TalkChess.com

[H.G.Muller]

I once proposed this too, except that I did not want to count depth as log of the number of opponent moves, but as your own moves. I don't think that makes much difference, and it is cheaper to implement.

The idea was that this would mainly help to solve those kind of problems that are impossible to solve for engines now (and easy for humans), where one side has overwhelmingly strong material, which is somehow trapped, and the key to beating him is preventing the material to escape. This means you have to search very deep along a tree with very low branching factor, in the face of parallel branches where the branching factor explodes. Balancing sub-tree sizes would be a big help.

It would also solve the problem encountered in real games, where you convert to a Pawn ending, in a natural way. Most programs use an ad-hoc method now (like extending 2 ply where you enter the Pawn ending) to solve that.

I never tried it out, though.