another tablebase idea

[Uri Blass]

1)It is not good to probe tablebases at every node because it slows the search too much.

2)probing the tablebases only when the depth is big may cause problems because of non consintent evaluation(a draw position may be evaluated as a draw if you probe tablebases and as +2 for white if you do not probe them).

I thought that it is possible to solve the problem by the following idea.

1)Have a pruning rule(if a capture leads to tablebase position then prune the tree before that capture unless the remaining depth is at least 10)

I guess this rule is going to have some small negative effect about the elo
of the program.

2)probe tablebases at every node of the search(and of course remember your probe in special hash tables that you never forget so you do not probe the same position twice)

Not many nodes of the search are going to be tablebases positions because of rule number 1.

I wonder how much elo you are going to lose by rule 1 and if you earn more than you lose by rule 2.