All engines I'm aware of focus on finding the best moves to win the position. I'm curious whether there are any that exist that attempt to present an honest evaluation of a position, with human capabilities in mind. For example, with each candidate move, associate a weighted probability that indicates the likelihood a human is to actually find a given move on the board.

I'm thinking of situations where computers can find winning lines, but the key moves are so counterintuitive or unnatural that most humans would never consider them.

This is clearly a challenging computer science task. I'm also not even sure there's a good use case for such an engine. But, I'm curious.

  • A good use case would be doing live commentary - doing commentary using stockfish or sesse is good if you want to compare people to computers, but if you're a human talking to other humans about what two humans are doing, you'll want an engine that thinks like a human. This is why such commentators tend to be high level players themselves - they are those engines that think like humans. – corsiKa Dec 1 '18 at 2:38
  • Related: chess.stackexchange.com/questions/13021/… – D M Dec 1 '18 at 2:45

No engine do what you describe because it’s too hard to code a imperfect human brain. Estimating positional complexity at humans level is also quite hard.

Your Answer

By clicking "Post Your Answer", you acknowledge that you have read our updated terms of service, privacy policy and cookie policy, and that your continued use of the website is subject to these policies.

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