Mobility information used in search

[xr_a_y]

Hi all,

a while ago I posted this (https://talkchess.com/forum3/viewtopic.php?f=7&t=71325) and that (https://talkchess.com/forum3/viewtopic.php?f=7&t=72981) about using some evaluation features data in search to tune pruning margin and reductions. This is breaking some kind of "evaluation / search orthogonality" of course which can make tuning a bit harder.
But I most notably used that to make SSE margins depend on danger level.

Since NNUE, as i'm not extracting all "features" from (NNUE) search anymore, I stopped to work on this thing for a while until the use of "threats" was proved interesting in some engines (pruning, probcut triggering, some margins adaptation, ...) after its introduction in Koivisto. At this point I (re)introduce again some danger dependency and added also some threats dependent stuff.

For some days, I'm now testing the use of mobility data a little more again. Because, here also, mobility was (before NNUE) used in Minic's search to try to tune some reductions and margins and to do some nullmove verification and it seems like there is some good stuff here.

Anyone has already tried something in this direction ? What were your findings / ideas ?