Internet engine matches

[Krzysztof Grzelak]

Match KINGSROW - ARES (2-move ballots)

Kingsrow 1.63 vs Ares 1.53d NNUE 0 wins, 1 losses, 157 draws, 0 unknowns

Kingsrow 1.63 x64

Threads = 16
Opening Book = Best Moves
Pondering = On
HashTable Size = 512 MB
DB cache Size = 30000 MB
The base ends = 8 Pieces
Time = 1 Min / 120 Moves

Ares 1.53d NNUE x64

Threads = 16
Book = On
TT-size = 32
Bucket-size = 4
Ponder = On
The base ends = 6 Pieces
Time = 1 Min / 120 Moves

Match played on a computer with the equipment.

Processor - AMD Ryzen Threadripper 1950X
Hard disc - Samsung 970 EVO 2 TB
Memory of frames - 128 GB DDR4 2400
System - Windows 10 64 bit Pro

[hendrikv]

[Event "Game 18, opening 32-28 16-21"]
[Date "2022.10.18"]
[White "Ares NNUE v1.53d"]
[Black "Kingsrow 1.63"]
[Result "2-0"]
[FEN "W:W28,31,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50:B1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,17,18,19,20,21"]
1. 31-26 11-16 2. 38-32 18-23 3. 34-29 23x34 4. 40x29 17-22 5. 28x17 19-23 6. 29x18 13x11 7. 26x17 12x21 8. 37-31 21-26 9. 32-28 26x37 10. 41x32 14-19 11. 35-30 19-23 12. 28x19 20-24 13. 30-25 24x13 14. 43-38 7-12 15. 45-40 1-7 16. 46-41 12-18 17. 41-37 7-12 18. 33-28 10-14 19. 40-34 13-19 20. 39-33 5-10 21. 44-40 8-13 22. 36-31 2-8 23. 31-27 15-20 24. 37-31 20-24 25. 49-43 24-30 26. 34-29 30-34 27. 27-22 18x36 28. 29-23 34x45 29. 25-20 14x25 30. 23x5 25-30 31. 43-39 30-35 32. 32-27 13-18 33. 28-22 9-13 34. 39-34 4-9 35. 42-37 11-17 36. 22x11 16x7 37. 5-10 7-11 38. 10-15 11-17 39. 38-32 9-14 40. 33-28 13-19 41. 34-30 35x24 42. 15x38 18-23 43. 38-33 6-11 44. 33-42 36-41 45. 37x46 11-16 46. 27-21 16x38 47. 42x7 14-20 48. 7-16 8-13 49. 16-27 3-9 50. 27-31 17-21 51. 28-22 19-23 52. 31-42 20-25 53. 42-24 21-26 54. 24x2 23-29 55. 22-17 29-34 56. 17-11 34-40 57. 11-6 40-44 58. 50x39 25-30 59. 2x35 45-50 60. 35-44 26-31 61. 6-1 9-14 62. 1-6 50-45 63. 44-40 45x43 64. 48x39 14-20 65. 46-41 20-25 66. 39-34 31-36 67. 6-1 25-30 68. 34x25 *

[Krzysztof Grzelak]

Match KINGSROW - ARES (3-move ballots)

Kingsrow 1.63 vs. Ares AI 1.91 23 wins ,1 losses, 962 draws, 0 unknowns

Kingsrow 1.63 x64

Threads = 16
Opening Book = Best Moves
Pondering = On
HashTable Size = 512 MB
DB cache Size = 30000 MB
The base ends = 8 Pieces
Time = 1 Min / 120 Moves

Ares AI 1.91 x64

Threads = 16
TT-size = 16
Book = On
Bucket-size = 2
Ponder = On
The base ends = 6 Pieces
Time = 1 Min / 120 Moves

Match played on a computer with the equipment:

processor - AMD Ryzen Threadripper 2990WX 3.0 GHz
hard drive - Samsung 970 EVO Plus NVMe M.2 SSD 2 TB
memory ram - Kingston HyperX Fury 128 GB DDR4 2400 MHz CL15
monitor - Nec MultiSync EA245 WMi
keyboard - A4Tech KV-300H Slim
mouse - Razer DeathAdder Essential
operating system - Microsoft Windows 11 Professional 64 bit

[Joost Buijs]

What on earth is 'TT-size = 16'? That means it used a hash-table of 65 KB, does not seem very appropriate with 16 threads sharing the hash-table.

That version of the engine does not know 'Bucket-size', it should be 'Bucket-shift = 2', it clearly was in the ini-file when I sent you this engine almost a year ago.

[Krzysztof Grzelak]

What on earth is 'TT-size = 16'? That means it used a hash-table of 65 KB, does not seem very appropriate with 16 threads sharing the hash-table.

That version of the engine does not know 'Bucket-size', it should be 'Bucket-shift = 2', it clearly was in the ini-file when I sent you this engine almost a year ago.

Sorry, I have to repeat the entire match from the beginning. As for the Bucket-size option, I think you gave me that and not Bucket-shift as you wrote - I have to check it because I have everything on the hard drive.

[Joost Buijs]

Sorry, I have to repeat the entire match from the beginning. As for the Bucket-size option, I think you gave me that and not Bucket-shift as you wrote - I have to check it because I have everything on the hard drive.

The ini file is still on my FTP-server, it shows:

threads = 16
tt-size = 32
bucket-shift = 2
dxp-port = 27531
book = on
ponder = on
egdb = on
egdb-path = "./ADB-6P"

I think the default is 'bucket-shift = 2', if it doesn't get this command it's not a problem. But if you really used a hashtable of just 65 KB this will cost a lot of search-depth, there is a chance of more collisions too.

The size of the hash-table is 2^TT-size, 2^16 = 65.536, if you want to get 512 MB use 'TT-size = 29'.

You don't have to repeat the match, it's not so important, maybe better to wait until I have the new version modified in such a way that it will work on your AMD Zen1+ PC's.

[MichelG]

What on earth is 'TT-size = 16'? That means it used a hash-table of 65 KB, does not seem very appropriate with 16 threads sharing the hash-table.

That version of the engine does not know 'Bucket-size', it should be 'Bucket-shift = 2', it clearly was in the ini-file when I sent you this engine almost a year ago.

Actually, i think there is a great lesson to be learned here. One should not leave the engine configuration to the end-user.

That is your resposibility. To the typical end user, numbers like hashtable size or 'threads' are meaningless terms. It is up to the programmer to adopt to the hardware it runs on.

Dragon is guilty on this as well, though it has some safeguards.

[Krzysztof Grzelak]

Match KINGSROW - ARES (3-move ballots)

Kingsrow 1.63 vs. Ares AI 1.91 20 wins, 1 losses, 965 draws, 0 unknowns

Kingsrow 1.63 x64

Threads = 16
Opening Book = Best Moves
Pondering = On
HashTable Size = 512 MB
DB cache Size = 30000 MB
The base ends = 8 Pieces
Time = 1 Min / 120 Moves

Ares AI 1.91 x64

Threads = 16
TT-size = 32
Book = On
Bucket-shift = 2
Ponder = On
The base ends = 6 Pieces
Time = 1 Min / 120 Moves

Match played on a computer with the equipment:

processor - AMD Ryzen Threadripper 2990WX 3.0 GHz
hard drive - Samsung 970 EVO Plus NVMe M.2 SSD 2 TB
memory ram - Kingston HyperX Fury 128 GB DDR4 2400 MHz CL15
monitor - Nec MultiSync EA245 WMi
keyboard - A4Tech KV-300H Slim
mouse - Razer DeathAdder Essential
operating system - Microsoft Windows 11 Professional 64 bit

[Krzysztof Grzelak]

Match KINGSROW - MOBY DAM (May 15 2026) (3-move ballots)

Kingsrow 1.63 vs. Moby Dam (May 15 2026) 91 wins, 1 losses, 894 draws, 0 unknowns

Kingsrow 1.63 x64

Threads = 8
Opening Book = Best Moves
Pondering = On
HashTable Size = 512 MB
DB cache Size = 30000 MB
The base ends = 8 Pieces
Time = 1 Min / 120 Moves

Moby Dam (May 15 2026) x64

Threads = 8
TT-size = 28
Opening book = Off
Pondering = On
The base ends = 6 Pieces
Time = 1 Min / 120 Moves

Match played on a computer with the equipment:

processor - AMD Ryzen Threadripper 2990WX 3.0 GHz
hard drive - Samsung 970 EVO Plus NVMe M.2 SSD 2 TB
memory ram - Kingston HyperX Fury 128 GB DDR4 2400 MHz CL15
monitor - Nec MultiSync EA245 WMi
keyboard - A4Tech KV-300H Slim
mouse - Razer DeathAdder Essential
operating system - Microsoft Windows 11 Professional 64 bit

[Krzysztof Grzelak]

Match KINGSROW - DAMAGE 16.1 (3-move ballots)

Kingsrow 1.63 vs. Damage 16.1 44 wins, 3 losses, 939 draws, 0 unknowns

Kingsrow 1.63 x64

Threads = 16
Opening Book = Best Moves
Pondering = On
HashTable Size = 512 MB
DB cache Size = 30000 MB
The base ends = 8 Pieces
Time = 1 Min / 120 Moves

Damage 16.1 x64

Threads = 16
Book = 0n
NNUE-file = nn_20210328.gnn
TT-persistent = On
TT-size = 24
BB-cache = 5
BB-size = 7
BB-preload = 7
The base ends = 7 Pieces
Time = 1 Min / 120 Moves

Match played on a computer with the equipment:

processor - AMD Ryzen Threadripper 2990WX 3.0 GHz
hard drive - Samsung 970 EVO Plus NVMe M.2 SSD 2 TB
memory ram - Kingston HyperX Fury 128 GB DDR4 2400 MHz CL15
monitor - Nec MultiSync EA245 WMi
keyboard - A4Tech KV-300H Slim
mouse - Razer DeathAdder Essential
operating system - Microsoft Windows 11 Professional 64 bit

[Joost Buijs]

Match KINGSROW - ARES (3-move ballots)

Kingsrow 1.63 vs. Ares AI 1.91 23 wins ,1 losses, 962 draws, 0 unknowns
Kingsrow 1.63 vs. Ares AI 1.91 20 wins, 1 losses, 965 draws, 0 unknowns

It surprises me that changing the hash table size barely affects the results.

That Ares version you have is just an untested experimental build I made for Sidiki a very long time ago, which you really wanted a copy of. Turns out it has a repetition detection bug, so it misses a draw every now and then. I've fixed it in the latest version, but I'm holding off on releasing it since there’s still a lot of other stuff I want to polish, and that takes time. The new version also uses a 7P EGDB instead of 6P, this won't make a big difference, but ocassionally it could help.

I also want to advise you to run tests like these on two computers. In my experience, running tests like these (with huge endgame databases and pondering) on a single computer makes the engines influence each other in a bad way, especially under Windows 11 (which is the worst OS Microsoft has ever produced).

[Sidiki]

Hi all,

Based on Scan open source with some improvements (search and new features), a 158 dxp games is still running against Kingsrow.
Perspective for Kingsrow 1.63 : 0 wins, 3 losses, 114 draws.

Both use 6 egtb for the test and 10 threads, pondering off, 1 min/ 99 moves.

Friendly, Sidiki

[Krzysztof Grzelak]

Match KINGSROW - ARES (3-move ballots)

Kingsrow 1.63 vs. Ares AI 1.91 23 wins ,1 losses, 962 draws, 0 unknowns
Kingsrow 1.63 vs. Ares AI 1.91 20 wins, 1 losses, 965 draws, 0 unknowns

It surprises me that changing the hash table size barely affects the results.

That Ares version you have is just an untested experimental build I made for Sidiki a very long time ago, which you really wanted a copy of. Turns out it has a repetition detection bug, so it misses a draw every now and then. I've fixed it in the latest version, but I'm holding off on releasing it since there’s still a lot of other stuff I want to polish, and that takes time. The new version also uses a 7P EGDB instead of 6P, this won't make a big difference, but ocassionally it could help.

I also want to advise you to run tests like these on two computers. In my experience, running tests like these (with huge endgame databases and pondering) on a single computer makes the engines influence each other in a bad way, especially under Windows 11 (which is the worst OS Microsoft has ever produced).

I see that I have nothing to look for on this forum.

[Joost Buijs]

Match KINGSROW - ARES (3-move ballots)

Kingsrow 1.63 vs. Ares AI 1.91 23 wins ,1 losses, 962 draws, 0 unknowns
Kingsrow 1.63 vs. Ares AI 1.91 20 wins, 1 losses, 965 draws, 0 unknowns

It surprises me that changing the hash table size barely affects the results.

That Ares version you have is just an untested experimental build I made for Sidiki a very long time ago, which you really wanted a copy of. Turns out it has a repetition detection bug, so it misses a draw every now and then. I've fixed it in the latest version, but I'm holding off on releasing it since there’s still a lot of other stuff I want to polish, and that takes time. The new version also uses a 7P EGDB instead of 6P, this won't make a big difference, but ocassionally it could help.

I also want to advise you to run tests like these on two computers. In my experience, running tests like these (with huge endgame databases and pondering) on a single computer makes the engines influence each other in a bad way, especially under Windows 11 (which is the worst OS Microsoft has ever produced).

I see that I have nothing to look for on this forum.

This is a very weird response. Is this also why they banned you at the Talkchess forum?

[Krzysztof Grzelak]

Kingsrow 1.63 vs. Ares AI 1.91 23 wins ,1 losses, 962 draws, 0 unknowns
Kingsrow 1.63 vs. Ares AI 1.91 20 wins, 1 losses, 965 draws, 0 unknowns

It surprises me that changing the hash table size barely affects the results.

That Ares version you have is just an untested experimental build I made for Sidiki a very long time ago, which you really wanted a copy of. Turns out it has a repetition detection bug, so it misses a draw every now and then. I've fixed it in the latest version, but I'm holding off on releasing it since there’s still a lot of other stuff I want to polish, and that takes time. The new version also uses a 7P EGDB instead of 6P, this won't make a big difference, but ocassionally it could help.

I also want to advise you to run tests like these on two computers. In my experience, running tests like these (with huge endgame databases and pondering) on a single computer makes the engines influence each other in a bad way, especially under Windows 11 (which is the worst OS Microsoft has ever produced).

I see that I have nothing to look for on this forum.

This is a very weird response. Is this also why they banned you at the Talkchess forum?

No - because a person is honest.