The Unofficial World Championships Of Computer Programs In International Draughts Blitz 2020

Discussion about development of draughts in the time of computer and Internet.
Krzysztof Grzelak
Posts: 927
Joined: Thu Jun 20, 2013 17:16
Real name: Krzysztof Grzelak

Re: The Unofficial World Championships Of Computer Programs In International Draughts Blitz 2020

Post by Krzysztof Grzelak » Mon Sep 14, 2020 17:11

stepanov.anatolievich wrote:
Mon Sep 14, 2020 03:50
Incredible tournament, incredible work!

On what game and when will be the next Championship?
I think next year.Will try to get - ask for the availability of the programs Damy and TDKing. I will also ask the author of Maximus to be able to start next year.

BertTuyt
Posts: 1435
Joined: Wed Sep 01, 2004 19:42

Re: The Unofficial World Championships Of Computer Programs In International Draughts Blitz 2020

Post by BertTuyt » Mon Sep 14, 2020 19:58

Krzystof, without any doubt i will share the next version of the Damage engine.
However my focus is to understand some results of the current version of Damage.

The 2 draws within a won position, due to move repetition, is clear. As I mostly test against KR, the endgame is never played and the program stops at a DB position. So here i know what to do. I did not implement all code.

What puzzles me is the losses with black (all 6 losses were with black).
It could be coincidence.
Or a very deep mysterious bug, which i did not find in the DXP matches against KR, so far.

The challenge I have now that I'm not able to reproduce these errors.
So you could help me.

To my knowledge you have used 2 computers during the tournament.

Could you do a test with the 7p db, but only 1 thread, and 25 ply fixed depth, you could start such a test from the GUI.
I'm interested to see the outcome of the 2 computers.
So could you make a screenshot how the engine output looks like.

The position below is from the game against Sjende Blyn.

Thanks for your support.
And although i already shared it, many thanks for organizing this tournament, we all enjoyed it!!!

Bert


black to play...

Image

User avatar
stepanov.anatolievich
Posts: 3
Joined: Fri Jan 17, 2020 10:05
Real name: Stepanov Oleg Anatolievich

Re: The Unofficial World Championships Of Computer Programs In International Draughts Blitz 2020

Post by stepanov.anatolievich » Mon Sep 14, 2020 20:01

>> I think next year.Will try to get - ask for the availability of the programs Damy and TDKing. I will also ask the author of Maximus to be able to start next year.

I mean you made not only International Draughts. I think another variants very much interesting too. International Give Away more aggressive. I hear about Edeon version and now Svirin made it for new Aurora. But of course Scan is the best and all Russian used it on correspondense games. May be possible ask another authors of strong programs made Give Away versions? Or you know some else?

Jelle Wiersma
Posts: 54
Joined: Wed Aug 12, 2015 17:45
Real name: Jelle Wiersma

Re: The Unofficial World Championships Of Computer Programs In International Draughts Blitz 2020

Post by Jelle Wiersma » Mon Sep 14, 2020 22:22

BertTuyt wrote:
Mon Sep 14, 2020 19:58
Krzystof, without any doubt i will share the next version of the Damage engine.
However my focus is to understand some results of the current version of Damage.

The 2 draws within a won position, due to move repetition, is clear. As I mostly test against KR, the endgame is never played and the program stops at a DB position. So here i know what to do. I did not implement all code.

What puzzles me is the losses with black (all 6 losses were with black).
It could be coincidence.
Or a very deep mysterious bug, which i did not find in the DXP matches against KR, so far.

The challenge I have now that I'm not able to reproduce these errors.
So you could help me.

To my knowledge you have used 2 computers during the tournament.

Could you do a test with the 7p db, but only 1 thread, and 25 ply fixed depth, you could start such a test from the GUI.
I'm interested to see the outcome of the 2 computers.
So could you make a screenshot how the engine output looks like.

The position below is from the game against Sjende Blyn.

Thanks for your support.
And although i already shared it, many thanks for organizing this tournament, we all enjoyed it!!!

Bert


black to play...

Image
Hi Bert, that bug is likely to be related to the transposition table, multi-threading, stopping the search or a combination of these.
Strange though it didn't show in your matches against Kingsrow; considering Krzysztof's tournament, it should be the case that several losses/draws of Damage against Kingsrow where caused by a totally wrong move.

By-the-way, forgot to ask in my previous post: Is Damage capable of supplying the 158 positions of the 158 DXP game match?

Jelle

Jelle Wiersma
Posts: 54
Joined: Wed Aug 12, 2015 17:45
Real name: Jelle Wiersma

Re: The Unofficial World Championships Of Computer Programs In International Draughts Blitz 2020

Post by Jelle Wiersma » Mon Sep 14, 2020 22:32

Krzysztof Grzelak wrote:
Mon Sep 14, 2020 17:11
stepanov.anatolievich wrote:
Mon Sep 14, 2020 03:50
Incredible tournament, incredible work!

On what game and when will be the next Championship?
I think next year.Will try to get - ask for the availability of the programs Damy and TDKing. I will also ask the author of Maximus to be able to start next year.
Hi Krzysztof, may be you can also ask Nicolas Guibert for the program Buggy, and Gijsbert Wiesendekker / Klaas Bor for the program GWD. Both strong programs. And thanks for organizing and running the tournament 13 days in a row!

Jelle

Krzysztof Grzelak
Posts: 927
Joined: Thu Jun 20, 2013 17:16
Real name: Krzysztof Grzelak

Re: The Unofficial World Championships Of Computer Programs In International Draughts Blitz 2020

Post by Krzysztof Grzelak » Mon Sep 14, 2020 23:20

Bert please.
Computer 1

Image

Computer 2

Image

BertTuyt
Posts: 1435
Joined: Wed Sep 01, 2004 19:42

Re: The Unofficial World Championships Of Computer Programs In International Draughts Blitz 2020

Post by BertTuyt » Mon Sep 14, 2020 23:24

Jelle, basically that is what is also expect, as the most tricky part is the multi-thread search (in my case YBWC).
But my puzzle is why does it occur so often in the tournament (and only black, but 6 might not be significant) and not so frequent in DXP matches.
Therefore my request to you for a DXP match against SB.

As (for an unknown reason I'm not able anymore to attach pictures or files in posts), below the dropbox link.
Result (from Damage perspective) 3 losses (all black), and 1 win (also black) and 154 draw.

In game 62 it could be that move 37 4-9 is a similar blunder as seen in the tournament

https://www.dropbox.com/sh/6gsinim40bew ... ewsWa?dl=0

Thanks for your support,

Bert

Krzysztof Grzelak
Posts: 927
Joined: Thu Jun 20, 2013 17:16
Real name: Krzysztof Grzelak

Re: The Unofficial World Championships Of Computer Programs In International Draughts Blitz 2020

Post by Krzysztof Grzelak » Mon Sep 14, 2020 23:25

Jelle Wiersma wrote:
Mon Sep 14, 2020 22:32
Hi Krzysztof, may be you can also ask Nicolas Guibert for the program Buggy, and Gijsbert Wiesendekker / Klaas Bor for the program GWD. Both strong programs. And thanks for organizing and running the tournament 13 days in a row!

Jelle
I have been asking for this programs for several years.I wanted to buy a Buggy to no avail. GWD also asked me to no avail.

BertTuyt
Posts: 1435
Joined: Wed Sep 01, 2004 19:42

Re: The Unofficial World Championships Of Computer Programs In International Draughts Blitz 2020

Post by BertTuyt » Mon Sep 14, 2020 23:47

Krzysztof,

in my case the move 19-23 is played after 28368395 nodes 25 ply, and 37194237 nodes 26 ply.
Score is the same 0.00

But i do not have an AMD machine, but it should not make a difference?

Here the dropbox link, could you check if you use the engine version with the same date?

https://www.dropbox.com/sh/pxqa6zc8bu5o ... WOqua?dl=0

Bert

Krzysztof Grzelak
Posts: 927
Joined: Thu Jun 20, 2013 17:16
Real name: Krzysztof Grzelak

Re: The Unofficial World Championships Of Computer Programs In International Draughts Blitz 2020

Post by Krzysztof Grzelak » Tue Sep 15, 2020 00:12

Yes, the same draughts engine.

BertTuyt
Posts: 1435
Joined: Wed Sep 01, 2004 19:42

Re: The Unofficial World Championships Of Computer Programs In International Draughts Blitz 2020

Post by BertTuyt » Tue Sep 15, 2020 01:22

Krzystof, sorry i did the test with the position after the move 23-29, now i can confirm i have the same results as you.
So a strange bug which for whatever reason occurred several times during the tournament.....
And which was not really obvious during the DXP matches.

Bert

Krzysztof Grzelak
Posts: 927
Joined: Thu Jun 20, 2013 17:16
Real name: Krzysztof Grzelak

Re: The Unofficial World Championships Of Computer Programs In International Draughts Blitz 2020

Post by Krzysztof Grzelak » Tue Sep 15, 2020 08:30

Bert.

Something will show you for some time. I will play a match using DXP. Kingsrow - Damage and will arrange the match as follows:

Kingsrow 1.62x64

Opening Book = Best Moves
HashTable Size = 512 MB
DB cache Size = 30000 MB
The base ends = 8 Pieces - WLD - DTW
Threads = 16
Time = 1 Min / 75 Moves

Damage 15.3x64

Book = 0
Threads = 16
Tt-shared = 1
Tt-size = 25
Bb-cache = 5
Bb-size = 7
Bb-preload = 7
Time = 1 Min / 75 Moves

The match will be played on the processor AMD Ryzen Threadripper 1950X. Write me or Ed if this is the correct match setup. You'll see something interesting.

Krzysztof Grzelak
Posts: 927
Joined: Thu Jun 20, 2013 17:16
Real name: Krzysztof Grzelak

Re: The Unofficial World Championships Of Computer Programs In International Draughts Blitz 2020

Post by Krzysztof Grzelak » Tue Sep 15, 2020 09:07

BertTuyt wrote:
Tue Sep 15, 2020 01:22
Krzystof, sorry i did the test with the position after the move 23-29, now i can confirm i have the same results as you.
So a strange bug which for whatever reason occurred several times during the tournament.....
And which was not really obvious during the DXP matches.

Bert
I have my theory of this error.

BertTuyt
Posts: 1435
Joined: Wed Sep 01, 2004 19:42

Re: The Unofficial World Championships Of Computer Programs In International Draughts Blitz 2020

Post by BertTuyt » Tue Sep 15, 2020 09:13

Krzystof, if you have a theory then please share it.
As you have played all the games, you might have seen something strange.

As the results on your computers and mine are exactly the same, I suspect that Jelle is right, and it is related to the hash-table and multi-threading, and maybe the way i abort the search...

Anyway thanks for your help, Im sure we will find the cause of the problem.

Bert

Krzysztof Grzelak
Posts: 927
Joined: Thu Jun 20, 2013 17:16
Real name: Krzysztof Grzelak

Re: The Unofficial World Championships Of Computer Programs In International Draughts Blitz 2020

Post by Krzysztof Grzelak » Tue Sep 15, 2020 09:31

Bert.

Please write if the match setup is correct - I know that he played such a match on his computer. I'll make an exception and play one match and show you something. It may be important to you.

My theory is this.

I have the impression that the program responds to movement every 3 seconds. The duration of the entire game is about 2 and a half minutes. I wondered what about the next time. After the 1-6 move from the game against Sjende Blyn, the program is unable to count the variant deeper due to time (too little time). The 3 seconds come and the program responds to the movement. Just after 3 seconds, the program makes a move - no matter what it is (whether it is a winner or a loser).

Post Reply