0 =  0.0%      5808 - 0 =  0.0%      1130 - 0 =  0.0%
35        424 - 0 =  0.0%      5801 - 0 =  0.0%      1127 - 0 =  0.0%
36        424 - 0 =  0.0%      5794 - 0 =  0.0%      1124 - 0 =  0.0%
37        424 - 0 =  0.0%      5785 - 0 =  0.0%      1121 - 0 =  0.0%
38        423 - 0 =  0.0%      5777 - 0 =  0.0%      1120 - 0 =  0.0%
39        423 - 0 =  0.0%      5772 - 0 =  0.0%      1115 - 0 =  0.0%
40        423 - 0 =  0.0%      5765 - 0 =  0.0%      1114 - 0 =  0.0%
41        423 - 0 =  0.0%      5764 - 0 =  0.0%      1113 - 0 =  0.0%
42        423 - 0 =  0.0%      5759 - 0 =  0.0%      1111 - 0 =  0.0%
43        421 - 0 =  0.0%      5757 - 0 =  0.0%      1106 - 0 =  0.0%
44        420 - 0 =  0.0%      5755 - 0 =  0.0%      1104 - 0 =  0.0%
45        420 - 0 =  0.0%      5749 - 0 =  0.0%      1100 - 0 =  0.0%
46        420 - 0 =  0.0%      5743 - 0 =  0.0%      1100 - 0 =  0.0%
47        420 - 0 =  0.0%      5740 - 0 =  0.0%      1099 - 0 =  0.0%
48        419 - 0 =  0.0%      5736 - 0 =  0.0%      1099 - 0 =  0.0%
49        417 - 0 =  0.0%      5735 - 0 =  0.0%      1096 - 0 =  0.0%
50        414 - 0 =  0.0%      5732 - 0 =  0.0%      1093 - 0 =  0.0%
51        414 - 0 =  0.0%      5729 - 0 =  0.0%      1091 - 0 =  0.0%
52        414 - 0 =  0.0%      5724 - 0 =  0.0%      1090 - 0 =  0.0%
53        414 - 0 =  0.0%      5720 - 0 =  0.0%      1087 - 0 =  0.0%
54        413 - 0 =  0.0%      5718 - 0 =  0.0%      1086 - 0 =  0.0%
55        412 - 0 =  0.0%      5716 - 0 =  0.0%      1084 - 0 =  0.0%
56        411 - 0 =  0.0%      5715 - 0 =  0.0%      1082 - 0 =  0.0%
57        411 - 0 =  0.0%      5710 - 0 =  0.0%      1080 - 0 =  0.0%
58        411 - 0 =  0.0%      5704 - 0 =  0.0%      1078 - 0 =  0.0%
59        411 - 0 =  0.0%      5702 - 0 =  0.0%      1077 - 0 =  0.0%
60        410 - 0 =  0.0%      5700 - 0 =  0.0%      1075 - 0 =  0.0%

I want to see such statistic of SF and Dragon. If SF [1 core] hardly changes any moves in comparison with SF [4 cores] while Dragon changes more frequently then you would have a point.

3. Question for you.
Match-one, SF1 vs Dragon, 1000 games, result 550 - 450, W=100 | D=900 | L=0
Match-two, SF2 vs Dragon, 1000 games, result 550 - 450, W=200 | D=700 | L=100

Which engine is stronger, SF1 or SF2 ?

Back to top Go down
http://rebel13.nl/
TheSelfImprover

TheSelfImprover


Posts : 3112
Join date : 2020-11-18

I wonder what is the reason that so many cores work for stockfish - Page 2 Empty
PostSubject: Re: I wonder what is the reason that so many cores work for stockfish   I wonder what is the reason that so many cores work for stockfish - Page 2 EmptyFri Jul 29, 2022 1:23 pm

Admin wrote:
The current elo system has reached its limits, engines have become too strong.


This is obviously not scientifically rigorous, but given how difficult SF has become to beat, maybe we can get an indication of the maximum possible rating under the elo system:

* get a list of SF games against opponents which are within 100 elo of it

* for each available range, (e.g. 2600-2700, 2700-2800 etc) find SF's loss rate

Having tabulated this data, and maybe drawn a chart for it (spreadsheet or chart-drawing web page), maybe we can see a pattern and make an informed guess as to where this loss rate will go to zero. At that point, it will be close to being unbeatable, and the maximum elo rating that will ever be achieved will probably not be much higher than that.

At the heart of chess are two fundamnetal imperatives:

1. avoid making a mistake

2. find the win when your opponent makes a mistakes

It looks as though SF is making sound progress on the first one.
Back to top Go down
Admin
Admin
Admin


Posts : 2608
Join date : 2020-11-17
Location : Netherlands

I wonder what is the reason that so many cores work for stockfish - Page 2 Empty
PostSubject: Re: I wonder what is the reason that so many cores work for stockfish   I wonder what is the reason that so many cores work for stockfish - Page 2 EmptyFri Jul 29, 2022 8:02 pm

TheSelfImprover wrote:
Admin wrote:
The current elo system has reached its limits, engines have become too strong.


This is obviously not scientifically rigorous, but given how difficult SF has become to beat, maybe we can get an indication of the maximum possible rating under the elo system:

* get a list of SF games against opponents which are within 100 elo of it

* for each available range, (e.g. 2600-2700, 2700-2800 etc) find SF's loss rate

That's basically what I have already done with SF11.

http://rebel13.nl/rebel13/stockfish-handicap-matches.html

For today I suggest (Mark?) to run a 10 or 20 game match in Arena with debug=on which creates a log file with all main lines. From the Arena log file I can create a ProDeo 1.6 statistic (see my previous post) that reveals the scaling of both engines.

Suggested match, SF15 vs Dragon 3, 16 cores, 4Gb hash, tc: 2 hours + 30 second increments

10 games will produce ~750 positions with full main lines for each engine, 20 games 1500 positions to create a reasonable scaling statistic.
Back to top Go down
http://rebel13.nl/
mwyoung

mwyoung


Posts : 880
Join date : 2020-11-25
Location : USA

I wonder what is the reason that so many cores work for stockfish - Page 2 Empty
PostSubject: Re: I wonder what is the reason that so many cores work for stockfish   I wonder what is the reason that so many cores work for stockfish - Page 2 EmptyFri Jul 29, 2022 8:51 pm

Admin wrote:
You are totally missing my points.

1. As for Graham test, 16 cores, tc 2 hours + 30 second increments, do you have any idea how strong the top engines play? Beyond our understanding, agree? With such hardware and time control it's only logical other engines catch up, do you understand that? Meaning, your conclusion that SF scales bad is pure speculation. We arrived at a playing strength the elo system (basically 1%=7 elo points) was never build for.

2. We need more data to make better assumptions. Here is one. In ProDeo I kept a statistic at which depth the engine still changed the best move, example from ~2005

Code:
                                    SEARCH OVERVIEW        ProDeo 1.6
                                    ===============        

Depth           Moves                    Moves                   Moves    
               Changed                  Changed                 Changed  
             Middle Game              End Game (1)            End Game (2)

 1      34228 - 29886 = 87.3%  16906 - 10766 = 63.7%  13273 - 7262 = 54.7%
 2      33402 - 16295 = 48.8%  16163 - 3713 = 23.0%  12732 - 3189 = 25.0%
 3      28353 - 11075 = 39.1%  13365 - 2350 = 17.6%   9839 - 2414 = 24.5%
 4      25356 - 7948 = 31.3%  12088 - 1948 = 16.1%   8373 - 1679 = 20.1%
 5      15013 - 4060 = 27.0%  11976 - 1723 = 14.4%   8039 - 1346 = 16.7%
 6      13949 - 3731 = 26.7%  11848 - 1467 = 12.4%   7871 - 1228 = 15.6%
 7      13860 - 3254 = 23.5%  11786 - 1645 = 14.0%   7743 - 1235 = 15.9%
 8      13521 - 2845 = 21.0%  11728 - 1104 =  9.4%   7569 - 1101 = 14.5%
 9      12840 - 2872 = 22.4%  11656 - 1113 =  9.5%   7266 - 1160 = 16.0%
10      11546 - 2473 = 21.4%  11464 - 804 =  7.0%   7118 - 1072 = 15.1%
11       9961 - 2074 = 20.8%  11020 - 978 =  8.9%   6855 - 903 = 13.2%
12       7293 - 1396 = 19.1%   9943 - 543 =  5.5%   6553 - 799 = 12.2%
13       5158 - 922 = 17.9%   9215 - 424 =  4.6%   6126 - 661 = 10.8%
14       3113 - 431 = 13.8%   8388 - 288 =  3.4%   5596 - 689 = 12.3%
15       1541 - 141 =  9.1%   7656 - 218 =  2.8%   4817 - 598 = 12.4%
16        767 - 26 =  3.4%   6960 - 102 =  1.5%   3907 - 386 =  9.9%
17        516 - 4 =  0.8%   6501 - 36 =  0.6%   3145 - 233 =  7.4%
18        461 - 0 =  0.0%   6252 - 21 =  0.3%   2558 - 159 =  6.2%
19        454 - 0 =  0.0%   6131 - 11 =  0.2%   2105 - 80 =  3.8%
20        448 - 0 =  0.0%   6063 - 7 =  0.1%   1819 - 49 =  2.7%
21        444 - 0 =  0.0%   6005 - 2 =  0.0%   1638 - 30 =  1.8%
22        443 - 0 =  0.0%   5965 - 1 =  0.0%   1508 - 18 =  1.2%
23        441 - 0 =  0.0%   5938 - 1 =  0.0%   1417 - 14 =  1.0%
24        437 - 0 =  0.0%   5918 - 1 =  0.0%   1339 - 13 =  1.0%
25        435 - 0 =  0.0%   5896 - 0 =  0.0%   1261 - 6 =  0.5%
26        434 - 0 =  0.0%   5876 - 1 =  0.0%   1208 - 6 =  0.5%
27        434 - 0 =  0.0%   5862 - 0 =  0.0%   1183 - 3 =  0.3%
28        433 - 0 =  0.0%   5850 - 0 =  0.0%   1170 - 0 =  0.0%
29        430 - 0 =  0.0%   5839 - 0 =  0.0%   1161 - 2 =  0.2%
30        430 - 0 =  0.0%   5834 - 0 =  0.0%   1155 - 2 =  0.2%
31        429 - 0 =  0.0%   5826 - 0 =  0.0%   1150 - 3 =  0.3%
32        428 - 0 =  0.0%   5816 - 0 =  0.0%   1139 - 0 =  0.0%
33        427 - 0 =  0.0%   5812 - 0 =  0.0%   1131 - 0 =  0.0%
34        426 - 0 =  0.0%   5808 - 0 =  0.0%   1130 - 0 =  0.0%
35        424 - 0 =  0.0%   5801 - 0 =  0.0%   1127 - 0 =  0.0%
36        424 - 0 =  0.0%   5794 - 0 =  0.0%   1124 - 0 =  0.0%
37        424 - 0 =  0.0%   5785 - 0 =  0.0%   1121 - 0 =  0.0%
38        423 - 0 =  0.0%   5777 - 0 =  0.0%   1120 - 0 =  0.0%
39        423 - 0 =  0.0%   5772 - 0 =  0.0%   1115 - 0 =  0.0%
40        423 - 0 =  0.0%   5765 - 0 =  0.0%   1114 - 0 =  0.0%
41        423 - 0 =  0.0%   5764 - 0 =  0.0%   1113 - 0 =  0.0%
42        423 - 0 =  0.0%   5759 - 0 =  0.0%   1111 - 0 =  0.0%
43        421 - 0 =  0.0%   5757 - 0 =  0.0%   1106 - 0 =  0.0%
44        420 - 0 =  0.0%   5755 - 0 =  0.0%   1104 - 0 =  0.0%
45        420 - 0 =  0.0%   5749 - 0 =  0.0%   1100 - 0 =  0.0%
46        420 - 0 =  0.0%   5743 - 0 =  0.0%   1100 - 0 =  0.0%
47        420 - 0 =  0.0%   5740 - 0 =  0.0%   1099 - 0 =  0.0%
48        419 - 0 =  0.0%   5736 - 0 =  0.0%   1099 - 0 =  0.0%
49        417 - 0 =  0.0%   5735 - 0 =  0.0%   1096 - 0 =  0.0%
50        414 - 0 =  0.0%   5732 - 0 =  0.0%   1093 - 0 =  0.0%
51        414 - 0 =  0.0%   5729 - 0 =  0.0%   1091 - 0 =  0.0%
52        414 - 0 =  0.0%   5724 - 0 =  0.0%   1090 - 0 =  0.0%
53        414 - 0 =  0.0%   5720 - 0 =  0.0%   1087 - 0 =  0.0%
54        413 - 0 =  0.0%   5718 - 0 =  0.0%   1086 - 0 =  0.0%
55        412 - 0 =  0.0%   5716 - 0 =  0.0%   1084 - 0 =  0.0%
56        411 - 0 =  0.0%   5715 - 0 =  0.0%   1082 - 0 =  0.0%
57        411 - 0 =  0.0%   5710 - 0 =  0.0%   1080 - 0 =  0.0%
58        411 - 0 =  0.0%   5704 - 0 =  0.0%   1078 - 0 =  0.0%
59        411 - 0 =  0.0%   5702 - 0 =  0.0%   1077 - 0 =  0.0%
60        410 - 0 =  0.0%   5700 - 0 =  0.0%   1075 - 0 =  0.0%

I want to see such statistic of SF and Dragon. If SF [1 core] hardly changes any moves in comparison with SF [4 cores] while Dragon changes more frequently then you would have a point.

3. Question for you.
Match-one, SF1 vs Dragon, 1000 games, result 550 - 450, W=100 | D=900 | L=0
Match-two, SF2 vs Dragon, 1000 games, result 550 - 450, W=200 | D=700 | L=100

Which engine is stronger, SF1 or SF2 ?


1. Is not speculation. As I have given evidence in many forms. So that is my conclusion based on the evidence. And the results back up my conclusion.  

Here is what 16 cores gives you with Stockfish.
Test 5.

Bench 1024 16 26.

2950X
Total time (ms) : 129677
Nodes searched  : 3077330368
Nodes/second    : 23730733

Winner 5900HX 8 cores vs 2950X 16 cores.
Even with the 2950X searching many more NPS

Winner 5900HX 4 cores vs 2950X 16 cores.
Even with the 2950X 16 cores searching 23.7 Mnps vs 5900HX 4 cores searching 9.8 Mnps.
lol!

https://prodeo.actieforum.com/t872-stockfish-what-is-really-better-faster-cores-more-cores-and-nps-vs-time-to-depth
https://prodeo.actieforum.com/t311-stockfish-070321-scaling-tournament-1-to-16-cpu-cores
https://prodeo.actieforum.com/t541-stockfish-17-09-21-scaling-test-1-to-16-cpu-cores

2. Is meaningless. It does not matter how many times, or how often a engine changes the "best" move. What matters is did the move change the balance of the position. Like changing a loss to a draw. And that is measured with ELO. It is called performance.

3.3. Question for you.
Match-one, SF1 vs Dragon, 1000 games, result 550 - 450, W=100 | D=900 | L=0
Match-two, SF2 vs Dragon, 1000 games, result 550 - 450, W=200 | D=700 | L=100

Which engine is stronger, SF1 or SF2 ?

They are the same in regards to TPR.

I do not know what peoples fascination is with draw rate. As that is a totally meaningless stat. In regards to performance.
Back to top Go down
Admin
Admin
Admin


Posts : 2608
Join date : 2020-11-17
Location : Netherlands

I wonder what is the reason that so many cores work for stockfish - Page 2 Empty
PostSubject: Re: I wonder what is the reason that so many cores work for stockfish   I wonder what is the reason that so many cores work for stockfish - Page 2 EmptyFri Jul 29, 2022 9:35 pm

Okay Mark, I give up.
Back to top Go down
http://rebel13.nl/
Uri Blass




Posts : 207
Join date : 2020-11-28

I wonder what is the reason that so many cores work for stockfish - Page 2 Empty
PostSubject: Re: I wonder what is the reason that so many cores work for stockfish   I wonder what is the reason that so many cores work for stockfish - Page 2 EmptyFri Jul 29, 2022 11:11 pm

mwyoung wrote:
Admin wrote:
And now you made Ed angry Very Happy

GRL - 20 cores

Code:
   # PLAYER                 :  RATING  ERROR  POINTS  PLAYED   (%)  CFS(%)    W    D    L  D(%)  OppAvg
   1 Stockfish_14           :  3617.5    9.7   408.5     600    68      96  227  363   10    61  3480.4
   2 Komodo-Dragon_2.5      :  3604.6   13.3   338.5     500    68     100  194  289   17    58  3470.6
   3 Komodo-Dragon_2        :  3542.5   12.2   299.5     500    60     100  137  325   38    65  3470.6
   4 Ethereal_13.25-NNUE    :  3446.4    8.2   250.0     600    42      92   60  380  160    63  3508.9
   5 Koivisto_6.16          :  3434.6    9.6   239.0     600    40      70   65  348  187    58  3510.9
   6 SlowChess_2.7          :  3430.2   10.7   235.0     600    39      79   52  366  182    61  3511.6
   7 RubiChess_2.2          :  3424.2   11.1   229.5     600    38     ---   46  367  187    61  3512.6

Code:
Engine           Games   Loss  Percent
Stockfish 14      600     10    1.66%
Dragon 2.5        500     17    3,40%
Dragon 2          500     38    7,60%
Ethereal 13.25    600    160   26.66%
Koivisto 6.16     600    187   31.16%
SlowChess 2.7     600    182   30.33%
RubiChess 2.2     600    187   31.16%

Unfortunately no SF15 and with Komodo could close in but even with Sf14 the pattern remains, SF hardly loses a game.

Secondly, Graham recently played a 80 game match between SF 15 and Dragon 3, 16 cores, time control 2 hours + 30 second increments

The result SF won 3, lost none and 77 draws.

http://talkchess.com/forum3/viewtopic.php?f=6&t=80053&sid=049ed403a77aca07119652211faa29e1

SF is way stronger, in no circumstance this will happen the other way around.

The current elo system has reached its limits, engines have become too strong.

Ed we are comparing ratings gap or also know as performance as core and/or time increase.

As you see I am exactly correct in what I am saying and showing.

For example in your data at 40/2 time controls.

Stockfish 14 at 20 cores Rating 3617.5
Dragon 2.5  at 20 cores Rating 3604.6

Difference at 20 cores 12.9 Elo

Stockfish 14 at 1 core 3689.3
Komodo-Dragon 2.5 at 1 core 3648.1

Difference at 1 core 41.2 Elo

"The current elo system has reached its limits, engines have become too strong."

Ed what are you talking about. That statement makes no sense.

"Secondly, Graham recently played a 80 game match between SF 15 and Dragon 3, 16 cores, time control 2 hours + 30 second increments

The result SF won 3, lost none and 77 draws."

I think we need more games to draw a conclusion.

But we do see also from CCRL.

Time control: From Jan 2020, equivalent to 2'+1" on an Intel i7-4770K

1 Stockfish 15 64-bit 8CPU 3744 +12 −12 66.5% −94.3 66.7% 1975
99.5%
2 Dragon by Komodo 3 64-bit 8CPU 3718 +16 −16 61.2% −61.7 72.5% 996
7


Ratings difference 26 Elo


Time control: Equivalent to 40 moves in 15 minutes on an Intel i7-4770k.

1 Stockfish 14 64-bit 4CPU 3538 +18 −18 66.9% −101.2 65.0% 874
68.6%
2 Dragon by Komodo 3 64-bit 4CPU 3532 +16 −16 63.0% −74.5 73.3% 1012
94.8%

Rating difference 6 Elo And now within the margin of error.

Now lets see how all the best stockfish's did playing Komodo 3 in each match.

I wonder what is the reason that so many cores work for stockfish - Page 2 K_m_r10

Stockfish did not win a single game playing Komodo 3.

I think my statement holds water!

"You are making a very bad assumption. You are assuming the whole truth can be found with only a 1 tier testing standard. In your rating list that is 40/2 and 1 core.

Stockfish NNUE is a very fast engine, but a horrible scaling engine. This allows other engines to catch up."

If stockfish NNUE is a horrible scaling engine relative to other engines then I guess other engines should be able to win a match against stockfish and not only draw against it if the time control is long enough.

Do you know about time control that is long enough to cause stockfish to lose games?
If yes then I would like to see games.

Admin likes this post

Back to top Go down
mwyoung

mwyoung


Posts : 880
Join date : 2020-11-25
Location : USA

I wonder what is the reason that so many cores work for stockfish - Page 2 Empty
PostSubject: Re: I wonder what is the reason that so many cores work for stockfish   I wonder what is the reason that so many cores work for stockfish - Page 2 EmptyFri Jul 29, 2022 11:25 pm

Uri Blass wrote:
mwyoung wrote:
Admin wrote:
And now you made Ed angry Very Happy

GRL - 20 cores

Code:
   # PLAYER                 :  RATING  ERROR  POINTS  PLAYED   (%)  CFS(%)    W    D    L  D(%)  OppAvg
   1 Stockfish_14           :  3617.5    9.7   408.5     600    68      96  227  363   10    61  3480.4
   2 Komodo-Dragon_2.5      :  3604.6   13.3   338.5     500    68     100  194  289   17    58  3470.6
   3 Komodo-Dragon_2        :  3542.5   12.2   299.5     500    60     100  137  325   38    65  3470.6
   4 Ethereal_13.25-NNUE    :  3446.4    8.2   250.0     600    42      92   60  380  160    63  3508.9
   5 Koivisto_6.16          :  3434.6    9.6   239.0     600    40      70   65  348  187    58  3510.9
   6 SlowChess_2.7          :  3430.2   10.7   235.0     600    39      79   52  366  182    61  3511.6
   7 RubiChess_2.2          :  3424.2   11.1   229.5     600    38     ---   46  367  187    61  3512.6

Code:
Engine           Games   Loss  Percent
Stockfish 14      600     10    1.66%
Dragon 2.5        500     17    3,40%
Dragon 2          500     38    7,60%
Ethereal 13.25    600    160   26.66%
Koivisto 6.16     600    187   31.16%
SlowChess 2.7     600    182   30.33%
RubiChess 2.2     600    187   31.16%

Unfortunately no SF15 and with Komodo could close in but even with Sf14 the pattern remains, SF hardly loses a game.

Secondly, Graham recently played a 80 game match between SF 15 and Dragon 3, 16 cores, time control 2 hours + 30 second increments

The result SF won 3, lost none and 77 draws.

http://talkchess.com/forum3/viewtopic.php?f=6&t=80053&sid=049ed403a77aca07119652211faa29e1

SF is way stronger, in no circumstance this will happen the other way around.

The current elo system has reached its limits, engines have become too strong.

Ed we are comparing ratings gap or also know as performance as core and/or time increase.

As you see I am exactly correct in what I am saying and showing.

For example in your data at 40/2 time controls.

Stockfish 14 at 20 cores Rating 3617.5
Dragon 2.5  at 20 cores Rating 3604.6

Difference at 20 cores 12.9 Elo

Stockfish 14 at 1 core 3689.3
Komodo-Dragon 2.5 at 1 core 3648.1

Difference at 1 core 41.2 Elo

"The current elo system has reached its limits, engines have become too strong."

Ed what are you talking about. That statement makes no sense.

"Secondly, Graham recently played a 80 game match between SF 15 and Dragon 3, 16 cores, time control 2 hours + 30 second increments

The result SF won 3, lost none and 77 draws."

I think we need more games to draw a conclusion.

But we do see also from CCRL.

Time control: From Jan 2020, equivalent to 2'+1" on an Intel i7-4770K

1 Stockfish 15 64-bit 8CPU 3744 +12 −12 66.5% −94.3 66.7% 1975
99.5%
2 Dragon by Komodo 3 64-bit 8CPU 3718 +16 −16 61.2% −61.7 72.5% 996
7


Ratings difference 26 Elo


Time control: Equivalent to 40 moves in 15 minutes on an Intel i7-4770k.

1 Stockfish 14 64-bit 4CPU 3538 +18 −18 66.9% −101.2 65.0% 874
68.6%
2 Dragon by Komodo 3 64-bit 4CPU 3532 +16 −16 63.0% −74.5 73.3% 1012
94.8%

Rating difference 6 Elo And now within the margin of error.

Now lets see how all the best stockfish's did playing Komodo 3 in each match.

I wonder what is the reason that so many cores work for stockfish - Page 2 K_m_r10

Stockfish did not win a single game playing Komodo 3.

I think my statement holds water!

"You are making a very bad assumption. You are assuming the whole truth can be found with only a 1 tier testing standard. In your rating list that is 40/2 and 1 core.

Stockfish NNUE is a very fast engine, but a horrible scaling engine. This allows other engines to catch up."

If stockfish NNUE is a horrible scaling engine relative to other engines then I guess other engines should be able to win a match against stockfish and not only draw against it if the time control is long enough.

Do you know about time control that is long enough to cause stockfish to lose games?
If yes then I would like to see games.



It will take a long time in some cases. You can calculate how long if you know the scaling curve of each engine.
And in some cases it will never happen depending on the engine. As stockfish still gets stronger.

Take Lc0 for example, it is almost the exact opposite of Stockfish. Slow, and not a fast calculator like Stockfish. But Lc0 can out perform stockfish given time.

Or take a Shannon type A chess engine vs Stockfish. At any normal time control Stock wins 100% of the time. But given almost unlimited time, or a super fast computer. Beyond physical limits Stockfish would lose, and mostly likely lose badly.


Last edited by mwyoung on Fri Jul 29, 2022 11:44 pm; edited 1 time in total
Back to top Go down
TheSelfImprover

TheSelfImprover


Posts : 3112
Join date : 2020-11-18

I wonder what is the reason that so many cores work for stockfish - Page 2 Empty
PostSubject: Re: I wonder what is the reason that so many cores work for stockfish   I wonder what is the reason that so many cores work for stockfish - Page 2 EmptyFri Jul 29, 2022 11:41 pm

mwyoung wrote:
I do not know what peoples fascination is with draw rate. As that is a totally meaningless stat. In regards to performance.


Hmmm...

Engine A gets 10 wins over engine B in 10 games. Looks as though engine A is better than engine B.

Engine A gets 10 wins and 990 draws with engine B in 1000 games. Looks as though the two engines are roughly equal.

That's performance. The bigger issue is that the proportion of draws between top engines is steadily rising. IMO this is telling us something fundamentally important about the game of chess: it's not proof, of course, but the implication is that without a mistake, a game cannot be won - and the mistakes are getting rarer.
Back to top Go down
Admin
Admin
Admin


Posts : 2608
Join date : 2020-11-17
Location : Netherlands

I wonder what is the reason that so many cores work for stockfish - Page 2 Empty
PostSubject: Re: I wonder what is the reason that so many cores work for stockfish   I wonder what is the reason that so many cores work for stockfish - Page 2 EmptySat Jul 30, 2022 12:01 am

TheSelfImprover wrote:
mwyoung wrote:
I do not know what peoples fascination is with draw rate. As that is a totally meaningless stat. In regards to performance.


Hmmm...

Engine A gets 10 wins over engine B in 10 games. Looks as though engine A is better than engine B.

Engine A gets 10 wins and 990 draws with engine B in 1000 games. Looks as though the two engines are roughly equal.

That's performance. The bigger issue is that the proportion of draws between top engines is steadily rising. IMO this is telling us something fundamentally important about the game of chess: it's not proof, of course, but the implication is that without a mistake, a game cannot be won - and the mistakes are getting rarer.

I think the whole issue is food for the better mathematicians among us, there must be a better formula than 1% = 7 elo.
Back to top Go down
http://rebel13.nl/
mwyoung

mwyoung


Posts : 880
Join date : 2020-11-25
Location : USA

I wonder what is the reason that so many cores work for stockfish - Page 2 Empty
PostSubject: Re: I wonder what is the reason that so many cores work for stockfish   I wonder what is the reason that so many cores work for stockfish - Page 2 EmptySat Jul 30, 2022 12:02 am

TheSelfImprover wrote:
mwyoung wrote:
I do not know what peoples fascination is with draw rate. As that is a totally meaningless stat. In regards to performance.


Hmmm...

Engine A gets 10 wins over engine B in 10 games. Looks as though engine A is better than engine B.

Engine A gets 10 wins and 990 draws with engine B in 1000 games. Looks as though the two engines are roughly equal.

That's performance. The bigger issue is that the proportion of draws between top engines is steadily rising. IMO this is telling us something fundamentally important about the game of chess: it's not proof, of course, but the implication is that without a mistake, a game cannot be won - and the mistakes are getting rarer.


Do not over think this....

Here is what I am saying.

1-0,0-1 = 1/2-1/2, 1/2-1/2

One has a 0% draw rate, one has 100% draw rate. But performance is equal.

And with more games, you are just repeating the above cycle.
Back to top Go down
mwyoung

mwyoung


Posts : 880
Join date : 2020-11-25
Location : USA

I wonder what is the reason that so many cores work for stockfish - Page 2 Empty
PostSubject: Re: I wonder what is the reason that so many cores work for stockfish   I wonder what is the reason that so many cores work for stockfish - Page 2 EmptySat Jul 30, 2022 12:31 am

TheSelfImprover wrote:
mwyoung wrote:
I do not know what peoples fascination is with draw rate. As that is a totally meaningless stat. In regards to performance.


IMO this is telling us something fundamentally important about the game of chess: it's not proof, of course, but the implication is that without a mistake, a game cannot be won - and the mistakes are getting rarer.

That is not knowable.

Just remember any 100% draw rate does not equal chess is a draw, or both player played perfectly.

Any more then a 100% winning score means chess is a win, or the winning player played perfectly.

And this includes stockfish, the only exception is when we have a table base solution.
Back to top Go down
Admin
Admin
Admin


Posts : 2608
Join date : 2020-11-17
Location : Netherlands

I wonder what is the reason that so many cores work for stockfish - Page 2 Empty
PostSubject: Re: I wonder what is the reason that so many cores work for stockfish   I wonder what is the reason that so many cores work for stockfish - Page 2 EmptySat Jul 30, 2022 6:40 am

mwyoung wrote:
Admin wrote:

3. Question for you.
Match-one, SF1 vs Dragon, 1000 games, result 550 - 450, W=100 | D=900 | L=0
Match-two, SF2 vs Dragon, 1000 games, result 550 - 450, W=200 | D=700 | L=100

Which engine is stronger, SF1 or SF2 ?


They are the same in regards to TPR.

I put the question on Talkchess.

http://talkchess.com/forum3/viewtopic.php?f=2&t=80384

9 votes so far and only one agrees with you scratch, 6 votes go to me Idea
Back to top Go down
http://rebel13.nl/
Chris Whittington




Posts : 1254
Join date : 2020-11-17
Location : France

I wonder what is the reason that so many cores work for stockfish - Page 2 Empty
PostSubject: Re: I wonder what is the reason that so many cores work for stockfish   I wonder what is the reason that so many cores work for stockfish - Page 2 EmptySat Jul 30, 2022 8:43 am

Admin wrote:
mwyoung wrote:
Admin wrote:

3. Question for you.
Match-one, SF1 vs Dragon, 1000 games, result 550 - 450, W=100 | D=900 | L=0
Match-two, SF2 vs Dragon, 1000 games, result 550 - 450, W=200 | D=700 | L=100

Which engine is stronger, SF1 or SF2 ?


They are the same in regards to TPR.

I put the question on Talkchess.

http://talkchess.com/forum3/viewtopic.php?f=2&t=80384

9 votes so far and only one agrees with you scratch, 6 votes go to me Idea

Both are same strength, by definition.
Back to top Go down
mwyoung

mwyoung


Posts : 880
Join date : 2020-11-25
Location : USA

I wonder what is the reason that so many cores work for stockfish - Page 2 Empty
PostSubject: Re: I wonder what is the reason that so many cores work for stockfish   I wonder what is the reason that so many cores work for stockfish - Page 2 EmptySat Jul 30, 2022 8:50 am

I wonder what is the reason that so many cores work for stockfish - Page 2 Wis10

Only 2 people out of 11 people on Talkchess got the correct answer!  That is telling.  lol!
Back to top Go down
Sponsored content





I wonder what is the reason that so many cores work for stockfish - Page 2 Empty
PostSubject: Re: I wonder what is the reason that so many cores work for stockfish   I wonder what is the reason that so many cores work for stockfish - Page 2 Empty

Back to top Go down
 
I wonder what is the reason that so many cores work for stockfish
Back to top 
Page 2 of 3Go to page : Previous  1, 2, 3  Next
 Similar topics
-
» Stockfish! What is really better? Faster Cores? More Cores?, and NPS vs. Time to Depth.
» Stockfish 17/09/21 Scaling Test. 1 to 16 CPU Cores.
» Stockfish 070321 Scaling Tournament, 1 to 16 CPU Cores.
» Stockfish 070321 Core Scaling Test. Cores for Testing 1, 2, 4, and 8.
» Testing the New Stockfish Net of 45 Mb, Stockfish 190521 vs Stockfish 13 (TC = 5m+5s) (32 Threads)

Permissions in this forum:You cannot reply to topics in this forum
ProDeo :: Computer Chess-
Jump to: