Author: leonid
Date: 15:46:07 08/10/00
Go up one level in this thread
On August 10, 2000 at 18:17:12, Tom Kerrigan wrote: >On August 10, 2000 at 17:42:24, leonid wrote: > >>On August 10, 2000 at 16:51:08, Tom Kerrigan wrote: >> >>>On August 10, 2000 at 15:45:06, leonid wrote: >>> >>>>On August 10, 2000 at 13:58:22, Tom Kerrigan wrote: >>>> >>>>>On August 10, 2000 at 07:54:32, leonid wrote: >>>>> >>>>>>On August 09, 2000 at 21:44:38, Tom Kerrigan wrote: >>>>>> >>>>>>>On August 09, 2000 at 17:32:48, leonid wrote: >>>>>>> >>>>>>>>On August 09, 2000 at 17:04:37, Tom Kerrigan wrote: >>>>>>>> >>>>>>>>>On August 09, 2000 at 15:59:24, leonid wrote: >>>>>>>>> >>>>>>>>>>>I don't recall Ed ever calling his search brute force. >>>>>>>>>>> >>>>>>>>>>>-Tom >>>>>>>>>> >>>>>>>>>>If it is so, now I see why my branching factor is so miserable. >>>>>>>>>> >>>>>>>>>>I asked above question when I tried to solve this position by brute force. For >>>>>>>>>>black side I looked up to 10 plys deep and it took already 12 min 17 sec. Move >>>>>>>>>>was wrong. Black knight goes to the position e2. And for finding right move I >>>>>>>>>>must go to the next 12 plys search. But this could take some next 6 hours. This >>>>>>>>>>is how my old question about branching factor came to me. It prohibit to my >>>>>>>>>>program to see very rapidly and reach far distance. >>>>>>>>> >>>>>>>>>You're the only person in the entire world who does these "brute force" >>>>>>>>>searches. >>>>>>>>> >>>>>>>>>-Tom >>>>>>>> >>>>>>>>When you want to know if your basic speed is the right one, only brute force >>>>>>>>search could say you so. This I remember from writing my program for finding >>>>>>> >>>>>>>I don't know what basic speed means, but I'm sure that there isn't a right one. >>>>>>>And a fixed-depth brute force search with no extensions and no quiescence search >>>>>>>won't tell you anything useful. >>>>>>> >>>>>>>-Tom >>>>>> >>>>>>Tom, if you compare two programs that do its search, but not by brute force, you >>>>>>actually compare "pruning technics" for both of them. But how much program with >>>>>>good pruning technics still miss from its potential, you will find by seeing its >>>>>>brute force speed only. >>>>> >>>>>What potential? Presumably the pruning techniques are increasing the program's >>>>>potential, otherwise the author wouldn't use them. >>>>> >>>>>-Tom >>>> >>>>The same pruning technics, in two programs that have different brute force base, >>>>winner will be program with best brute force speed. But by the same talken, it >>>>could be said that program with best pruning technincs could be speeded even >>>>more by speeding its brute force part. Sometime this brute force speeding will >>>>simply forgotten when program already shine with its advanced pruning >>>>capability. >>> >>>The speed of a forward-pruning program can be tested (and improved) just as >>>easily as any other program. >>> >>>My point is that your "brute force" searches are extremely stupid and there's no >>>reason for anyone to do them ever. >>> >>>-Tom >> >>Everybody can choose its own way. In the first part of my program I did exectly >>like you said, pruning technics first and brute force second. My second part I >>go opposit way. Experience say me that this way is logical. For you entire game >>start and end with one part. Here we are different. > >I made a copy of my program that does your brute force searches, i.e., no check >extension, no quiescence search, and no null move. I called it "Stupid." > >I played a match between Stupid and my program. > >Stupid lost 20 games in a row. It usually got mated around move 30. Once in a >while it would last for 50-60 moves. > >So basically, you can add 3 things that are well-understood and that everybody >has and you can immediately increase your program's strength by 400+ points, or >you can continue down your brute force path and never have a strong program. > >-Tom And now find your own program that do brute force worst that actual one. After putting everything that you took off your first program into both of them, find what is best. Now, you see my point. I not diminish importance of pruning and everything else in the program - not at all. Too early inclusion of all those final additions can obscure the fact that its initial and most important part stays weak and unaccomplished. Anyway, I am not that ready today to be in some kind of wrangling about words. Had happy day! Found two bugs. One, about bounds that was there for more that one year. Second, still must eliminate. Also some old one. Now I am taking the positions from my old "Chess life" and make solve them on mine and Rebel and Hiarcs. Sometime fatal moments come up. The most useful is for me Rebel. For good or bad reason, he is speedy. Leonid.
This page took 0.01 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.