I think, SF takes time to think. The move it tells instantly may not be the best move.
So after few seconds, it tells you another move which would be the best (after SF calculates it)...
chess.com analysis contradicts itself on next move often
Sort:
Jan 23, 2021
0
#3
Depth 10 is sometimes inaccurate, even depth 18 can be inaccurate from time to time.
I've also seen this. As suggested by @Uninterrupted, it might depend on analysis time. When this happens there probably isnt a great deal of difference between the variations. My guess is the engine will rank one better than the other even if the difference is small. Often, more useful is the underlying idea expressed by the move.
Jan 23, 2021
0
#5
Yeah, you need to let the depth get past the twenties for the Eval Bar and the lines, and let the "best move, inaccuracies, etc" categories get all the way to depth eighteen. That's when the computer starts to become accurate, although as Mark said, if the difference between variations is very small, the computer will have a hard time making up it's mind without a lot more time.
Do other people see this? Why isn’t the analyzer relatively stable move to move? It makes it difficult to trust.