This is typical of engines that are set to play at lower ratings. They set them to select moves that aren’t the best move at some rate but they do it erratically in weird spots that look nothing like human play. It’s hard to tell sometimes whether they’re blundering or you’re falling victim to some disgusting engine line that there’s no way a 1600 would find.
Just read the rest of the thread and others making similar points. Someone should just get a database of games at various strengths and train a neural net AI to imitate the play.
Update: I’ve beaten the 1600 and 1800 versions of Beth Harmon after a few tries. Now I’m up to the 2400, and I really don’t see myself beating that one so I’m calling it quits. Still losing to all kinds of 1100 rated real players
Yeah, what’s the deal with Stockfish’s ratings on Lichess? It says level 4 is 1700, but I can beat the pants off it at level 4 and I’m sure I’m not anywhere near 1700.
I’ve been playing some arena tournaments lately - they’re fun and a good way to play opponents with a wide variety of ratings. If you win your first game or two there’s a good chance you end up paired with some very strong players. I’ve had two games where I’ve had winning positions against 2200+ players in the middlegame but I’m definitely not good enough to convert them yet. Got my highest rated win last night against a 1950 player on lichess, though!
So this is cool, a new chess AI called Maia was trained from human games in an attempt to play a human-like style. Different levels to mimic human players of different ratings.
Another example of computer chess doing things almost no human player would do. Am I the asshole for not resigning, or is Stockfish the asshole for dicking around for three moves before playing the super-obvious checkmate?
Just re the chess mastery vs intelligence debate a while back, Nakamura said in a YouTube video today taken from stream that his IQ tested out at 102. Who knows how accurate that was and definitely no aids debates about the usefulness of IQ as a measure of intelligence, but just a data point for the incongruence of chess aptitude vs general intelligence.
What Nakamura does have is a strikingly good memory, not just for chess positions but for stuff like “oh yeah I played this line vs Shirov in 2012 I think it was”. Similarly, Carlsen was #1 in the world for a little while in Fantasy Premier League and reports at the time said that a big advantage of his was his prodigious memory, being able to remember everything about every player in the league.
I think getting into competitive chess will dramatically improve anyone’s memory. Its a very intense training regimen for studying many commonly used openings and the resultant flavors, PLUS even in the more creative mid to late game a ton of the skill is based on honed pattern recognition.
