Author: Christophe Theron
Date: 23:59:35 11/30/00
Go up one level in this thread
On November 30, 2000 at 20:20:15, Bruce Moreland wrote:
>On November 30, 2000 at 14:25:32, Severi Salminen wrote:
>
>>
>>>>Well, I'm not performing any illegal captures (which leaves king in check) - so
>>>>if a piece or a pawn is pinned to (?) king it can't maybe capture and SEE
>>>>doesn't see that. I ment no other pins. But maybe king pinning situations are
>>>>_very_ rare that one doesn't have to consider them.
>>>>
>>>>Severi
>>>
>>>If you have code that lets you try it both ways, there is one way to find out.
>>>My intuition is that even if it's not particularly rare, it isn't critical to
>>>avoid a few mistakes at the tips.
>>
>>I don't have either MVV/LVA nor SEE. I just wanted to know if there are _any_
>>reasons to pick MVV/LVA instead of SEE (other than simplycity). But apparently I
>>must go for SEE, sounds like a few days of programming :(
>>
>>Severi
>
>MVV/LVA is a very quick and very dirty SEE. SEE gives you a better resolution
>of the swapping sequence. I think it would be insane to use MVV/LVA to prune
>out losing captures, because you don't detect them. You can get away with using
>a SEE to do it.
>
>I can't think of any reason not to use SEE at least to get better move ordering,
>other than that a SEE is computationally expensive. This means that whether
>it's better or not is implementation dependent.
>
>I think that most implementations would get a higher node rate with MVV/LVA, but
>they'd search more deeply with a SEE, at least if losing captures are pruned
>out.
>
>bruce
It is actually rather easy to write a SEE. Think of it as an alphabeta search
algorithm. The difference is that the move generator is degenerated: it
generates only one capture at any level (the capture by the smallest piece that
has not been already used to capture on the target square).
I have an SEE which is programmed like a recursive alphabeta search, with a
couple of optimizations. It is easy to write a simple one. Then it takes a
little more time to improve it to handle aligned pieces (a bishop attacking thru
a friendly queen for example), but with the framework of the alphabeta search
it's no big deal.
Christophe
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.