Author: leonid
Date: 14:29:51 02/09/00
Go up one level in this thread
On February 09, 2000 at 16:41:37, Heiner Marxen wrote: >On February 09, 2000 at 16:09:41, leonid wrote: > >>On February 09, 2000 at 15:21:34, Heiner Marxen wrote: >> >>>On February 09, 2000 at 14:03:15, leonid wrote: >>> >>>>On February 09, 2000 at 13:39:38, José de Jesús García Ruvalcaba wrote: >>>> >>>>>On February 09, 2000 at 08:44:53, leonid wrote: >>>>> >>>>>>On February 08, 2000 at 20:11:50, Dann Corbit wrote: >>>>>> >>>>>>>2bnN1k1/6P1/4p3/1P2N1P1/3K4/8/8/8 w - - >>>>>> >>>>>>Move G5 - G6. Take 0.055 sec. >>>>>> >>>>>>AMD 400Mhz. Solved by Mate Solving Logic. >>>>>> >>>>>>Leonid. >>>>> >>>>>Hi Leonid! >>>>> Have you compared the solving time of your Mate Solving Logic against solving >>>>>times of Chest? >>>>>Thanks in advance, >>>>>José. >>> >>>Chest uses 0.15 seconds on a P/133. This is quite comparable to leonid´s >>>time. >>> >>>>Hello, Jose! >>>> >>>>No. Will be interesting to try one day. This logic was created long time ago and >>>>I practically never touched it after its creation. Expect come back one day and >>>>speed it even more. Now I see that this is possible after the ideas that came to >>>>me while writing my Positional Logic. For now, Mate Solving Logic, don't use any >>>>hash tables, reference tables or any memory demanding technics. >>> >>>The hash table speeds up Chest for about 14% (conservative estimate). >>>Within a mate in 4 there is not much chance for repeated positions. >>> >>>Heiner >>> >>> >>>>When I tryed this logic for solving the mate some 4 years ago it was the >>>>speediest against all other best games that I could find. Impecable as well. >>>>Result was found after long statistics counted in thousands of positions. >>> >>>A comparison with Chest might be worthwhile, but I would use larger (deeper) >>>jobs. The deeper, the more opportunities to do tricky speed ups. >>>We may both learn from the other program, where it is significantly faster. >>> >>>>Hasta la vista! >>>>Leonid. >>> >>>Have a nice day! >>>Heiner >> >>Thanks, Heiner, for comments! Quit interesting. >> >>I found through the Rebel that game speed could be improved up to 70% when hash >>tables are ignited. Could only guess how much in real life "mate solving logic" >>could be speeded up. > >That depends very much upon the depth and the board contents. >The deeper, the better works the hash table. And Fine #70 is the classic >example where hash is essential: two kings walking around, creating >many different ways to the same position. The speed factor >(time without hash) / (time with hash) is virtually unbounded. My very >conservative estimate sometimes is as large as 36. >Normal values are around 2 to 4, which matches the 70% you have seen. > >Really deep jobs just *need* a hash table (IMO). > >>All the best in your logic writing! >> >>Leonid. > >Thanks a lot! > >Heiner I think that my finding about 70% was mainly done while making search 8 and 10 plies deep, brute force. I tried this around June last year when I bought Rebel 10 for finding the speed of my "positional logic". I never found exactly the speed of "positional logic", but the is the next story. By the way, Rebel is the best game to find so many useful data about the speed that hash tables usage give, positions/second rate and so on. At least the rate to compare I found there for sure. Heiner, I expect to speed the "mate solving logic" (beside hash) mainly by alining the moves in the way that I did in "positional logic". Taking from the "past" move that are the most succesful and put them at the head of the line. Until now all alignement was done after the logic - checking moves goes first. Do you use the moves from (the past) the hash for alignement or not? If so, how effective it can be expected? Leonid.
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.