It walks like a clone, it quacks like a clone ...

[AndrewGrant]

Ps. Please be kind sir

huh? perpetrator playing victim, seems pretty standard these days

let me refresh your memory, you wrote this:

Almost no discussion that is relevant to development of chess engines goes on in this forum, why should we be active here? Talkchess really is not the centre of the chessprogramming world.

with about 20 posts total and exactly 0 posts in the programming subforum
have you ever visited there? have you ever responded to or helped someone there?

posts like yours aren't "kind" or even useful

seems like a typical Dunning-Kruger to me

to sum up: I don't give a shit about you or your Koivisto team, "sir"

I think you missed his point. Top engine devs are not lurking around in the talkchess subforums anymore. They are in discord servers. Whether that be Stockfish, Leela, or smaller ones like their own Koivisto one or OpenBench's discord.

Talkchess is a place to post releases, games and results, and call out clones. No real developments are taking place here any more. Its been a long time since the old guard of Hyatt, Tord, Lefler, or others -- typed up lengthy posts in the programming forum.

[mar]

I think you missed his point. Top engine devs are not lurking around in the talkchess subforums anymore. They are in discord servers. Whether that be Stockfish, Leela, or smaller ones like their own Koivisto one or OpenBench's discord.

Talkchess is a place to post releases, games and results, and call out clones. No real developments are taking place here any more. Its been a long time since the old guard of Hyatt, Tord, Lefler, or others -- typed up lengthy posts in the programming forum.

hmm, maybe because "top engine devs" never post anything even remotely useful in the programming subforum, if at all.

also - the world doesn't revolve about top engine devs, like I sad, many new and aspiring programmer can find help there, which is way more useful in general than discussing patches (or whatever) in some private discord server

[AndrewGrant]

I think you missed his point. Top engine devs are not lurking around in the talkchess subforums anymore. They are in discord servers. Whether that be Stockfish, Leela, or smaller ones like their own Koivisto one or OpenBench's discord.

Talkchess is a place to post releases, games and results, and call out clones. No real developments are taking place here any more. Its been a long time since the old guard of Hyatt, Tord, Lefler, or others -- typed up lengthy posts in the programming forum.

hmm, maybe because "top engine devs" never post anything even remotely useful in the programming subforum, if at all.

also - the world doesn't revolve about top engine devs, like I sad, many new and aspiring programmer can find help there, which is way more useful in general than discussing patches (or whatever) in some private discord server

Possibly because most of engine development has left the realm of human understanding. Its mostly trial and error. Those with the best instincts only have a marginal edge in predicting success of ideas.

Idk, back in the original days of talkchess (CCC), there were not resources like the wiki, youtube videos, tons of blogs and articles. Nor were there dozens of extremely high strength open source engines with varying degrees of code quality.

Death by neglect? or Death by irrelevance? or somewhere in between.

[mar]

Possibly because most of engine development has left the realm of human understanding. Its mostly trial and error. Those with the best instincts only have a marginal edge in predicting success of ideas.

Idk, back in the original days of talkchess (CCC), there were not resources like the wiki, youtube videos, tons of blogs and articles. Nor were there dozens of extremely high strength open source engines with varying degrees of code quality.

Death by neglect? or Death by irrelevance? or somewhere in between.

like I said - people who start need to understand the basics, speaking of AB engines there're still many big ideas present in all the engines across the elo spectrum whether you like it or not. after all - you need a working, bulletproof movegenerator first, even though it's not where the strength comes from.

wikis/videos/articles/open source won't help you to track problems, so for some it's still quite relevant (some discussions still happen though).
perhaps you imply that engines below a certain elo range are irrelevant - could be, but some (most?) of us do this for fun

I understand that the modern way of learning - "dissect and reengineer" is more efficient and I don't really have a problem with that.
I have problem with people crapping on a whole forum with post #20. try it elsewhere and let me know how it went

[AndrewGrant]

Possibly because most of engine development has left the realm of human understanding. Its mostly trial and error. Those with the best instincts only have a marginal edge in predicting success of ideas.

Idk, back in the original days of talkchess (CCC), there were not resources like the wiki, youtube videos, tons of blogs and articles. Nor were there dozens of extremely high strength open source engines with varying degrees of code quality.

Death by neglect? or Death by irrelevance? or somewhere in between.

like I said - people who start need to understand the basics, speaking of AB engines there're still many big ideas present in all the engines across the elo spectrum whether you like it or not. after all - you need a working, bulletproof movegenerator first, even though it's not where the strength comes from.

wikis/videos/articles/open source won't help you to track problems, so for some it's still quite relevant (some discussions still happen though).
perhaps you imply that engines below a certain elo range are irrelevant - could be, but some (most?) of us do this for fun

I understand that the modern way of learning - "dissect and reengineer" is more efficient and I don't really have a problem with that.
I have problem with people crapping on a whole forum with post #20. try it elsewhere and let me know how it went

What exactly is the incentive to help people who come into the forum working on their first ever move generator? Obviously its a kind thing to do. Maybe you can argue it helps bring more people into the community, which is a net gain in the long run. But on an individual level, why should I, or a Stockfish dev, or someone else spend the time weeding through someone's first ever chess code with hundreds of lines to hunt down their bugs for them?

Seems like asking someone a question when you can just google it -- even though its great to talk to someone about it instead of just reading something that has already been written on the topic. I'm interested in the followup questions you know? Not the entry ones.

Just my thoughts.

[hgm]

By us I meant the Koivisto team...

Never heard of it!

[Michel]

I think you missed his point. Top engine devs are not lurking around in the talkchess subforums anymore. They are in discord servers. Whether that be Stockfish, Leela, or smaller ones like their own Koivisto one or OpenBench's discord.

I think that's a pity. Discord is a chat program and the signal to noise ratio is usually very low. It is not a good place to preserve information (there are no threads for example). Sadly many people seem addicted to it.

[Madeleine Birchfield]

I think you missed his point. Top engine devs are not lurking around in the talkchess subforums anymore. They are in discord servers. Whether that be Stockfish, Leela, or smaller ones like their own Koivisto one or OpenBench's discord.

Talkchess is a place to post releases, games and results, and call out clones. No real developments are taking place here any more. Its been a long time since the old guard of Hyatt, Tord, Lefler, or others -- typed up lengthy posts in the programming forum.

hmm, maybe because "top engine devs" never post anything even remotely useful in the programming subforum, if at all.

also - the world doesn't revolve about top engine devs, like I sad, many new and aspiring programmer can find help there, which is way more useful in general than discussing patches (or whatever) in some private discord server

The programming and technical topics subforum of Talkchess, and Talkchess as a whole in general, is mostly for alpha-beta engines that run on the CPU. Most beginner chess engine developers and network trainers coming from the machine learning community and looking to develop Leela/AlphaZero style engines with GPU backends, MCTS search, and deep residual convolutional neural networks are talking on the Leela discord instead of Talkchess about those technical topics, or otherwise on github and various other machine learning forums. Stuff like piece square tables, NNUE, minimax/alpha-beta search, null move reductions, quisecence search, move ordering heuristics, etc, commonly found on the programming subforum in Talkchess do not help such an engine developer at all.

Most of those efforts go unnoticed here in this forum, apart from the well known Leela project, projects by already prominent members of Talkchess (Allie - gonzochess, Scorpio - Daniel Shawul, Stoofvlees - GCP, a0lite - dkappe), and commercial clones of Leela like Fat Fritz. And a0lite and Scorpio 3.0 isn't even on any of the rating lists.

[hgm]

It seems more like they are set in their own ways, not really interested in other opinions. Aren't Leela-type NN engines still using that crappy MCTS/UCT, and aren't NNUE engines still using a network architecture designed for Shogi?

[Madeleine Birchfield]

Aren't Leela-type NN engines still using that crappy MCTS/UCT

For Leela type NN-engines, the alternatives to PUCT tested so far, like betaMCTS and RENTS/TENTS, have been incorporated into various engines and found to be slightly worse than Leela's PUCT when using the same net. I have seen no discussion about MCTS altermatives to PUCT on Talkchess whatsoever, with people preferring to talk about the latest hot topics in CPU engines (NNUE, cloning and simex, handicap games vs Komodo, etc.)

AB/minimax is inferior to PUCT for Leela-style networks because AB/minimax search won't be able to search deep enough with massive and deep neural networks compared to PUCT/MCTS, and overall deeper search depth makes the engine stronger, even if from time to time the engine might miss a critical line. This has been known in the Leela community for almost two years now. If AB/minimax wasn't inferior to PUCT/MCTS, Allie would have been using AB/minimax by now, as it originally started as an attempt to combine deep learning with AB/minimax search.

This is why I view Talkchess as primarily just a forum for traditional CPU based engines.

It seems more like they are set in their own ways, not really interested in other opinions....aren't NNUE engines still using a network architecture designed for Shogi?

This issue has been discussed on Talkchess extensively in the past six months. There would be a lot more variety of architectures if they all developed their own architectures from scratch like what Pedone, Ethereal, Orion, and Seer did, but most CPU engine developers in the Talkchess community that adopted NNUE have decided just to copy the nodchip architecture from Stockfish and tinker with it. And then you have older people who have been in the community for two or three decades like mclane who are still complaining about the use of alpha-beta minimax search as opposed to 'selective type B' search, ignoring the fact that selectivity heuristics make the minimax search into a type B search.. So yes, it does seem that most people in the Talkchess community are set in their own ways, not really interested in other opinions.