Can you make a professional Janggi (korean chess) engine?

[maksimKorzh]

Hi @cjssh1002 re: Janggi's rules are similar to xiangqi rules

• apart from cannons moving diagonally if I'm not mistaken?

re: f you are going to do this, please do janggi as well

• I think adding jiangi as variant is possible and should be easilly achieved without dropping performance. So adding jiangi variant might be an option, however not the prior. I still need to do lots of things.

re: If you develop a janggi specialized engine, it will be a good insurance for the future.

• I thought the same about xiangqi... well, it doesn't seem to be so, we'll see

re: "professional" engine:

• this is the most arguable. If you're a developer and have a look at the source code, you'll understand what I mean. But from the perspective of playing strength my version should be at least as strong as Fairy Stockfish.

So let's say I accept this issue and will keep it open, however I don't promise quick results. Hope you understand.

[maksimKorzh]

@cjssh1002 Did you have a look at current xianqqi stockfish code base? (movegen.cpp in particular)? What do you think?

Let's go the following way:

1. I embed alpha-beta search
2. We work on jianggi variant

How about that? I can ping you when done.

Another note: What GUI do you use to play jiangqi with fairy stockfish? I think my CCBridge Arena can be altered to serve as jianggi GUI, did you have a look at my GUI? https://www.youtube.com/watch?v=ZEnVvoXBPho

[maksimKorzh]

No, Fairy Stockfish's code won't work because of bitboards. I'm using array based move generator here. I will search for the rules when the alpha-beta routine is done. Anyway only movegen and eval code would be different while all the rest should be the same.

I'm just wondering why do you need other than Fairy Stockfish engine? Do you want to run engine matches? Does Fairy Stockfish support winboard protocol natively?

[maksimKorzh]

@cjssh1002 A few notes:

1. Fairy Stockfish is developed by a single person - Fabian Fichter, not by a team as far as I'm aware.
2. Fairy Stockfish is not the only jiangqi engine, H.G. Muller, maintainer of winboard is now working on jiangqi engine: http://talkchess.com/forum3/viewtopic.php?f=7&t=76465 It should have a very promising evaluation and obviously compatible with winboard.

I clearly understand your requests from the user's perspective, I had something similar in mind which resulted creating my ccbridge arena - a proof of concept gui to play games between UCI/UCCI xiangqi engines, but when I got what I currently have I started feeling even more disappointed then earlier)

Another issue regards to interaction between engine and GUI. I tried uci2wb adapter and it works but it's too complicated for the end user to setup the engine this is probably the reason why Fabian added native CECP support to Fairy Stockfish to run natively in winboard. It's possible to reuse that obviously but anyway with my current skills you can't expect much improvement comared to Fairy Stockfish. The reason for that is that for evaluation I use material and piece-square tables (to give engine an idea of where to develop it's pieces) only. I don't know for sure, but I guess that's something similar to what is used in Fairy Stockfish itself. So the "style" of play would be very very similar to Fairy Stockfish. I just have a faster move generator and potentially might have as good search as Fairy Stockfish uses (but without threads for now). In order to get "interesting" and "human-like" style of play NNUE is needed. Fabian is going to make it for various variants for Fairy Stockfish and I'm gonna reuse it then (I'm too dumb to create my own NNUE).

Just tried to share my concerns as a developer. Btw are you a developer? Did you ever write a chess engine? Did you ever think to do so if not?

[maksimKorzh]

Thanks for your questions. Your request is accepted, however I can't say for sure If I go for implementing jiangqi or not. If yes - you'll be the first person to know about it)

Without having a decent search in the engine thinking about different variants is pointless and I've just started developing search: https://www.youtube.com/watch?v=3sLgdJGJGk0

It takes quite a bit of time to bring it to the state somewhat close to Fairy Stockfish)

[maksimKorzh]

@cjssh1002 I just had a look at jiangi rules at wikipedia, well... it's a different game compared to xiangqi and I've realized that I just don't want to complicate current engine's move generator. However the game itself (jiangqi) seems to be really exciting. Here's the point - despite the fact it's not obvious to be able to play a game in order to create playing engine, still I prefer to master a game at somewhat novice level first. This is what I did before coding xiangqi engine. So to start a jianggi engine I first need to lose around 100 jianggi games as a human player)

Now here's the idea that has just come to my mind - the project you're requesting might be done much faster if we deal on a paid bases (fixed price paid by milestones). You can mail me to freesoft.for.peole@gmail.com for details.

Some benefits you might have in this case:

1. Engine may be implemented in different programming languages with different purposes in mind, for instance:
   • in javascript (this would allow to play in browser directly, see playable example: https://maksimkorzh.github.io/wukong-xiangqi/src/gui/xiangqi.html but also within winboard, see how I played my xiangqi engine written in javascript versus maxQi by H.G. Muller, maintainer of Winboard: https://www.youtube.com/watch?v=NmXhAUnKLe0)
   • in C/C++ (would be faster and stronger but won't have own UI, however it's possible to make a custom UI like my CCBridge Arena: https://github.com/maksimKorzh/ccbridge-arena)
2. Priority (paid projects are prior to hobby projects, so I can drop what I currently do and start working on jianggi project instantly)
3. Customization & revising (literally on-demand feature requests/upgrades/support etc.)

Potentially this can be turned into commercial project you might want to start selling one day (but in this case it would be a custom stuff, no longer derived from stockfish to avoid licencing issues)

I would say that this is something I'm interested to discuss in particular.

On the other hand on free bases somewhat similar project might occur in a year or two, but again like I can't promise for sure because my hobby stuff flows on it's own - I mean I do something for free IF that is interesting to me and WHEN it is interesting for me and not on demand, hope you understand)

[maksimKorzh]

re: it seems impossible to surpass Fairy Stockfish in the short term

• I'm happy you understand it)

re: Finally, in fact, I am currently developing janggi with the Winboard developer

• you mean with H.G. Muller? Congrats!

re: ...because I have to help the Winboard developer...

• yeah, for sure it's better to help H.G. Muller especially bearing in mind you're a pro player. I believe this should result in great engine!

re: ...as a result, ... users can make donations

• haha) I though this would be so for xiangqi, but I've gifted fo free all I've developed)))

Today I played a few jiangqi games versus fairy stockfish - this was really fun) Jiangi has some unique charm no that no other variant have) I managed to beat random mover but obviously got crushed by any stronger levels)))

To be honest making a playing engine for whatever variant is not really that big deal if you have experience with chess programming, however the tricky thing is to set up proper goals ones you start a project. It's still a big issue for me - my original goals are always "simplest possible proof of concept implementation" no matter how complicated project is and this always leads to a working prototype I completely lose interest to after starts initially working. With this current xiangqi stockfish I have the same issue - I already see that I can bring it to somewhat a decent playing strength but unfortunately I can't see much sense in further development for some reason. All I do kind of lives as a youtube show/series and ones it's completed I usually start a new one... I wish I could bring all the stuff I've done so far together into a single product, but it's a complete mystery to me what kind of stuff can lead to a common denominator.

[maksimKorzh]

Hi @cjssh1002 I've made a traditional janggi theme for winboard/xboard: http://talkchess.com/forum3/viewtopic.php?f=2&t=76959 maybe you get interested.

Also this xiangqi stockfish project is now stopped due to some conflicts I have in my head))) So I'm now starting to research janggi and hope to come up with another engine, just like you wanted) I will first port my wukong-xiangqi from JS to C and then alter the movegen to play janggi. It should be at least as strong as HGM's J 0.2b, btw did you already test that?

P.S. It's such a pleasure to watch engines playing janggi with traditional board and pieces!

[maksimKorzh]

Hi @cjssh1002

First of all there's a typo in the email, it should be: freesoft.for.people@gmail.com (peoPle, not peole)

re: images on windows 10

• Some theory first. I use wine (windows environment emulator, windows 10 is emulated in particular) to run winboard. So the issue you're facing is 99% image path issue, not OS issue. I can't guarantee that it would work with your current version of winboard, but the version HGM has sent me should work (download this repo: https://github.com/maksimKorzh/chess_programming see /janggi/WinBoard/WinBoard folder for the exact version of WinBoard, it's configured to run janggi by default).
• Now regarding images. You need to specify YOUR OWN PATH to images becasue otherwise it would most likely rely on the path specific to my environment. In the video I'm showing how to do that.

Start with board texture, this should be trivial:

1. Open View=>Board Themes...
2. Specify Light-Square texture and Dark-Square texture (pick up "Janggi.bmp" from /theme_winboard folder), click on Use board textures checkbox.

It should display the board texture. If not try a whatever custom ".bmp" file with the same size (width and height in pixels) this would tell whether the issue in the texture encoding or not.

3. Specify Directory with piece bitmaps as well

It SHOULD work, please let me know if it helps.

re: engine related questions

• all ideas are very exciting, give me a bit of a time, I'll creae a new repo for janggi engine and we'll move this discussion there

Looking forward.

P.S. Email is also fine if you prefer private discussion.

[maksimKorzh]

@cjssh1002 I moved the discussion here: https://github.com/maksimKorzh/wukong-janggi/issues/1 This issue would be closed.