Are we missing too much from the 50-move rule implementation to engines, whether HCE (Hand-crafted evaluation) or/and NN (Neural Network)?

We know that certain endgames are a win without the rule. Would the NN training time be significantly extended by removing the rule? If so, would there be enough benefit for it, i.e. engines evaluation more optimistic in some situations and converting drawn positions (by the 50 move rule) to a win?

Of course we are talking about special and rare cases over all.

  • From my POV this question is less related to chess and more to AI. Apr 28, 2022 at 6:48
  • *=50-move rule is lengthy and redundant. Related to computer chess as 50-move rule is a chess rule.
    – JSBach49
    Apr 28, 2022 at 8:05
  • 1
    Definitely a pure chess question and a good one. op may like to check out r-mobility (a tiebreaker score used in engine chess) Apr 30, 2022 at 15:55
  • 3
    This is a confusingly worded question. It would help if you explain a bit what the context is, your assumptions, etc, and spell out the acronyms. NN? HCE? May 2, 2022 at 5:34
  • Missing to much, i.e. Some positions are under evaluate because engine would need far more than 50 move to convert it into a win. The NN training is the self play reinforcement learning part, without the 50 move rule some games would become very long and would probably increase the training time dramaticaly.
    – JSBach49
    May 2, 2022 at 7:19


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.