http://www.computerchess.org.uk/ccrl/4040/Does this mean no human can beat a Chess engine anymore?
Has anyone ever beat lichess stockfish at level 8?
level 8 has only 400ms and a single thread to make a move. It's therefore not representative of the stockfish engine.
Thinkabit beat stockfish level 8 on a number of occasions. Here is a link.
http://en.lichess.org/KXDCCWvH@thibault, so our stockfish has not its full potential unleashed, it's crippled :). What's it's ELO with the 400ms/1thread setting? In the future if lichess happens to get better servers, would you consider increasing the engine power?
@LEGENDARY_BOSS thanks for the game! thinkabit seems 500elo over the engine as it's crippled to 2400elo.
Chess engines have been stronger than humans for a few years now. The Elo rating we use for human players doesn't really work for chess engines though - partially because we don't have GMs, etc. constantly playing against them.
Ratings are always relative the players in the pool, and the absolute numbers are completely arbitrary.
You could take Stockfish, Houdini, and Komodo, all of which are much stronger than any human, and make them play a bunch of games against each other, but start them all of at 1600, and you'd end up with ratings centering around 1600.
Obviously that wouldn't mean that those engines are 1600 players in human terms :)
Most rating lists started out (years and years ago) trying to set the numbers so that the top engines had ratings roughly equivalent to the top humans, so that it would be easy for people to relate the two.
Modern engines are so much stronger than those engines of yesteryear that the best are now much, much higher than typical elite human ratings, which is why you see numbers like 3200 on those lists, and even newer lists try to set the mean rating so that the absolute numbers look similar to those older sites, because that's what people are used to seeing.
All that actually matters with rating are differences, not the absolute numbers, and even the differences only matter if you're comparing players who have derived their ratings from playing in the same group of players.
That's why it's misleading at best to compare FIDE ratings and engine ratings on lists (and perhaps deceitful, if used by marketing to show how strong a certain engine is compared to humans).
So, the completely correct answer is that we can't really say exactly how much stronger Stockfish on some particular hardware is compared to, say, Carlsen.
Now, if we want to fudge a little, there IS of course historical data on human-engine play, like the Kramnik-Fritz, Kasparov-Fritz, Kasparov-Junior, etc.
The number of games played is still so low that you can't really infer all that much from it, but modern engines are about 500 points stronger than the Junior and Fritz versions that held Kramnik and Kasparov to draws in those matches.
Also, Pocket Fritz (just the product name for a Chessbase product that actually was running Hiarcs 13) won the Mercosur Cup tournament in 2009 with a performance rating of around 2900, on a 2009 smartphone. Modern engines are around 250 points or so stronger than that Hiarcs on equal hardware.
Once you factor in the hardware improvements since 2002-2003 (for the Kasparov and Kramnik matches) and 2009 (for the Mercosur cup) in addition to the huge improvements in software, the gap between Carlsen and the most recent Stockfish on, say, a desktop i5, is probably around 300-400 points.
It might be a bit higher or lower than that, as it's not clear that the same things that make engines play better against each other necessarily have the same effect on their play against humans.
The absolutely true part of all this, though, which Magnus himself admitted, is that the best engines are quite obviously vastly superior to the best humans now.
I'd love to see a game with half an hour per move between all of the worlds computing power combined and god
I think that engines might not have really beaten humans yet, or at least haven't beaten decisively - because players like Kasparov, Kramnik etc. are probably not the best anti-computer players in the world, they earned their titles playing against humans not computers.
Chess engines are still fairly dumb in many positions, all they can do is calculate, and that's limited by geometric progression of the "chess tree" - until we see true AI chess engine humans still have a chance. It doesn't even require for you to be a master level player to be able to draw against top engines by avoiding tactical lines (and sometimes even win in endgames), if such games were rated engines would lose lots of points like that.
Please show me a game where you beat highest level stockfish.
Even Carlsen doesn't stand a chance. He might win 1 in 15 or something.
Sorry man I disagree. They don't need AI, their calculating ability is miles ahead of even the best humans.
Oh even level 8 is not the strongest.
Still I'd like to see you beat level 8 without takebacks.
