Chess is governed by tactics (unlike go), and search is the only way to deal with tactics.
The only probabilistic element in current chess engines I'm aware of are piece/square tables. They aren't used for move decisions but for move ordering to speed up search.
There are currenly some efforts to fine-tune evaluation parameters with large quantities of sample positions but so far the results aren't very promising.
Hello, this might be an old question for many of you, but I haven't found definitive answers about this. I am aware how chess engines work when games are analyzed in principle, and I've read general information about chess engines making moves in games. There is brute force, there are game records, and probably other algorhithms I am not too familiar with.
So, the question is about predictive probability. I am wondering whether there has any analysis been made of perhaps a million or more games which resulted in some sort of pattern with which probabilities that a that a human player will make a certain move in a given position.
I am not sure whether the sameple size would never be good enough, or maybe there is no application for this. I know similar methods are used by Go engines, but how would it work exactly?