We are Fish. Resistance is useless.

[Rodolfo Leoni]

I think that Stockfish has reached its own limit and can't progress anymore. The SF OS team has started from a well know point and has used well know technology but now they are in a "local maximum". The only way to progress would be to full rewrite the code, starting from different ideas.

"Sorry captain, the fish starship has been captured by a black hole!"

I think limits aren't easy to measure anymore. When 10.000 games are needed to start thinking engine maybe is 5 ELOs stronger, then we could conclude no real progress is possible.

Rodolfo

P.S. To Lyudmil Tsvetkov... You didn't understand (or didn't like) this joke as well as some other people didn't understand (or didn't like) your "Che tu possa morire" joke.

Much to learn you still have, ObiTsvetKenobi.

[Lyudmil Tsvetkov]

ObiTsvetKenobi.

I like this one.

[Lyudmil Tsvetkov]

I think that Stockfish has reached its own limit and can't progress anymore. The SF OS team has started from a well know point and has used well know technology but now they are in a "local maximum". The only way to progress would be to full rewrite the code, starting from different ideas.

"Sorry captain, the fish starship has been captured by a black hole!"

that might not be fully true.

I guess SF can still achieve quite a lot elo on their current code.

problem is, when code necessarily grows(not in terms of lines count, but in terms if bigger number of eval terms and more refined conditional search), how to successfully tune a much larger pool of parameters?

they are having difficulties fine-tuning even the current set.

[Rodolfo Leoni]

ObiTsvetKenobi.

I like this one.

Beware the blunder side of chessforce. Keep concentrated on your road to become Jedi Chess Master. Keep concentrated and say aloud:

"Sotto la panca la capra campa; sopra la panca la capra crepa."

When perfectly you'll say that, a first step you completed will have.

[lantonov]

I think that Stockfish has reached its own limit and can't progress anymore.

The reality is somewhat different. I come to different conclusion after looking at this curve https://nextchessmove.com/dev-builds

[Guenther]

I think that Stockfish has reached its own limit and can't progress anymore.

The reality is somewhat different. I come to different conclusion after looking at this curve https://nextchessmove.com/dev-builds

Thanks very interesting. Actually over the given time frame the progress (in selfplay) still looks linear.
May be with an exception between 10/15 and 03/16, here the progress seems to have been a bit higher.

Guenther

[stegemma]

I think that Stockfish has reached its own limit and can't progress anymore.

The reality is somewhat different. I come to different conclusion after looking at this curve https://nextchessmove.com/dev-builds

Thanks very interesting. Actually over the given time frame the progress (in selfplay) still looks linear.
May be with an exception between 10/15 and 03/16, here the progress seems to have been a bit higher.

Guenther

We can know if we are now at a local maximum only if we come back to this graph after some month since now. Of course, if we are at the bottom of the hill and we can't see forward, we cannot know that we cannot raise again. Mine is just an IMHO and a bid.

[lantonov]

It is normal to expect a slowdown in progress but it is not a fact yet, at least within the past 2-3 months.

[stegemma]

(I mean "top" of the hill, not "bottom")

[Rodolfo Leoni]

It is normal to expect a slowdown in progress but it is not a fact yet, at least within the past 2-3 months.

You honor me with your visit in the Fish cube.

Thanks for the link. I'm happy to see it's improving quickly. I didn't believe it was possible.