The future of computer chess

[Dann Corbit]

Both hardware and software have advanced in Elo exponentially.
Not only is the hardware c0 * 2^k stronger, where k is time in years, but the software is c1 * 2^k stronger.
Is the chess more beautiful, or more inscrutable?
Beats me.

[towforce]

The problem is that programmers have not 6502 to test so I guess engines of today without adjustments cannot work on very old hardware.

Well... there are 6502 emulators available - link.

Let's be clear - nobody is actually going to do this - but if I were going to do it, I'd ask for suggestions from dev websites - but my first thought would be:

1. Transpile the source code to a language that can be compiled for the 6502

2. Compile the source code for a 6502

3. Run it against an a 6502 emulator

I suspect that the process would be tricky - but probably possible. Some of the original programs were written in assembly language though, so are likely to be efficient on that platform. A program written for a modern CPU is not likely to be efficient on a 6502.

All in all, the programs written in the golden era of dedicated chess computers are likely to continue to be the best chess programs ever written for 8-bit CPUs. They live on thanks to the work that Franz Huber did recreating them in emulation.

[Uri Blass]

The problem is that programmers have not 6502 to test so I guess engines of today without adjustments cannot work on very old hardware.

Well... there are 6502 emulators available - link.

Let's be clear - nobody is actually going to do this - but if I were going to do it, I'd ask for suggestions from dev websites - but my first thought would be:

1. Transpile the source code to a language that can be compiled for the 6502

2. Compile the source code for a 6502

3. Run it against an a 6502 emulator

I suspect that the process would be tricky - but probably possible. Some of the original programs were written in assembly language though, so are likely to be efficient on that platform. A program written for a modern CPU is not likely to be efficient on a 6502.

All in all, the programs written in the golden era of dedicated chess computers are likely to continue to be the best chess programs ever written for 8-bit CPUs. They live on thanks to the work that Franz Huber did recreating them in emulation.

I do not believe that the programs written in the golden era are likely to continue to be the best in case somebody make the effort to translate the best relatively simple chess engines of today to that hardware.

Even something that is not efficient and 10 times slower should be able to beat them because constant speed advantage of the old engines is going to be smaller than the advantage new engines get from a better algorithm(both search and evaluation) that people found later.

[Viz]

The only reason why this programs will always remain the best on this hardware is because no one would ever bother to make a better program for it.
For times of their creation it was an epic achievement, of course. But with modern day knowledge this things would be beaten pretty easily after literally weeks of work of someone who is capable of writing smth for this hardware.
Like 4ku on 1 core has almost the same rating as crafty 25.2 while having 4kb weight executable. This is difference between modern knowledge and ancient knowledge.

[mar]

The only reason why this programs will always remain the best on this hardware is because no one would ever bother to make a better program for it.
For times of their creation it was an epic achievement, of course. But with modern day knowledge this things would be beaten pretty easily after literally weeks of work of someone who is capable of writing smth for this hardware.
Like 4ku on 1 core has almost the same rating as crafty 25.2 while having 4kb weight executable. This is difference between modern knowledge and ancient knowledge.

two things:
- 4ku is already stronger than Crafty (single core)
- it's not technically a 4kb executable, but a script with compressed minified source code which executes a compiler to produce a binary

with this logic, you could easily write a much smaller script which clones a git repo instead and produces a much stronger engine; of course the compiler itself weighs way more than 4kb

[Uri Blass]

The only reason why this programs will always remain the best on this hardware is because no one would ever bother to make a better program for it.
For times of their creation it was an epic achievement, of course. But with modern day knowledge this things would be beaten pretty easily after literally weeks of work of someone who is capable of writing smth for this hardware.
Like 4ku on 1 core has almost the same rating as crafty 25.2 while having 4kb weight executable. This is difference between modern knowledge and ancient knowledge.

two things:
- 4ku is already stronger than Crafty (single core)
- it's not technically a 4kb executable, but a script with compressed minified source code which executes a compiler to produce a binary

with this logic, you could easily write a much smaller script which clones a git repo instead and produces a much stronger engine; of course the compiler itself weighs way more than 4kb

I agree but if you take hardware from 2010 people always write better engines for the same hardware and I believe stockfish can easily run on hardware from 2010 and there are hundrends of elo improvement between shredder that was the top engine in 2010 and stockfish.

I see no reason that it is impossible to do it for significantly older hardware.
ssdf is the only list that I know to have the same engines with different hardware but the problem is that they usually prefer not to test new engines with a very old hardware even if they can use the old hardware.

IHere is a comparison for the same hardware that show improvement of software with the same hardware(both were the best of their time)
2 Stockfish 16 x64 1800X 3.6 GHz 3582 50 -48 200 58% 3523
32 Stockfish 8 x64 1800X 3.6 GHz 3415 19 -19 1520 70% 3271

96 Deep Rybka 3 Athlon 1.2 GHz 3068 36 -35 372 53% 3044
140 Shredder 7.04 UCI Athlon 1.2 GHz 2796 21 -20 1269 66% 2683

The main problem with very old hardware(not hardware that is 10 years old but hardware that is 40 years old) is that nobody use them so programmers have no interest to develop engines that can use them.

It is clear to me that if they wanted they could do something better for the very old hardware and it does not mean to say something bad about the people who started computer chess because they had less knowledge in the first place to develop strong engines and today we know better.

I do not claim that the programmers of today could do something better in case they lived 40 years ago to develop chess engines but only that it seems obvious that with the knowledge that they have today they can develop something better software for the hardware that people had 40 years ago.

[smatovic]

[...]
This is difference between modern knowledge and ancient knowledge.

Maybe you can elaborate a bit about your view on modern and ancient knowledge? What is modern chess programming about, what was old chess programming about?

--
Srdja

[towforce]

Maybe you can elaborate a bit about your view on modern and ancient knowledge? What is modern chess programming about, what was old chess programming about?

I'll start the list:

1. Auto-tuning

[smatovic]

....well, maybe first define the timeframe what oldschool and newschool is, Discord members mention that CPW and TC are oldschool, and that "we" are missing the point of modern chess programming, so, I myself am interested, what I am missing in context of "what modern chess programming is about", or alike, what is modern chess programming about, and, what was old chess programming about?

--
Srdja

[Viz]

....well, maybe first define the timeframe what oldschool and newschool is, Discord members mention that CPW and TC are oldschool, and that "we" are missing the point of modern chess programming, so, I myself am interested, what I am missing in context of "what modern chess programming is about", or alike, what is modern chess programming about, and, what was old chess programming about?

--
Srdja

Actual statistical testing to get actual statistical signifficance is the most important part which wasn't the case prior to rybka more or less and why rybka was so dominant at it prime.
Ofc other things - SPSA, details of how to implement heuristics that gain - like LMR can be implemented to gain 5 elo or to gain 100 elo depending on implementation, but this is true for almost any known heuristic like null move pruning or futility pruning.