Oct. 19th, 2013

The bugs mentioned in my last month's post have all been fixed in the GitHub repository, except for this one:

In pulling moves from the hash table (or the q-search cache) for use in the quiescence search, the code could select moves that do not fit the conditions that would apply if no cached move is present. For example it is possible that a non-capture, non-promotion checking move could be retrieved and used at a depth at which checks in the q-search are not normally allowed.

I did some measurements and this is not frequent enough to worry about.