Similar to how GTO won't make an exploitable play if an unexploitable one is available. But if so,does that not mean that there are situations that come up where the human GM will win more often than the computer because the weaker opponent is not apt to see the forced checkmate, and if he doesn't, the advantage will swing back to him.

Nope. Computers are waaaay too far ahead of the best human players that anything humans do doesn't help except for veeeeeeery specific positions. But a human can't even get to that kind of position on his own, he'd need the help of the top computer to do so.

We're well past the point of human + computer vs. computer being interesting at all.

There are maybe some settings on some engines that have to do with what you're talking about, but, I think computers more or less just play whatever the "best move" is according to how they make decisions and time constraints they're playing under. That can lead to some weird stuff with some of the pure neural networks - I have no idea if it was resolved but Leela used to just keep the game going after it was up a bunch of material - rather than get to checkmate faster (the actual checkmate was further than it was calculating and all lines showed 100% win chance so it just was making somewhat "random" moves until it won).

The human GM would certainly not win more often by using "exploitation" but he could win faster this way (which would matter in time handicap games), at least faster than if he was trying to play "gto" chess.

Yes, hi level chess programs will see and react to threats that GMs won't. While the human GM might win faster because of this, he won't actually win any more games. The same phenomenon is seen with say GMs in large open tournaments against soft fields, vs expert players in those same fields. The expert players will make some unsound attacks that their weak opponents will not be able to counter and win quickly. It will take the GMs much longer to win, because they will play very prophylactically and achieve no risk subtle advantages, but they will win all their games (and then the tournament) whereas eventually the experts will make an unsound attack against a player who knows how to defend and they will lose.

The best play chess programs play at 3500 elo. Even the best human player probably couldn't draw stockfish 11.

Computers might be better than humans and they beat our asses everytime, yet, human chess popularity is still super high, especially now given the mini-series Queens Gambit. No one gives a rat's ass about computer chess. So, ultimately, humans won

Wat? I'm not sure this really is true or makes sense. I would be pretty surprised if GMs take longer on average to beat a similar level of opposition than an expert.

Well it's not really my claim, but that of GM Finegold of the Atlanta Chess Club, a popular chess twitch streamer and youtuber. In one of his videos he was talking about how if you are playing in an elite tournament like the Sinquefield Cup then if you had a score of like 4 wins and 3 draws you would probably win, but if you are playing an open tournament with a huge field and you want to win then you need to win every single game. And that it will take him a long time to win games because he will play very conservatively, constantly building up small advantages, whereas lesser rated players than him will just charge ahead and attack and, and usually they will beat their opponents much quicker then him, but sometimes they will lose, whereas he wins every game, because he doesn't take any risks and always makes the best move.


Have you ever seen Anatoly Karpov play? His style is the epitome of patience and careful planning. The Boa Constrictor they call him. He always plays the most boring moves possible, a master of closed positions, he just subtly rearranges his pieces, gains tiny advantages, never takes any risks, never gives you anything, and then somehow your pieces end up on your back rank and his are greatly positioned and you lose the game. And this is vs other super GMs, people like Kramnick.

I mean, even if that's true, Finegold is only saying it is for himself and he (as far as I remember) constantly talks about how his style is old man chess.

Yes, that was Karpov's style. But that doesn't mean he wouldn't blow 2200 players off the board in short order.

This is really nonsense. "Play the most boring moves possible"? The man´s oeuvre is chokefull of brilliant games. And it´s not Kramnick.

https://youtu.be/qlX_806ZwOk?t=2663

here is the Finegold clip

I don't dispute that. Karpov was the best or second best player in the world for a very long time. But like Petrosian he played a very cautious, defensive style, focused on the accumulation of small advantages; winning with no risk. That's why he was the best in the world.

Haven´t listened to the whole Finegold clip, but I don´t hear him say Karpov only plays the most boring, careful moves. During every stage of his carreer Karpov has played brilliant attacking games, Gik (1968), Hort (1971, a true gem), Korchnoi´s dragon slaying in the the 1974 candidates final, all the way up to his massacring of Topalov in Linares 1994. His natural inkling was probably to control counterplay ahead of initiating active play himself, but that doesnt mean he didn´t have a very evil eye for positional attacking play.

Actually that clip is just him talking about how he takes a lot longer to beat a 1400 rated player than a 2200 rated player would.

How? Why?

Finegolds a moron

In the same way when you give Stockfish a won position with loads of enemy pieces en prise but no clear path to mate. Stockfish will carry on playing strategically looking 25 moves ahead whereas the average player would just grab the pieces immediately without much thought and win faster.

Taking pieces immediately enter inferior lines of the search tree as far as the computer is concerned but lead to a quicker more straightforward win after its 25 move horizon.

Sometimes you can show the computer the moves then back up to the original position and it will adjust the principal lines.