Here is another classic case in point. Here Rybka 4 960 after a long think, has concluded that the position is almost completely equal:
Black to move:
Could someone please explain how Rybka 4 960 comes to this conclusion considering it is up a minor piece but down a rook? You might think that white is about to launch an attack after e4/Bb5 but it is black's turn to move and he has ...c5/Nc7 (protecting b5) or ...c5/Qc7 (removing the pin and shielding).
Cheers
This forum is for showing to you why it is that the current latest generation of computers struggle to play Chess960, and thus why it is that human intelligence predominates. We all know that the 960 computers do not have a human generated database to fall back to and so must struggle to understand an opening pattern that they cannot see because their event horizon is not great enough. Feel free to post examples or to comment.
Here is a simple and classic example. This is the very latest and greatest 960 engine trying to decide whether Qxh2 is ok. Please be aware that the computer did not play the opening in this case. This is just an example of a computer analysing a position and not even understanding a medium level tactic! This is Rybka 4 960 running on a dual core CPU and still not getting it even after ten minutes! Rybka thinks that white will only get a small advantage if Qxh2, and so it is ok to play:
"Hmmmm 10 minutes later and I think I am not too bad if I take h2 because after c5 I have Nf7 and I should be ok. After c5 I could go to Nc8 but that is probably not as good"
!?!?
Alright I hear you that in the next passage of play if white takes the knight fxg5...fxg5+ we have Kg1 for example. The position is very comical. It's really hard for the silicon believe me
. White's king protection is very weird granted. Chess960 is amazing. Seriously though, I think the computer has a lot of trouble evaluating a position in chess960 mid games when pieces are so terribly trapped as black's pieces are here. This kind of entrapment is nowhere near as common in SP518 (standard chess).
In the computers defense there was no alternative but Qxh2 and the computer did not play the opening and on a machine with more cores Rybka would probably understand that ...Nf7 is not a good idea. However I doubt that the computer would ever understand that white has a big advantage after Qxh2.
The point is that the computer is not as all powerful as we would believe it to be. Without a safety net of human knowledge to fall back on in the Chess960 opening, it is all just too deep for the computer. If the computer does not quickly understand this medium difficulty situation, how is it to truly understand the 960 opening strategy without human guidance? A genius player like Nakamura or Carlsen would regularly beat the best chess 960 engines of this day because they would be able to achieve winning midgames where the probability of blundering is not great. That is only my opinion.
Comments?
Cheers