On most websites, such as chess.com and lichess.org, computers aren't allowed, but when playing under ICCF rules engines are allowed. With the increase in strength of engines especially from the neural network approach, starting with AlphaZero, what do humans have to add? Are ICCF correspondence games simply one 3500 machine vs another 3500 machine nowadays? What do humans add to these high-level correspondence games?

  • There is a bit of chess intuition to be added still but mostly it is about choosing the right tools for the right job. Picking an engine that understands the current position well, deciding which line to analyze how deeply, which of the engine's suggestions look more promising than the others. So arguably it is more about understanding chess engines these days, rather than understanding chess itself.
    – koedem
    Commented Nov 27, 2021 at 22:31
  • @Herb: That's the typical "time dependent question" just (oops, time-dependent) referred in Meta :-) Having read fairly recent (son of oops) first hand reports on correspondence chess, I have the impression a bit of human intervention is still needed. This would make up for an interesting scientific experiment: does Carlsen+computer reliably fares better than chimpanzee+computer? Commented Nov 28, 2021 at 10:28
  • 1
    First note: I don't think this should be closed. The referenced question dates back 8+ years, in which time the progress in computer chess has been tremendous. Commented Nov 28, 2021 at 11:37
  • 1
    @BrianTowers since you took executive decision on this - what's the general policy regarding such questions where the (old) question/answer are now outdated? Is it to update the question and provide a new answer, within the same (old) question? Commented Nov 28, 2021 at 13:20
  • 1
    This might be the better target: chess.stackexchange.com/questions/19107/…
    – Allure
    Commented Nov 29, 2021 at 3:48