Unofficial World Championship Of Computer Programs In International Draughts 2021

[Krzysztof Grzelak]

I arrived late!

You are not too late. Joost will share when the engine is ready.

[Joost Buijs]

Krzysztof,

I've been working on the engine for several days now and did remove a lot of small bugs, I also made some other improvements.

Right now I'm busy getting pondering working, it is almost there but I want to make sure that it doesn't introduce unwanted behavior. The problem is that I never designed the engine to use pondering, a decision that I now regret, so I had to make rather big changes at several locations in the source code.

Let's say that I will send you the final tournament version this weekend, being it with or without pondering. I will make sure that it runs without problems and that it won't give you a headache by using it.

Joost

[Krzysztof Grzelak]

I have no problem Joost. Thanks Joost.

[Joost Buijs]

Krzysztof,

When I use pondering my engine seems to be working without any problems now. I still want to do some more tests to make sure that it is perfectly stable. As it looks now I can post a link to the final tournament version on the forum later today, this gives you the opportunity to test it on your machine(s) a few days before the tournament starts.

Joost

Edit:

There are a few changes:

It won't accept more threads than you have physical cores (Sidiki had an issue with this running 16 threads on a 2 core machine).
The network file now has a number to avoid confusion with other nets.

The ini file now has:

threads = 16
tt-size = 32
bucket-size = 4
dxp-port = 27531
ponder = on

threads: can be 1 to 64
tt-size: can be 26 to 34 (32 means 4 GB, it's a power of 2 so 30 means 1 GB)
bucket-size: can be 1 to 4 (the number of tt-entries in a bucket, 2 or 4 is probably best)
ponder: can be 'on' or 'off' (default off)

It is compiled for the AMD Zen1, I assume it will work on your Threadripper. When you encounter issues, please let me know so I can fix them.

The major problem with the NNUE network is that it is overall about 3 times slower as a pattern system and that it has some difficulties understanding positions with just a few pieces. This can be fixed with an EGDB, currently I don't have one and I want to see what it can do on it's own. It still plays a very decent game, the fun part is that it often plays totally different moves than pattern programs would do.

Joost

[Joost Buijs]

Krzysztof,

Maybe it's a nuisance, but after carefully watching the output of my engine I found that it didn't select the right ponder-move when the score of the opponents move >= beta. I fixed this behavior in a new version 1.42b. It has no impact on stability but this fix could make the engine a little bit stronger.

I will remove the link to the 1.42a version, you can download the 1.42b version here: https://www.dropbox.com/s/84avb2tv0gxze ... b.zip?dl=0

Sorry for all the hassles. I think I've been programming too much lately.

Joost

[Krzysztof Grzelak]

Thanks Joost.

[pontel]

I arrived late!

You are not too late. Joost will share when the engine is ready.

Thanks in advance to Joost and Grzelak!

[Krzysztof Grzelak]

The final page of the tournament will be released on the last day before the tournament due to frequent changes on the tournament pages.

[Joost Buijs]

Krzysztof,

As promised I've added a small book to my engine. It contains about 100k positions, not very big but I couldn't make more of it on such a short notice. There also was a small bug in the time-control routine that made my engine move to fast, this is now fixed.

I bumped the version to 1.43a, maybe you want to change that on the tournament site.

You can download the latest version over here: https://www.dropbox.com/s/uhm518hfvzcgm ... a.zip?dl=0

I've tested it very briefly, everything looks fine but there is always a possibility that things might go wrong.

Thanks a lot for organizing this tournament! It's amazing that you are willing to spend such an enormous amount of your time on it.

It is my intention to watch the games on the tournament site when I am at home.

Joost

[Krzysztof Grzelak]

Thanks Joost. Recently I came back from work.

[Yves]

Hello, Joost.
Please let me know if it is possible to do play Ares as initiator against Truus who is the server.
Sincerely, Yves.

[Joost Buijs]

Hello, Joost.
Please let me know if it is possible to do play Ares as initiator against Truus who is the server.
Sincerely, Yves.

Hello Ives,

Unfortunately this is not possible at the moment, I already made the code that is needed to be initiator, but it still has to be implemented in the engine. It is on my to-do list, when I have time (and feel like it) I will certainly add it.

Probably I have to add some things in the ini-file like initial number of moves and initial time and send a GAMEREQ message, I will take a look at it.

Joost

[Krzysztof Grzelak]

Therefore, that one of the developers of your program is not released, the tournament changes occur. There will be 34 rounds in the tournament, there will be no pause program. I ask all programmers to check if possible the program settings for the tournament. The tournament starts tomorrow at 8:00 CEST. I wish all the developers all the best and the best possible place in the tournament. Official website of the tournament https://www.uwccpi2021.draughtsprograms.eu/ - I invite you to watch.

[stepanov.anatolievich]

Unfortunately, here this page not work. Can it be realized on YouTube like last year?

[pontel]

The tournament has already started very interesting...
In advance, congratulations Grzelak!

A question, are you using any opening books for Flits 3.1?
I realize that Flits would hardly lose some of the perts I've seen if he was without a book, for example.

I happen to have the version here, and I noticed that he doesn't usually play these moves. His book when activated is pretty incomplete, I ended up adapting the 1.5 version book to see if it gets stronger.