The problem with LMR in suprtactical positions.

[metax]

So Rcc3 is a killer?

[bob]

So Rcc3 is a killer?

Or a hash move. NO way of knowing which without dumping the tree to find the ordering info, which is a pain. But it clearly wasn't reduced. Otherwise the knight capture would be the first q-search move and the next move, being a check, would not be in the search at all since I do qchecks on the first ply only, at present.

[Uri Blass]

So Rcc3 is a killer?

Or a hash move. NO way of knowing which without dumping the tree to find the ordering info, which is a pain. But it clearly wasn't reduced. Otherwise the knight capture would be the first q-search move and the next move, being a check, would not be in the search at all since I do qchecks on the first ply only, at present.

I do not understand why it is not possible that Rcc3 is reduced.

It is possible to have the following:

Rcc3 remaining depth after Rcc3 is 1 and not 2(reduced)
Nxe5 remaining depth after this move is 0(next move is the qsearch)
Rh3 mate(first ply of the qsearch)

Uri

[bob]

So Rcc3 is a killer?

Or a hash move. NO way of knowing which without dumping the tree to find the ordering info, which is a pain. But it clearly wasn't reduced. Otherwise the knight capture would be the first q-search move and the next move, being a check, would not be in the search at all since I do qchecks on the first ply only, at present.

I do not understand why it is not possible that Rcc3 is reduced.

It is possible to have the following:

Rcc3 remaining depth after Rcc3 is 1 and not 2(reduced)
Nxe5 remaining depth after this move is 0(next move is the qsearch)
Rh3 mate(first ply of the qsearch)

Uri

That might be possible. I didn't look at the depth carefully. However, if Rh3 occurs in the first ply of q-search, it will certainly be found, so you could well be correct...