@Talkchess guys: What is pre-historic chess programming about?

[AndrewGrant]

I'm curious to know what chess programming was like in the stone ages, when authors seldom tested their changes with any statistical significance. It must have been more fun back then, as one could find a +10 elo idea was ease, as the space was so unexplored. I guess some things remain the same now even though, regarding low-effort cloners.

Discuss

[Viz]

I kinda miss times where you could make +10 elo with crappy implementation of original ideas.
Nowadays sf gainers average some 0,5 elo/patch, sad days.

[BrendanJNorman]

Here's how I see it.

"Prehistoric" chess programming was when chess nerds (I say this with great affection) used programming to try to make computers play good chess.

Modern chess programming is when programming nerds (also no disrespect ) use chess to try to show how good they are at programming.

In the first case, chess is the passion and guiding light.

In the second, chess is a tool that is being used.

Just an long-time observer's view. No interest in your tears or flame-throwers.

[smatovic]

I started 2010 with GPGPU and chess, back then people talked about 100x speedups via GPUs, and everybody was excited about what you can do with GPUs and what not, exploring new territory....now the path is laid down.

--
Srdja

[AndrewGrant]

"Prehistoric" chess programming was when chess nerds (I say this with great affection) used programming to try to make computers play good chess.

Modern chess programming is when programming nerds (also no disrespect ) use chess to try to show how good they are at programming.

First time you've said something I agree with

[Rebel]

I started in 1980, if memory serves me well still active members are here like Chris, HGM and Joost started even earlier. Speaking for myself, I had zero knowledge, no single reference point, it was like sitting on an uninhabited island, no internet, no email, no forums, no contact with anyone. I started in BASIC defining an array of 64 bytes and wondered what to do next. And from one thing came another, my first versions did not even had alpha/beta because I never heard of it.

[BrendanJNorman]

"Prehistoric" chess programming was when chess nerds (I say this with great affection) used programming to try to make computers play good chess.

Modern chess programming is when programming nerds (also no disrespect ) use chess to try to show how good they are at programming.

First time you've said something I agree with

I'm a smart dude, so it probably won't be the last time.

[hgm]

The most important difference is that in those days there existed many chess engines, and if you did not test your ideas by playing gauntlets against a variety of opponents, the results were often meaningless. That required 4 times as many games to get the type I and type II errors below the desired thresholds than using self play, and of course computers were several orders of magnitude slower, so that it took a lot of resources to generate the games in the first place.

Now that all engines are basically the same, playing gauntlets still amounts to self play, and makes no sense anymore. It is much easier to design an engine for just beating one particular opponent, than it would be to develop an engine that is generally good.

[mar]

what makes you think people around here don't test patches?

I don't do sprt (sue me), not a member of any OB instance or discord and surprisingly I can still breathe and enjoy the hobby

Ed's IIR originally got me 20 elo, back when my engine was 3k using HCE;
some ideas may gain more than elsewhere - not everything is 3.6k. this is perfectly normal

btw - could you check your pyrrhic repo? there's a PR waiting to fix a reinit crash after tb_free due to dangling pointers

[Uri Blass]

The most important difference is that in those days there existed many chess engines, and if you did not test your ideas by playing gauntlets against a variety of opponents, the results were often meaningless. That required 4 times as many games to get the type I and type II errors below the desired thresholds than using self play, and of course computers were several orders of magnitude slower, so that it took a lot of resources to generate the games in the first place.

Now that all engines are basically the same, playing gauntlets still amounts to self play, and makes no sense anymore. It is much easier to design an engine for just beating one particular opponent, than it would be to develop an engine that is generally good.

I disagree.
Today there are more chess engines relative to what you had in the past.

It is clear that the top engines stockfish,lc0,Berserk,KomodoDragon are different engines.
I do not understand why you claim that they are basically the same.