Author: leonid
Date: 14:43:16 08/04/00
Go up one level in this thread
On August 04, 2000 at 13:53:27, Tom Kerrigan wrote: >On August 04, 2000 at 07:34:08, leonid wrote: > >>On August 04, 2000 at 00:46:19, Tom Kerrigan wrote: >> >>>On August 03, 2000 at 23:12:12, leonid wrote: >>> >>>>On August 03, 2000 at 17:59:47, Tom Kerrigan wrote: >>>> >>>>>On August 03, 2000 at 17:31:04, leonid wrote: >>>>> >>>>>>On August 03, 2000 at 14:14:38, Tom Kerrigan wrote: >>>>>> >>>>>>>I have 3 search functions: >>>>>>> >>>>>>>root_search() >>>>>>>search() >>>>>>>quiesce() >>>>>>> >>>>>>>The first two are almost identical, except root_search has some extra stuff for >>>>>>>the root of the search tree. >>>> >>>>Tom, I sent my response but came here once again to be sure in the most >>>>important question. You have "some extra stuff for the root of the search tree". >>>>What is for you "root"? Ply zero, or ply that will be reached when X numbers of >>>>plys will be searched? For me, after your terminology, ply numbers (x-1) and (x) >>>>are very distinct from all others. >>> >>>I clearly explained where the root of the tree is. It's the first node that you >>>search. >>>-Tom >> >>Then really you have 100% identical ply search. Completely different from mine. >>Now I am even not sure if Bob H. do it differently. Maybe I was lost in >>terminology. > >Hyatt and many other people do it just like I do. This very good news! Now I see this part more clearly. For sure my way of talking (higher ply and lowest ply) is the reason for confusion. Have seen this few times. > >>Two plys from which search start done differently in my program but this >>difference is not crucial for the speed. Two plys from the start are different >>because in first of them you are forced to see all moves (I neglect here few > >My search function looks at all moves at every ply (ignoring null move for the >moment). This is critical if you want a good program. > >>exceptions). Zero ply and one below mainly because of tactics, even if the part >>was hardly even started. If only your way of writing is one that is very commune >>it could be the reason why in Fritz I see so high NPS numbers. Sometime they are >>just to high to even approch them and still my tactics is in embrion stage. >>Tactics still must lower my NPS in unknow proportion. >> >>Tom, what will be your estimation for the lost in time or NPS for the program >>that have its tactics compeltely written and the same that don't have them? > >I don't have any special code for tactics. I just have my alpha-beta search. So >I don't understand your question. I do expect that you have basic part of your logic that find better move through material echange. And other part that care about "horizon effect", pawns structure and so like. If my understanding is right (after what I could read here), this second part (beyond material echange) is tactical part of program. So my question was: How mach time is spent in tactical part compared with material echange? This could give me some idea how much my speed, or NPS, will go down when tactical part will be accomplished. Found why my last "stange position" was done differently by Rebel. I put black king on the wrong place. Once king on its place, Rebel responed like my program and Genius 4. Leonid. >-Tom
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.