You do LOTS of evaluations. This would be multiple LOTS of SEE calls.

A hash table doesn't grow as you add entries. You just pick one, using some sort of hashing function, which selects a specific entry that you are going to write to. you can probably get to 100% utilization but it is very difficult. But 99+% utilization is quite common. With hashing, you are never ...

if you store some recognizable value in each entry (such as using the common "age" field) then you can look at all entries at the end of the search and look at the age value to see which entries were just written this search and which were left over from last search (and which were therefore not ...

As a general rule, native word length is the optimal reference size. The only exception would be when you are using an array or such, and you start to factor in L1 and L2 cache. If going from 64 bits down to 16 bits improves cache utilization, you would see a gain. Otherwise the cpu actually has to ...

I am not quite sure what you are asking for. "anyone else have bought it" doesn't make sense to me...

The correct way is to simply update material according to what happens. IE Pxe8=Q should add +9 for queen -1 for losing the pawn that promoted, plus whatever is captures. If the opponent then captures, he subtracts 9, which leaves the SEE score at -1 for losing the pawn.

This "corruption" nonsense REALLY gets tiring. If you like Trotsky that much, move in with him. There's no sign of corruption in TCEC, they made the correct decision and everyone has moved on but the two of you apparently.

When you add LMR that won't work so well for ordering. I did this from the middle 70's until 4-5 years ago. What I do now is a q-search on each root move to get at least a starting value for each move. I sort on this, and then leave the move list alone except when a move fails high, where I ...

Normally you don't see a lot of fail highs and fail lows in the same search. UNLESS you use mtd(f) which is where the two-bound tables originated, since you keep re-searching the same iteration with a zero width window where the window shifts around. It is quite common to see a search fail high ...

Actually there can be a problem.

Simple case: you search very deeply while pondering and store entries, most of which are >= beta or <= alpha. Your opponent plays a different move and you start getting hash hits that suggest a fail high or fail low, but when you relax the bound, you can't see ...