Is chess programming now boring ???

[laurietunnicliffe]

It seems I find myself looking at the list of entries here from back to front. From the oldest to the newest.
This is because all the "juicy stuff" was being invented/tested/implemented all that time ago.
It somehow seems more thrilling to have an idea, implement it, and test it, than to just rely on a neural network
to learn patterns. And of course, you can now just look up StockFish to see how everything is done.
I'm afraid, I have missed the golden age of chess development (although I did buy a lot of tabletop chess computers in their hay day).
I suppose now that chess computers are so far ahead of humans, there is not a lot to be gained. But I guess people will continue and try to "solve chess".
I could get more excited by trying to write more efficient code or more compact code or better algorithms even if the playing strength is not improved.
Anyhow, that's how I see it.
How about you ?

[smatovic]

my 2cents...

The Next Big Thing in Computer Chess?
viewtopic.php?t=81858

Fruit fly races on steroids?
viewtopic.php?t=83267

--
Srdja

[op12no2]

Now and again chess players in General Topics ask for less super-human and more interesting (playing style) programs to have some fun with. CCRL has divisions that can accommodate such programs; so you get the best of both worlds. ELO being just one dimension; there are others kinda-thing. May be worth a poll?

I would guess that creating such a program can be challenging and rewarding if you set yourself some constraints to work within. For example: never looking at any other code (and maybe also the wiki), no overt evaluation function as such - it's all done incrementally with the hash update etc., minimum executable size, develop a syntax so players can define their own evaluation function to play around with, no searching (e.g. just super-fast random playouts, tiny net, fun 'personalities' like moving everything towards to opponents king or the opposite, no tuning automation, etc.

[hgm]

The problem I have with it is that "chess programming" for top-tier programs is no longer programming. It has become data analysis, using standard software, which you sometimes can configure to better fulfil you needs.

But there still can be programming challenges in areas where no 'big data' is available. E.g. in chess variants, where there are no game databases available, and generating millions of games yourself would be at odds with a design goal of configurability by the user for new variants without any tuning effort. It is also much more rewarding, because there are many more people that would use a chess-variant engine than there would be of an engine for orthodox Chess (that is no Stockfish).

[chesskobra]

I would like to see a good endgame engine, based on hand crafted evaluation. After recently finding an example of a 3400+ rated engine taking 123 moves to win Q+B+3P against 1P, I don't trust the ratings. I would be happy to see an engine that plays at FM or IM level in all phases of the game - without too much computation, with hand crafted evaluation. Older engines like crafty and phalanx still fascinate me, even though I have not looked at their code.

[lithander]

For me programming is not a science but a craft. I'm not involved in chess programming hoping to publish a paper. I'm here to hone my skills and have some fun.

The way the community is set up with rating lists and enthusiasts running engine vs engine matches gives you something to aspire to without ever realistically expecting to dethrone one of the top 5 engines. The product (my chess engine) is not very valuable in itself but there are numerous of optimizations techniques to understand and implement. The reward is that you hone your craft and get to climb the Elo ladder as tangible proof of making progress.

You can define the terms of your journey in a way that makes it extra rewarding. So for example C# was considered unsuited for chessprogramming by many which motivated me to get as close to C/C++ performance levels as possible. I'm learning a lot about a language and runtime that I also use professionally. BMI2 and AVX2 intrinsics are so exotic that I never had to use them outside of chess programming.

While you can "just look up StockFish to see how everything is done" you can also not do that. I don't have to look at other engine's source code to first understand and then implement something like null move pruning or late move reductions. Also, developing Stockfish is a collaborative effort with over a hundred individual contributors. How would someone become knowledgable enough to contribute? Probably many of them have chess programming experience prior to their Stockfish contributions.

I agree with hgm that the machine learning aspects transform what started as a set of programming challenges into something else and I already see it becoming increasingly tedious. But it also came with it's own set of unique challenges that I found interesting.

So TLDR: Is it going to be my lifelong hobby? No. Have the past ~3 years been fun and educative? Yes!

[Whiskers]

Now and again chess players in General Topics ask for less super-human and more interesting (playing style) programs to have some fun with. CCRL has divisions that can accommodate such programs; so you get the best of both worlds. ELO being just one dimension; there are others kinda-thing. May be worth a poll?

I would guess that creating such a program can be challenging and rewarding if you set yourself some constraints to work within. For example: never looking at any other code (and maybe also the wiki), no overt evaluation function as such - it's all done incrementally with the hash update etc., minimum executable size, develop a syntax so players can define their own evaluation function to play around with, no searching (e.g. just super-fast random playouts, tiny net, fun 'personalities' like moving everything towards to opponents king or the opposite, no tuning automation, etc.

I’m trying something of this variety and it’s much more fun than Willow dev.
viewtopic.php?t=83309

[op12no2]

I’m trying something of this variety and it’s much more fun than Willow dev.
viewtopic.php?t=83309

Cool, I'll follow along; I like the blog topics.

[syzygy]

After recently finding an example of a 3400+ rated engine taking 123 moves to win Q+B+3P against 1P, I don't trust the ratings.

8-piece tablebases will solve that!

[mclane]

I think Patricia is a very interesting playing engine.