How many plies do chess programs search practically forward?

[Uri Blass]

Here is my idea to test it.

Give your chess program many chess games to analyze every position with clear hash for 1,000,000 nodes(let say with 1 cpu and 1024 mbytes hash to have deterministic results).

memorize the maximal ply that the program get during the search and give a distribution of that number.

You can repeat it also with different number than 1,000,000 nodes.

Note that in order to do it in stockfish you need to change the code because the selective depth that stockfish reports is not the maximal depth that it practically get but only a lower bound.

I think that it may be interesting to see distribution for different programs.

Something like:
100,000 positions
128 plies 35 positions
127 plies 56 positions
126 plies 102 positions
....