Internet engine matches

[Wieger Wesselink]

Let me add to this that I consider it very kind of Jelle to supply his program to Krzysztof, so that it could take part in this tournament. Furthermore I am very happy that Krzysztof organizes these tournaments, since it gives us an idea of the playing strength of the current draughts engines.

[BertTuyt]

I agree with Wieger that Krzysztof did a great job in organizing the tournament, which for me was a reason to re-join this forum.

Where a tournament gives a clue what the best Tournament Engine is (although I guess we suffer her from sufficient statistics), the Engine Match determines the strongest engine.

Here my personal belief is that Kingsrow and Scan are strongest, and at least the result for Maximus seems to confirm this statement partially.

I played 5 (2 ballot, 158 games) matches this week Kingsrow - Scan, with thinking time (for every game) 1 - 2 -3 - 4 and 5 minutes.
In a previous match I reached a black hole situation with 158 games, and 158 draws, which was now not the case.
Will share results later today in the evening, with the pdn files.
A teaser, based on the results it is still not decided which of the 2 (Kingsrow or Scan) is better.

I also started a match between Damage 14.0 and Kingsrow, unfortunately the program crashed.
This was new for me, so I need to dig deeper into this.
Basically Krzysztof experienced the same as Damage 14.0 does not work on his ThreadRipper.
In my case I also never experienced a crash, but now I use a new computer with 6 cores, so need to find out what is going on (not in the next 2 weeks, due to Business obligations).

Bert

[Sidiki]

Let me add to this that I consider it very kind of Jelle to supply his program to Krzysztof, so that it could take part in this tournament. Furthermore I am very happy that Krzysztof organizes these tournaments, since it gives us an idea of the playing strength of the current draughts engines.

Hi Bert,
I see the idea, yes, share a program with Krzysztof for take part to tournament and self-use itsn't a problem for me, and i already said it that every engine owner is free to share, sell, release his program.

The point that i no accept is to bring restrictions to certains engines on a forum and use the others for exhibition. We saw matches Damage, Kingsrow,Scan, Dragon, Horizon that bring many exchange of ideas on this forum. All this to improve program and in the same time international draughts world.

Several GMI use today all the greatest available programs to perfect their game.
You Bert and several programmers, i mean Fabien, Ed, Jan-Jaap, Michel are very very kind, no doubt about, and this it's only because you love the game and human.

Chess community it's one of the best, because of their passion and share.

Sidiki

[BertTuyt]

Herewith the results and .pdn files related to the match Kingsrow 1.60b - Scan 3.1
From the perspective of Kingsrow 1.60b

1 min/game 1W, 2L, 155D
2 min/game 0W, 1L, 157D
3 min/game 0W, 1L, 157D
4 min/game 0W, 1L, 157D
5 min/game 2W, 1L, 155D

Bert

[hendrikv]

Cool combination from KR in game 77
https://toernooibase.kndb.nl/opvraag/ap ... ed=1198147

[Event "Diversen KNDB "][Site ""][White "Kingsrow (Ed Gilbert)"][Black "Scan (Fabien Letouzey)"][Result "2-0"][GameType "20"][Round "1"][Date "2019.09.22"][WhiteTime ""][BlackTime ""][WhiteUrl "https://toernooibase.kndb.nl/opvraag/li ... ?SpId=6376"][BlackUrl "https://toernooibase.kndb.nl/opvraag/li ... SpId=15107"][WhitePhotoUrl "https://toernooibase.kndb.nl/Afbeelding ... s/6376.jpg"][BlackPhotoUrl "https://toernooibase.kndb.nl/Afbeelding ... /15107.jpg"][WhiteFlagUrl "https://toernooibase.kndb.nl/Clublogos/USA.gif"][BlackFlagUrl "https://toernooibase.kndb.nl/Clublogos/Frankrijk.gif"] /FEN "W:W26,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50:B1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,20,24"/ 1. 37-31 18-23 2. 41-37 12-18 3. 31-27 07-12 4. 33-28 24-29 5. 28x19 14x23 6. 39-33 10-14 7. 33x24 20x29 8. 35-30 14-19 9. 46-41 09-14 10. 36-31 05-10 11. 41-36 17-22 12. 30-25 15-20 13. 34-30 20-24 14. 40-35 10-15 15. 45-40 04-09 16. 44-39 11-17 17. 39-33 14-20 18. 25x14 09x20 19. 30-25 01-07 20. 25x14 19x10 21. 33-28 22x33 22. 35-30 24x44 23. 50x19 13x24 24. 27-21 16x27 25. 31x13 08x19 26. 26-21 17x26 27. 37-31 26x28 28. 38-33 29x38 29. 43x05

[Ed Gilbert]

Last year I played thousands of games between scan 3 and kingsrow 1.60b, and concluded that scan is stronger. I no longer have the results of those tests, but I do remember that while they are not that far apart, scan is the clear winner. I should also remind everyone that Fabien really broke new ground with his technique of using board patterns with logistic regression to make a very strong eval, which he generously shared with us. Without this help kingsrow would be much weaker than it is now.

Scan is the strongest international draught program, of that I have no doubt. Regardless of the results of small tournaments consisting of a few tens of games that might sometimes have other winners, no other program can be considered the strongest until it has beaten scan in a match of many hundreds, or preferably thousands, of games.

-- Ed

[Jelle Wiersma]

I am sorry to write so, but you have to ask the developer of the program Sjende Blyn.

Krzysztof.

Why didn't you ask him? You asked me.

But all right: Jelle, can Krzysztof play a match Sjende Blyn vs Scan 3.1?

Jan-Jaap

Hi Jan-Jaap,
You are right !!!! I think that all of the members of this forum are as a family or friend, i suppose it.
I already asked this question to Krzysztof : NO ANSWER.
And you are the second person.

Some persons among us refuse to share anything. We know that the fact that Ed Gilbert and Bert initiated the match engines on this forum brought so match to all the programs by sharing weakness of strongest parts of the engines, without forget Michel of Dragon ;this was followed by Scan of Fabien that completely share and make free his program and code.

This also brought a huge progress in the world of draughts because many program after this release was improved in strenght.
You, Jan-Jaap accepted to make a Windows version of Maximus and share it with Krzysztof for bring a plus in the last tournament and the world of draughts.

Why some program are so strange, Sjende Blyn, Damy.
Anyone is free to share or don't sell his program. Kingsrow was a selled program, but it's free now. All this to help the draughts world community. I dont ask to everybody to give their programs, so their works and time for free, no. If the owner want to sell or no, it's a legal choice. But when you come in a community and only take all of the others members and refuse to share something, i think that this it's very very very very unkind.

I asked to Jelle 1 or 2 years ago to share or sell me Sjende Blyn to compare with Dragon, Damage, Kingsrow and Scan. The answer was NO.
He however accepted only to share it with Krzysztof for the tournaments with all the updates.
I asked to Krzysztof to share, and the answer was NO again.

A simple dxp game to have also an idea of the strenght of the program is't also forbidden. A very big shame.
So, this is for Krzysztof, have asked to Ed, Bert, Michel, Fabien and Jan-Jaap before making all these dxp games between their programs?

I see that, Bert already gave the answer, there is some tournament programs and match program. The second category are the strongest, and Sjende Blyn it's among the first it's why all this kind of mystery it's done around a match. However, Jelle wrote that a match was done against Scan and finished by draw, we never saw the games.

Sidiki

Hmmm, maybe I should monitor this forum to correct some false statements about me...

[Sidiki] I asked to Jelle 1 or 2 years ago to share or sell me Sjende Blyn to compare with Dragon, Damage, Kingsrow and Scan.
I retrieved your mail of April 28th, 2018: "I want to learn his playing style to help my game improvement. I have only a few game of your wonderful program. Please, i beg you topermit to me to get a copy to train Thank.
- It mentions personal training, not playing matches against other programs.


[Sidiki] However, Jelle wrote that a match was done against Scan and finished by draw, we never saw the games.
I cannot remember I wrote about this somewhere on this forum, please show me!. The only source I can re-call for this is an excel sheet with some rating calculations I mailed to Bert: Somewhere in 2017 I played a match of 100 games against Scan 2.0, each program 15 minutes per game, the result was 100 draws.
I think tournaments are more exiting.
I never played a match against Scan 3.x, Maximus, Kingsrow or Dragon, if I wanted to, I would have done so.


[Sidiki] A simple dxp game to have also an idea of the strenght of the program is't also forbidden.
When I gave my program to Krzysztof to play in his tournaments, I did not make a restriction.

So Sidiki, I don't like these kind of statements, especially not when their intend seems to be to "force" me into a series of matches...

And now hopefully more informational for this community:
Sjende Blyn won the Unofficial World Championship, but the same program may loose an individual match against the four just mentioned programs. To elaborate on this, I share my answer when Krzysztof mailed me on September 21st with a (bit unclear) question about playing matches:

I do not entirely understand what you mean...
However, if you want Sjende Blyn to play matches against other top programs, I should develop a different engine:

A man, or a program, will not win a tournament when it plays draws.
The current engine takes (big) risks in order to win against weaker engines / humans.
The drawback is, that it also increases the chance to loose against top engines like Scan etc.

Like Jan-Jaap (Maximus) said on the forum in the "Unofficial World Championship" topic:
"In a tournament winning is important, in a match not losing is important."

This means that for matches, my engine should not take these risks anymore, but should play for draw...
Seems boring...!

While writing this, I just started thinking about how to program this...
When I have developed something, I will let you know, but this may take a while, and may never come due to lack of time for that development.


Maybe this explains.

[BertTuyt]

What we could introduce in Draughts Programs is a Contempt Factor, and before the Tournament the Author decides what the factor is for each opponent.
How to program it is another story.....

Bert

[Krzysztof Grzelak]

I agree with Wieger that Krzysztof did a great job in organizing the tournament, which for me was a reason to re-join this forum.

Where a tournament gives a clue what the best Tournament Engine is (although I guess we suffer her from sufficient statistics), the Engine Match determines the strongest engine.

Here my personal belief is that Kingsrow and Scan are strongest, and at least the result for Maximus seems to confirm this statement partially.

I played 5 (2 ballot, 158 games) matches this week Kingsrow - Scan, with thinking time (for every game) 1 - 2 -3 - 4 and 5 minutes.
In a previous match I reached a black hole situation with 158 games, and 158 draws, which was now not the case.
Will share results later today in the evening, with the pdn files.
A teaser, based on the results it is still not decided which of the 2 (Kingsrow or Scan) is better.

I also started a match between Damage 14.0 and Kingsrow, unfortunately the program crashed.
This was new for me, so I need to dig deeper into this.
Basically Krzysztof experienced the same as Damage 14.0 does not work on his ThreadRipper.
In my case I also never experienced a crash, but now I use a new computer with 6 cores, so need to find out what is going on (not in the next 2 weeks, due to Business obligations).

Bert

Once again, thank you everyone. I will help you a bit, Bert, regarding Damage.During program testing, I noticed that:

Write that you have 6 cores = 12 threads - the program will work
If you had 9 cores = 18 threads - as far as I remember the program should still work
I have procesor AMD Ryzen Threadripper 1950X - 16 cores = 32 threads the program will not work an error appears.
I tested it on VirtualBox.

[Krzysztof Grzelak]

A question came to my mind: how many people watched such a tournament?

[Krzysztof Grzelak]

Match SJENDE BLYN - KINGSROW (2-move ballots)

Postponed to another date.

[Ed Gilbert]

Krzysztof, thank you for running the match.

I would like to point out what looks like a problem with your match configuration. You have sjende blyn set to use 16 search threads and pondering on, and kingsrow to use 8 search threads and pondering on. That means that both program are searching at the same time using a total of 24 threads. But your CPU only has 16 cores. If I have all these details correct, it means the two programs are interfering with each other, unless they are running on separate computers. The simple solution is to turn pondering off when both programs are running on the same computer.

edit: I will add that even if the sum of search threads of the two programs was no more than 16, the two programs are probably still interfering with each other to some extent, as they share access to some cache blocks, hard drive, and perhaps other resources. My advice is to never use pondering when both programs are running on one computer.

-- Ed

[Krzysztof Grzelak]

Hi Ed.

I am not able to change it. Jelle is making the program available up to the processor and the memory of frames. The thinking is still including.

[Ed Gilbert]

I think Jelle should be able to give you a way to turn pondering off. This is not just a problem for DXP matches. It is a problem in any of your tournaments.

[Krzysztof Grzelak]

I think Jelle should be able to give you a way to turn pondering off. This is not just a problem for DXP matches. It is a problem in any of your tournaments.

I think the problem during tournaments is not. Tournaments are played on two separate computers. Pondering doesn't matter.