Intended for use in conjunction with move ordering.

Since strong moves are searched first, it stands to reason that later moves in the list are not as good.

We therefore reduce the search by one ply when searching later moves, reducing the amount of effort spent on moves we expect to be bad.

The starting point of LMR varies between programs, but generally the first four or five moves are searched fully and any moves after this are reduced.

LMR will be counterproductive in engines with poor move ordering, ruining the tactical capabilities of the search.

Late Move Reduction