UCI_Elo

Ferdy · Post by **Ferdy** » Sat Jul 06, 2019 11:40 pm

Ferdy wrote: ↑Sat Jul 06, 2019 8:38 pm

nodes = e ^ ((uciElo + 200) / 297.12)

was

Code: Select all

nodes = e ^ ((uciElo + 976.7) / 297.12)

Repeat the tournament of UCI_Elo 1500 but now using the new formula

Code: Select all

nodes = e ^ ((uciElo + 200) / 297.12)

my node limit now is
nodes = e^(5.72) = 305

The result of this tournament is much better compared to the previous one. It still needs adjustment though.

Code: Select all

   # PLAYER                        :  RATING  POINTS  PLAYED   (%)
   1 Cheese 2.0 ucielo 1500        :  2354.0   193.0     200    97
   2 Cheng 4.39 ucielo 1500        :  2197.4   183.0     200    92
   3 Amyan 1.72 ucielo 1500        :  1738.5   145.5     200    73
   4 Ufim v8.02 ucielo 1500        :  1586.7   125.0     200    63
   5 Rhetoric 1.4.3 ucielo 1500    :  1422.4    97.0     200    49
   6 MadChess 2.2 ucielo 1500      :  1384.4    90.0     200    45
   7 D2019.2.37.53 ucielo 1500     :  1384.4    90.0     200    45
   8 Discocheck 5.2 ucielo 1500    :  1325.2    79.0     200    40
   9 Arasan 21.3 ucielo 1500       :  1142.5    47.0     200    24
  10 CT800 V1.34 ucielo 1500       :  1066.7    35.5     200    18
  11 Hiarcs 14 ucielo 1500         :   897.7    15.0     200     8

Done another RR with formula adjusted to:

Code: Select all

nodes = e ^ ((uciElo + 500) / 297.12)

@ Elo 1500, nodes is around 840.

D2019.2.37.53 ucielo 1500 is closer now to 1500.

Code: Select all

   # PLAYER                              :  RATING  ERROR  POINTS  PLAYED   (%)
   1 Cheese 2.0 ucielo 1500              :  2171.7  184.4    83.5      88    95
   2 Cheng 4.39 ucielo 1500              :  2035.9  154.8    78.5      88    89
   3 Fruit reloaded v3.21 ucielo 1500    :  1760.6  105.7    64.0      88    73
   4 Amyan 1.72 ucielo 1500              :  1694.2   98.3    59.5      88    68
   5 MadChess 2.2 ucielo 1500            :  1530.0   95.3    47.0      88    53
   6 D2019.2.37.53 ucielo 1500           :  1523.7   91.3    46.5      88    53
   7 Ufim v8.02 ucielo 1500              :  1511.2   93.3    45.5      88    52
   8 Rhetoric 1.4.3 ucielo 1500          :  1397.8   93.0    36.5      88    41
   9 Discocheck 5.2 ucielo 1500          :  1338.7   97.3    32.0      88    36
  10 CT800 V1.34 ucielo 1500             :  1113.5  114.5    17.0      88    19
  11 Arasan 21.3 ucielo 1500             :   972.1  134.8     9.5      88    11
  12 Hiarcs 14 ucielo 1500               :   950.5  143.5     8.5      88    10

Ferdy · Post by **Ferdy** » Sun Jul 07, 2019 1:10 am

Ferdy wrote: ↑Sat Jul 06, 2019 11:40 pm

Ferdy wrote: ↑Sat Jul 06, 2019 8:38 pm

Code: Select all

nodes = e ^ ((uciElo + 200) / 297.12)

was

Code: Select all

nodes = e ^ ((uciElo + 976.7) / 297.12)

Repeat the tournament of UCI_Elo 1500 but now using the new formula

Code: Select all

nodes = e ^ ((uciElo + 200) / 297.12)

my node limit now is
nodes = e^(5.72) = 305

The result of this tournament is much better compared to the previous one. It still needs adjustment though.

Code: Select all

   # PLAYER                        :  RATING  POINTS  PLAYED   (%)
   1 Cheese 2.0 ucielo 1500        :  2354.0   193.0     200    97
   2 Cheng 4.39 ucielo 1500        :  2197.4   183.0     200    92
   3 Amyan 1.72 ucielo 1500        :  1738.5   145.5     200    73
   4 Ufim v8.02 ucielo 1500        :  1586.7   125.0     200    63
   5 Rhetoric 1.4.3 ucielo 1500    :  1422.4    97.0     200    49
   6 MadChess 2.2 ucielo 1500      :  1384.4    90.0     200    45
   7 D2019.2.37.53 ucielo 1500     :  1384.4    90.0     200    45
   8 Discocheck 5.2 ucielo 1500    :  1325.2    79.0     200    40
   9 Arasan 21.3 ucielo 1500       :  1142.5    47.0     200    24
  10 CT800 V1.34 ucielo 1500       :  1066.7    35.5     200    18
  11 Hiarcs 14 ucielo 1500         :   897.7    15.0     200     8

Done another RR with formula adjusted to:

Code: Select all

nodes = e ^ ((uciElo + 500) / 297.12)

@ Elo 1500, nodes is around 840.

D2019.2.37.53 ucielo 1500 is closer now to 1500.

Code: Select all

   # PLAYER                              :  RATING  ERROR  POINTS  PLAYED   (%)
   1 Cheese 2.0 ucielo 1500              :  2171.7  184.4    83.5      88    95
   2 Cheng 4.39 ucielo 1500              :  2035.9  154.8    78.5      88    89
   3 Fruit reloaded v3.21 ucielo 1500    :  1760.6  105.7    64.0      88    73
   4 Amyan 1.72 ucielo 1500              :  1694.2   98.3    59.5      88    68
   5 MadChess 2.2 ucielo 1500            :  1530.0   95.3    47.0      88    53
   6 D2019.2.37.53 ucielo 1500           :  1523.7   91.3    46.5      88    53
   7 Ufim v8.02 ucielo 1500              :  1511.2   93.3    45.5      88    52
   8 Rhetoric 1.4.3 ucielo 1500          :  1397.8   93.0    36.5      88    41
   9 Discocheck 5.2 ucielo 1500          :  1338.7   97.3    32.0      88    36
  10 CT800 V1.34 ucielo 1500             :  1113.5  114.5    17.0      88    19
  11 Arasan 21.3 ucielo 1500             :   972.1  134.8     9.5      88    11
  12 Hiarcs 14 ucielo 1500               :   950.5  143.5     8.5      88    10

Add Piranha 0.5, a 1500 ccrl 40/4 engine, to see these UCI_Elo 1500 engines in the ccrl 40/4 rating scale.

Code: Select all

  # PLAYER                              :  RATING  ERROR  POINTS  PLAYED   (%)
   1 Cheese 2.0 ucielo 1500              :  2001.4  214.5    91.5      96    95
   2 Cheng 4.39 ucielo 1500              :  1847.3  174.0    85.5      96    89
   3 Fruit reloaded v3.21 ucielo 1500    :  1570.2  129.1    69.0      96    72
   4 Piranha 0.5                         :  1500.0   ----    63.5      96    66
   5 Amyan 1.72 ucielo 1500              :  1487.8  121.8    62.5      96    65
   6 D2019.2.37.53 ucielo 1500           :  1338.6  119.7    49.5      96    52
   7 Ufim v8.02 ucielo 1500              :  1327.4  124.0    48.5      96    51
   8 MadChess 2.2 ucielo 1500            :  1310.6  118.5    47.0      96    49
   9 Rhetoric 1.4.3 ucielo 1500          :  1213.3  126.6    38.5      96    40
  10 Discocheck 5.2 ucielo 1500          :  1140.5  133.1    32.5      96    34
  11 CT800 V1.34 ucielo 1500             :   919.6  159.1    17.0      96    18
  12 Hiarcs 14 ucielo 1500               :   782.7  179.2     9.5      96    10
  13 Arasan 21.3 ucielo 1500             :   782.7  173.0     9.5      96    10

There was an estimate by Kai for FIDE and CCRL ratings.

Code: Select all

FIDE rating = 0.70*CCRL rating + 840 ELO points

For D2019.2.37.53 ucielo 1500, its FIDE rating could be around:
FIDE rating = 0.7*1339 + 840 = 1777

Try CT800 ucielo 1500
CCRL 40/4: 920
FIDE rating = 0.7*920 + 840 = 1484
So this is the nearest to FIDE rating of 1500.

pedrox · Post by **pedrox** » Sun Jul 07, 2019 3:12 am

Ferdy wrote: ↑Sat Jul 06, 2019 11:32 pm I am only interested on UCI_Elo and UCI_LimitStrength options of the engine.

Hi Ferdy, I would be happy if you can do a DanaSah test. Version 7.9 with LimitStrength.

https://www.mediafire.com/file/9u7sn4n7 ... a.zip/file

I have configured the default with 1500 points of Elo and the type of opponent engine (the engine allows another 2 types of opponent: chesscomputer of the 80s-90s and human).

When the engine starts, it usually indicates some information:

Code: Select all

2019-07-07 02:15:23,510<--1:info Elo engine: 1500
2019-07-07 02:15:23,511<--1:info Elo real: 1450
2019-07-07 02:15:23,512<--1:info nps_elo: 1737
2019-07-07 02:15:23,514<--1:info RandF: 275

I use 50 points less than elo because version 7.9 is about 50 points better than version 6.5 with which I tried to do the regulation.

My formula for Elo between 1100-2200 is:
nps_elo = (double)(pow(2,((double)elo)/100.0) / 100.0) * (double)((2200.0 - elo) / 100.0);
My number of nodes is approximately double that of yours, I try to compensate with a random factor in the evaluation, for that Elo I could add or subtract to eval something between 0 and 137 cp (RandF/2) (I do this for extend the regulation almost as a random move engine and have nodes).

If instead of the opponent engine you select human, I have tried to apply an own formula similar to:
FIDE rating = 0.70*CCRL rating + 840 ELO points

Elo 700-1900 --> diff_chesscomputer_engine = (465 * (double)(1.0 - (double)((1900.0-elo)/1200.0)));
The number 465 I have obtained by comparing the list CCRL with the list WIKI-Elo aktiv for dedicated machines (chesscomputer), a engine with an Elo x could have up to + 465 points in the wiki, this difference is not linear and varies if we talk about 700 or 2000 points Elo. For example my experience is that Darky engine with an approximate Elo of 1500 points plays like the Mephisto Roma 32 bit.

Elo human = Elo engine + diff_chesscomputer_engine + diff_human_chesscomputer = 1500 + 310 + 70 = 1880

If you use Arena and change the configuration, the engine must be initialized, I change the options when I receive an isready command and as you know Arena does not send it.

I have not yet calibrated the regulation of the engine, I have also found some errors in danasah 6.5, but I hope it is not too far from your 1500 points, I hope there are not more than +-100 points.

If it is not a nuisance it would be nice to see if danasah plays in 1500 engine like Piranha and in 1500 human like CT800.

Ferdy · Post by **Ferdy** » Sun Jul 07, 2019 3:50 am

pedrox wrote: ↑Sun Jul 07, 2019 3:12 am If it is not a nuisance it would be nice to see if danasah plays in 1500 engine like Piranha and in 1500 human like CT800.

Currently running Danasah on gauntlet.

Code: Select all

Rank Name                          Elo     +/-   Games   Score   Draws
   0 DanaSah 7.9 ucielo 1500       -76     202      14   39.3%    7.1%
   1 D2019.2.37.53 ucielo 1500     inf     nan       2  100.0%    0.0%
   2 Cheng 4.39 ucielo 1500        inf     nan       2  100.0%    0.0%
   3 Cheese 2.0 ucielo 1500        inf     nan       2  100.0%    0.0%
   4 Amyan 1.72 ucielo 1500        inf     nan       1  100.0%    0.0%
   5 Discocheck 5.2 ucielo 1500       0     nan       2   50.0%    0.0%
   6 Arasan 21.3 ucielo 1500      -191     nan       2   25.0%   50.0%
   7 Hiarcs 14 ucielo 1500        -inf     nan       1    0.0%    0.0%
   8 CT800 V1.34 ucielo 1500      -inf     nan       2    0.0%    0.0%
   9 Ufim v8.02 ucielo 1500        nan     nan       0    nan%    nan%
  10 Rhetoric 1.4.3 ucielo 1500     nan     nan       0    nan%    nan%
  11 MadChess 2.2 ucielo 1500      nan     nan       0    nan%    nan%
  12 Fruit reloaded v3.21 ucielo 1500     nan     nan       0    nan%    nan%

14 of 96 games finished.

pedrox · Post by **pedrox** » Sun Jul 07, 2019 4:19 am

Thanks !, the first results seems that I will stay a little short.

Ferdy · Post by **Ferdy** » Sun Jul 07, 2019 4:26 am

pedrox wrote: ↑Sun Jul 07, 2019 4:19 am Thanks !, the first results seems that I will stay a little short.

Final results, Danasah is doing good. It is close to CT800.

Code: Select all

   # PLAYER                              :  RATING  ERROR  POINTS  PLAYED   (%)
   1 Cheese 2.0 ucielo 1500              :  1996.5  205.7    99.5     104    96
   2 Cheng 4.39 ucielo 1500              :  1842.7  171.3    93.5     104    90
   3 Fruit reloaded v3.21 ucielo 1500    :  1555.4  129.2    76.0     104    73
   4 Piranha 0.5                         :  1500.0   ----    71.5     104    69
   5 Amyan 1.72 ucielo 1500              :  1488.2  122.1    70.5     104    68
   6 Ufim v8.02 ucielo 1500              :  1325.5  116.8    55.5     104    53
   7 D2019.2.37.53 ucielo 1500           :  1315.1  123.7    54.5     104    52
   8 MadChess 2.2 ucielo 1500            :  1273.8  125.9    50.5     104    49
   9 Rhetoric 1.4.3 ucielo 1500          :  1211.3  123.9    44.5     104    43
  10 Discocheck 5.2 ucielo 1500          :  1135.5  128.2    37.5     104    36
  11 DanaSah 7.9 ucielo 1500             :  1135.5  132.3    37.5     104    36
  12 CT800 V1.34 ucielo 1500             :   867.4  153.8    17.0     104    16
  13 Arasan 21.3 ucielo 1500             :   754.4  175.3    10.5     104    10
  14 Hiarcs 14 ucielo 1500               :   734.7  176.8     9.5     104     9

Remove Piranha and set CT800 to 1500, to see how far the other engines from CT800.

Code: Select all

   # PLAYER                              :  RATING  ERROR  POINTS  PLAYED   (%)
   1 Cheese 2.0 ucielo 1500              :  2601.9  233.5    91.5      96    95
   2 Cheng 4.39 ucielo 1500              :  2466.7  214.4    86.5      96    90
   3 Fruit reloaded v3.21 ucielo 1500    :  2180.4  168.9    71.0      96    74
   4 Amyan 1.72 ucielo 1500              :  2131.2  162.4    67.5      96    70
   5 Ufim v8.02 ucielo 1500              :  1948.2  146.1    52.5      96    55
   6 D2019.2.37.53 ucielo 1500           :  1936.9  147.5    51.5      96    54
   7 MadChess 2.2 ucielo 1500            :  1925.7  153.2    50.5      96    53
   8 Rhetoric 1.4.3 ucielo 1500          :  1835.8  141.5    42.5      96    44
   9 DanaSah 7.9 ucielo 1500             :  1778.3  147.3    37.5      96    39
  10 Discocheck 5.2 ucielo 1500          :  1772.4  139.5    37.0      96    39
  11 CT800 V1.34 ucielo 1500             :  1500.0   ----    17.0      96    18
  12 Arasan 21.3 ucielo 1500             :  1383.7  157.7    10.5      96    11
  13 Hiarcs 14 ucielo 1500               :  1342.4  167.1     8.5      96     9

pedrox · Post by **pedrox** » Sun Jul 07, 2019 4:34 am

but this was played with opponent: engine?
Then, the engine plays weaker than expected

Ferdy · Post by **Ferdy** » Sun Jul 07, 2019 4:37 am

pedrox wrote: ↑Sun Jul 07, 2019 4:34 am but this was played with opponent: engine?
Then, the engine plays weaker than expected

Yes opponent is engine.

pedrox · Post by **pedrox** » Sun Jul 07, 2019 4:53 am

Ferdy wrote: ↑Sun Jul 07, 2019 4:37 am
pedrox wrote: ↑Sun Jul 07, 2019 4:34 am but this was played with opponent: engine?
Then, the engine plays weaker than expected
Yes opponent is engine.

One last question. Time control is equivalent to CCRL 40/4?

Ferdy · Post by **Ferdy** » Sun Jul 07, 2019 4:57 am

pedrox wrote: ↑Sun Jul 07, 2019 4:53 am
Ferdy wrote: ↑Sun Jul 07, 2019 4:37 am
pedrox wrote: ↑Sun Jul 07, 2019 4:34 am but this was played with opponent: engine?
Then, the engine plays weaker than expected
Yes opponent is engine.
One last question. Time control is equivalent to CCRL 40/4?

I use TC 1m+1s.
According to my PC's speed, I can match CCRL 40/4 with TC 40 moves in 2 minutes.

UCI_Elo

Re: UCI_Elo

Re: UCI_Elo

Re: UCI_Elo

Re: UCI_Elo

Re: UCI_Elo

Re: UCI_Elo

Re: UCI_Elo

Re: UCI_Elo

Re: UCI_Elo

Re: UCI_Elo