Internal Iterative Deepening questions

Discussion of chess software programming and technical issues.

Moderators: Harvey Williamson, bob, hgm

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
User avatar
hgm
Posts: 25056
Joined: Fri Mar 10, 2006 9:06 am
Location: Amsterdam
Full name: H G Muller
Contact:

Re: Internal Iterative Deepening questions

Post by hgm » Tue Sep 27, 2011 4:27 pm

But why do you think it is worth having a best move when you are likely to fail low? Searching with the enlarged window is very expensive, even at the reduced depth. It will give you the best move, but if that best move will also fail low at full depth, it will have saved you exactly zero that you searched it first.

My guess is that simply searching without hash move (starting with the MVV/LVA best capture) would be much more efficient.

Saving a hash move on a fail low will not help very much. It could still be a very poor move. So it is not so clear whether on average the move with the largest upper bound will be better than the best MVV/LVA capture. So it could be better to try the latter (perhaps make it hash move) unless you have evidence to the contrary (e.g. because it had an upper bound much below current eval).

sedicla
Posts: 149
Joined: Fri Jan 07, 2011 11:51 pm
Location: USA

Re: Internal Iterative Deepening questions

Post by sedicla » Tue Sep 27, 2011 4:43 pm

I understand, seems the research is not necessary anyways.
I'm using the mvv/lva in the sort.

User avatar
hgm
Posts: 25056
Joined: Fri Mar 10, 2006 9:06 am
Location: Amsterdam
Full name: H G Muller
Contact:

Re: Internal Iterative Deepening questions

Post by hgm » Tue Sep 27, 2011 5:04 pm

Note that the IID pilot search, even if it does not provide you with a hash move for the current node because that fails low, will still have stuffed the hash table with cut moves for the following levels of the tree where the opponent has the move. So even if it inadvertantly fails low inwhat you thought would be a PV node, it will still speed up the following full-depth search.

Post Reply