expected elo gain from depth dependent null move reduction ?

[MahmoudUthman]

My search is a PVS with transposition table "not used in QS ,null move pruning & check extension " , and my evaluation considers material , advanced mobility , PST , advanced pawn evaluation "+PawnHashTable", and king safety , most likely it would rank between 2450 to 2550+ on the CCRL 40/4 list .

*the null reduction I use is :R=2 +1 if depth >=8 "the search doesn't descend directly into QS from a null move", how much elo gain should I expect from making the reduction depth dependent , I know it depends on many factors but I don't have the facility to test with low error margins if the gain isn't much , that's why I'm asking ?

[cdani]

An optimized null reduction can win 10ths of elo, even more than 50, but it depends on a lot of things.

[AndrewGrant]

I would expect the gain to be large enough to prove with less than 10,000 games. Just make sure your time control is long enough to trigger the differences in your search routine.

[MahmoudUthman]

should I start optimizing it -(making it depth dependent and testing different formulas )- before or after implementing other common pruning/reduction techniques (futility , LMR , razoring ,...etc) ?

[cdani]

You can try, but as one feature tend to be affected by the others, if you had optimized this one then you will have to optimize it again.

I tend to do just a few tries until the win is more or less in the expectations, and then I go to another feature.

And from time to time, once the engine is clearly stronger, I retest some like the null move.