Chess324 Top 15 Round-Robin (free engines only)

[Modern Times]

Following on from my previous work on this under CCRL, I have created a new version of this for free engines only.

Tournament Scores:

Code: Select all

1   Stockfish 16.1      2056.0 / 2800
2   Berserk 13          1833.0 / 2800
3   Obsidian 13         1674.5 / 2800
4   Caissa 1.19         1619.5 / 2800
5   PlentyChess 2.1.0   1562.5 / 2800
6   Alexandria 7.0.0    1446.0 / 2800
7   Clover 6.2          1445.5 / 2800
8   RubiChess 20240112  1332.5 / 2800
9   Viridithas 13.0.0   1277.5 / 2800
10  Uralochka 3.41      1191.0 / 2800
11  Titan 1.1           1161.5 / 2800
12  Seer 2.8.0          1157.0 / 2800
13  Stormphrax 5.0.0    1124.0 / 2800
14  Rebel 16.3          1062.0 / 2800
15  Lizard 10.4         1057.5 / 2800

No surprises really.

Stats:
White wins: 8'632 (41.1%)
Black wins: 790 (3.8%)
Draws: 11'578 (55.1%)
White score: 68.7%

Draw rate low due to the unbalanced openings.

Removal of the commercials Torch, Ethereal 14.25 and Komodo Dragon 3.3 allowed additional free engines to enter, being Stormphrax, Rebel and Lizard.

Tournament Conditions:
Time Control: game in 2 minutes plus 1 second increment, 2'+1"
Hardware: Ryzen 9 5900X, Intel i7-12700, Intel 10900
CuteChess GUI, concurrency 12 on the Ryzen, and 8 on the Intels
Format: Round-robin using Stefah Pohl's unbalanced 3-move book Chess324_xxl_big_+090_+119.pgn played reversed sides per pairing.
Adjudication: Syzygy 5-men only.
Endgame tablebases: Syzygy 5-men.
Each engine 512B hash, 1 thread, ponder off

[Modern Times]

Reluctantly - because I've lost all faith in Elo calculations - you get different answers from Ordo and bayeselo, and different answers from each of them depending on what parameters you use. You don't know what is the truth.

Ordo output of the tournament:

Code: Select all

# PLAYER                   :  RATING  ERROR  POINTS  PLAYED   (%)       W     D    L  D(%)  OppAvg
   1 Stockfish 16.1        :    3833      9  2056.0    2800  73.4    1340  1432   28  51.1    3650
   2 Berserk 13            :    3770      9  1833.0    2800  65.5    1096  1474  230  52.6    3655
   3 Obsidian 13           :    3729      8  1674.5    2800  59.8     881  1587  332  56.7    3658
   4 Caissa 1.19           :    3716      9  1619.5    2800  57.8     841  1557  402  55.6    3659
   5 PlentyChess 2.1.0     :    3701      8  1562.5    2800  55.8     755  1615  430  57.7    3660
   6 Alexandria 7.0.0      :    3673      7  1446.0    2800  51.6     640  1612  548  57.6    3662
   7 Clover 6.2            :    3673      9  1445.5    2800  51.6     631  1629  540  58.2    3662
   8 RubiChess 20240112    :    3645      8  1332.5    2800  47.6     552  1561  687  55.8    3664
   9 Viridithas 13.0.0     :    3632      8  1277.5    2800  45.6     486  1583  731  56.5    3665
  10 Uralochka 3.41        :    3611      8  1191.0    2800  42.5     422  1538  840  54.9    3666
  11 Titan 1.1             :    3603      8  1161.5    2800  41.5     373  1577  850  56.3    3667
  12 Seer 2.8.0            :    3602      8  1157.0    2800  41.3     383  1548  869  55.3    3667
  13 Stormphrax 5.0.0      :    3594      8  1124.0    2800  40.1     413  1422  965  50.8    3667
  14 Rebel 16.3            :    3578      8  1062.0    2800  37.9     322  1480  998  52.9    3669
  15 Lizard 10.4           :    3577      8  1057.5    2800  37.8     287  1541  972  55.0    3669

ordo -p Chess324.pgn -a 3833 -A "Stockfish 16.1" -U 0,1,2,3,4,5,7,8,9,10,11 -N 0,1 -n 8 -V -D -s 100 -J -g Chess324groups.txt -j Chess324headtohead.txt -o Chess324_OrdoRatings.txt -c Chess324_OrdoRatings.csv

I think these give the same ratings as the parameters that SPCC use for its UHO list.

Pgn available on request, p.m. me your email adress if interested.

[Graham Banks]

I'm pleased that you're continuing this rating list, as a lot of people followed it.

May I suggest that the next few worth testing are the latest versions of Peacekeeper, Clarity, Starzix and Motor.

[Modern Times]

It is a Top 15 round-robin tournament with engine rankings, not a ratings list.

I'm not aware that any of the ones you mention are close to qualifying for the Top 15. My approach is to keep an eye on the usual ratings lists and when I see something that looks like it may qualify I run some test matches and proceed further or not.

Typo in original post, should be
Each engine 512MB hash, 1 thread, ponder off

[Modern Times]

Clarity 7.2.0 looks like it may be close to breaking in, I'll run some test matches to see.

[Graham Banks]

Clarity 7.2.0 looks like it may be close to breaking in, I'll run some test matches to see.

[Modern Times]

Yes - Clarity 7.2.0 beats Lizard and Rebel head-to-head, so full run now underway. Very likely to enter the Top15 and push out one of those.