Author: Uri Blass
Date: 02:18:15 10/27/05
Go up one level in this thread
On October 26, 2005 at 04:40:16, Fabien Letouzey wrote:
>On October 26, 2005 at 01:47:34, Uri Blass wrote:
>
>>I thought from the name that it is about move ordering but it is obvious from
>>the content that it does nothing about move ordering(maybe it returns some value
>>that is relevant for move ordering but it certainly does not help to exlain move
>>ordering.
>
>>int move_order(int move) {
>>
>> ASSERT(move_is_ok(move));
>>
>> return ((move & 07777) << 2) | ((move >> 12) & 3);
>>}
>
>This is a static move ordering for debugging, e.g. a1a2, a1a3, etc ...
>
>---
>
>Real move-ordering code is in sort.cpp, and the implementation is admitedly
>complicated.
>
>Move ordering is:
>- transposition-table move
>- "SEE >= 0" captures (+ promotions) in MVV/LVA order (i.e. winning and equal
>captures mixed)
>- 2 killers
>- quiet moves ordered by history scores
>- losing captures in MVV/LVA order
>
>To my knowledge this is identical to the order proposed in Heinz et al.'s book
>(except that they don't specify interaction between SEE and MVV/LVA), although I
>reached it by experimentation.
>
>Fabien.
Thanks.
I am not sure if it is identical because the question is if the definition of
SEE is the same and hostory scores are computed in the same way.
I doubt if it is a good idea to order losing captures last because there are
often cases when they are winning captures.
simple example that I composed.
[D]r4qk1/5bp1/8/5p2/6p1/8/5PRP/R2Q2KB w - - 0 1
Fruit needs depth 3 to see Rxa8 Qxa8 Rxg4 because it probably considers Nxf4 as
a losing capture in the qsearch.
Note that Movei can see it at depth 1 but it is not because of better see but
because of less pruning in the qsearch that I am not sure if it is productive.
Movei prune losing captures but not in the first plies of the qsearch and here
it cannot see the list of captures at depth 1.
[D]r3rqk1/5bp1/8/5p2/6p1/8/R4PRP/R2Q2KB w - - 0 1
Uri
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.