I assume that in theory it would be possible for a player to reach the point where he could never lose any game. That no matter how good his opponent, it would just result in a stalemate. In theory. My question is how close the top players in the world are to that stage.
Meaning, if a theoretical “perfect” or “ultimate” player (or computer) played these guys, to what extent would they dominate them? Win every game? Win 50% and draw the rest? I assume the theoretical “perfect” player would never lose, so the only question is what percentage they would win.
This might be a stupid question, me not being a big chess guy. But FWIW, that’s the question.
I dunno. Humans have off days. It can’t be helped. Historically, it seems that no matter how good you are someone always comes along later on that can best you. Even if only a single time. This is not limited to chess, either.
I wonder if the current crop of grandmasters would dominate the players from a century ago. After all, hasn’t the game developed quite a bit the last century?
A 100 years ago an 8 year old prodigy defeated a whole room filled with masters.
It seems likely that white has an advantage, so it’s possible that a perfect player could always win when playing white and always draw when playing black (against an imperfect player). But, if white ends up theoretically not having an advantage, then two perfect players would always play to a draw (which seems to be practically how it’s going lately, right?).
So, depending on the solution to chess, a top chess player could either always draw or always win on white and draw on black.
In even much simpler games than chess, a player with optimal strategy will dominate a player with anything less than optimal strategy to a significant degree.
Checkers, for example, is one of the more complicated games that has been solved, and it’s significantly less complicated than Chess. Two optimal checkers players will draw, but one optimal checkers player will likely almost always beat a sub-optimal but very good checkers player, because the sub-optimal player likely needs to make only one sub-optimal play to doom themselves to a loss. That claim is not necessarily true for every game, but I believe it to be very likely true for chess.
All the 7-piece endgames of chess are solved, and enough of them require play that is so far beyond the capabilities of a human (hundreds of moves to reach a win) that it’s quite likely that an optimal-against-a-human strategy could involve maneuvering into such a position; one that a human is simply incapable of playing out of.
So, it’s unknown whether when/if chess is solved it will always be a draw or white will always be the victor (or, I guess possible but extremely unlikely, black will always be the victor) when two optimal players face each other. But I’m pretty confident that an optimal chess player would literally always beat every human player with either black or white. Our brains just can’t handle the complexity.
For comparison, when AlphaChess made its debut against the (then) strongest computer in the world, it went 30 wins, 70 draws, 0 losses out of 100 games. And that’s against a computer that it was already basically impossible for any human to beat.
The fact that draws are so common at the high end (among humans and computers alike) certainly suggests that black can always force a draw with optimal play, but the question still isn’t completely resolved (and might never be).
Well, the top engine has a rating of about 3500. Let’s give the too human grandmaster a rounded off 2900. That’s a huge difference. Plugging it into ELO win percentage calculators, that translates to an overall 96.5% win rate for the engine, and a 3.5% draw rate for the grandmaster, and an impossibly small win rate for the grandmaster. And the engines aren’t perfect yet.
So I’d say 95%+ outright win rate for the “perfect” chess player against the best grandmaster right now. Feel free others to check my logic.
Also need to factor in tournament play and the influence of the clock.
At one stage I played a little our correspondence chess with a mate who routinely kicked my arse “over the board”. Don’t know his ranking and I didn’t have one. Possibly because he thought he had my number he played quickly. I took much longer and regularly found stronger plays. I would have lost 95% over the board but roughly broke even in correspondence mode.
I watched a video of Hikaru Nakamura, who is currently #11 in the world and was #2 at his peak, play an engine with the goal of not being “adopted.” I’m not sure if that term is a streamer chess term, but it’s winning 10 out of 10. Hikaru lost all 10 matches and it wasn’t even the top engine with a lot of computer.
Current top computer engines will win nearly 100% of games against any human players if not 100%. Unless humans get waaay better for some reason, this number should go to 100% in the near term future.