lichess.org
Donate

Is Stockfish AI level 8 rating 2200 or 2500 read description

I want to start discussion among players here rated 1900+ ,how do you think AI level 8 is closer to 2200 or 2500 elo,I can't assume that on my own becouse i am still too weak chess player so I ask stronger players for discussion thank you.
Closer to 2500. It handles tactics like 3000. In openings, endgames and closed positions it is weaker, but with opening book and endgame table bases on fast hardware more like 3000.
dependent on what you have on your computer (( ram - cpu, etc )
storckfich 6 is aboud 3318 ELO.My guess that strockfich 8, that you can find at this site, is around that rating
It's extremely difficult to compare human ratings with computer ratings.

Engines use brute force where as humans use technique. It doesn't miss very many tactics, but is lacking in the strategy, and innovation of a strong GM.

Comparing engines to human GMs used to be like comparing a ninja to a fat tub of blob that could easily be refuted chopped up easily...

Nowadays it's more like those same ninjas have to face a 600 pound gorilla that has chainmail protecting it from the blade, also the gorilla gets a sword too, and knows how to use it.

In the future lets say 50 years from now... it will be more like those same ninjas gotta fight against a godzilla.

If quantum computing comes around hundreds of years form now, and they manage to some how create a 32 piece table base playing against such an opponent would be like trying to defeat god.

Engines are growing in strength much faster than humans.

I think as a player rated about 2000 I'd have a better chance against a 2500 player slipping up, and giving me a tactical gift than a "2500" engine, but both are slim chances.
A strength number isn't difficult to use to compare computers and humans, but it requires that the two play a lot of games together. Whether you use elo or glicko both are good enough to eventually figure out the right rating.

On lichess a lot of players play vs. Stockfish but we aren't using that data to adjust the ratings forr some reason. I know on other sites (at least in the past) some bots were allowed to be rated. Maybe ICC or Chesstempo.

That being said, computers are pretty strange and lopsided players by human standards. They tend to have godly tactics as a basis, but their mistakes are often strange and not at all like the mistakes humans might make.

I think saying they are lacking strategy and innovation of a strong GM is kind of selling them short and also giving too much credit to humans. Humans are basically doing their own kind of pruning and weighted searches of their own, with some brute force checking. Humans and computers both just have different weaknesses. In the end the rating number doesn't tell the whole story, only the story about the frequency of (probability of) wins, losses, and draws. If a computer program is 2600, don't think that means it knows everything a 2600 human does. It is far worse than that, but makes up for it with immunity to human factors like depth horizon, psychology, working memory, and human error. But that doesn't mean a computer can't ultimately have every chess ability of a human. It just means they will always have whatever human abilities they do, plus all these benefits.

This topic has been archived and can no longer be replied to.