4

I saw this rating score chart recently:

Elo rating scores

Is it true that cyborg chess players are (that much) better than the best chess programs by themselves?

6
  • 3
    Probably not. Stockfish 15 is already roughly 3700 Elo (it's hard to measure that strong) so the chart is inaccurate.
    – qwr
    Aug 28, 2022 at 21:09
  • 2
    most chess engine ratings are only against other engines. I asked a separate question on stockfish's human calibrated rating. related: chess.stackexchange.com/questions/18893/…
    – qwr
    Aug 28, 2022 at 21:23
  • Well, first of all, I don't know any "cyborg" chess players, so I'm going to have to do some quick research and try to explain my thinking. Google gives a lot of examples with Deep Blue, so I'm going to assume that a cyborg chess player is a human who plays with an engine. (Lichess also mentions that cyborgs are allowed to play casual games against each other, and defines them as people with engines) Computers are vastly stronger than people, but I know an IM who frequently plays in correspondence ICC tournaments where you can use engines. I suppose that the fact that those tournaments exist w
    – cirrin
    Aug 28, 2022 at 21:38
  • Well, I guess the most frequent cyborg player are correspondence ones. In a lot of correspondence competitions engines are allowed.
    – emdio
    Aug 29, 2022 at 6:08
  • 2
    Generally this has always been the case in correspondence chess. A talented human "guiding" an engine will be a bit better than just the engine by itself. Recently though (since like maybe 2019), this gap has decreased, due to how strong top engines like Stockfish and Leela have gotten. Referring to correspondence databases isn't as necessary for say, people researching opening theory; it's basically enough to just turn on current Stockfish to directly evaluate a position. Aug 29, 2022 at 17:05

1 Answer 1

6

Yes, it is true that "cyborgs" are better than the best chess engines. That's why the best correspondence chess players still beat engines. (Sort of. By far the most common result of such games from the opening position is a draw.)

See source (part 2 can be found from the same link), which is an interview with a former correspondence chess champion.

Today chess engines are much stronger than the best humans and many people wonder about the role of humans in correspondence chess. What can you do that the engines cannot do?

It is indeed impossible to achieve any significant result in today’s correspondence chess without engines and databases. But we humans play, not the engines, and the input of humans mainly affects two areas: a) the choice of a suitable opening, and b) steering the engine toward (or away) from certain types of position.

If you want to be successful in top correspondence chess you can only play a certain set of openings because you simply cannot afford one single sub-optimal move – if you do, you will sooner or later regret it. That’s as certain as death and taxes.

How well you guide your engines depends on your general chess knowledge. The better your chess knowledge (the significance of pawn structures, good bishop, bad bishop, etc.) the better you will do here – today’s engines are very strong but they still misjudge positions. If you have enough time and patience and composure you can feed the computer with more good ideas than your opponent – exactly the process described by former World CC Champion GM Ron Langeveld in an interview on the ICCF website.

2
  • 4
    Is this still true with the advent of neural network engines (2018) that have a much better positional understanding than their predecessors?
    – qwr
    Aug 29, 2022 at 18:48
  • 1
    @qwr yes, but as engines keep getting closer and closer to perfect play, the margin keeps getting smaller, and more and more corespondence games are drawn.
    – Allure
    Aug 30, 2022 at 0:15

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.