Author: Larry Griffiths
Date: 10:54:10 08/11/00
Go up one level in this thread
On August 11, 2000 at 05:57:22, leonid wrote: >On August 11, 2000 at 00:02:39, Larry Griffiths wrote: > >>Leonid, >> >>I am wondering now if I am confusing BruteForce with Minmax. I always thought >>BruteForce was generating ALL the moves (or another name for MinMax). > >It could be that I am confused with name. It happen to me often. Never mind. >Will just say what I see as Brute Force. Brute Force look all the possibility in >given position before it reach its decision. It is not forced to see all the >moves in each ply for doing so. In MiniMax you see all moves (nodes) for each >ply before saying your result. > > >>Anyway, I ran your position to 6 plys and here are the results... > >Some data still is useful for me even if I am doubious about my Brute Force >branching factor. > >> >>TCBoard - RunMinMax() >> >>Elapsed Time= 1125.38 seconds >> >>Ply ---CAPTURES-- -----MOVES----- -----TOTAL----- ---Invalid--- >> 1 5 30 35 2 >> 2 130 1,065 1,195 93 >> 3 6,613 35,468 42,081 2,107 >> 4 199,119 1,282,187 1,481,306 107,805 >> 5 8,725,158 43,530,271 52,255,429 2,555,427 >> 6 281,688,872 1,585,505,759 1,867,194,631 132,999,965 >> _____________ _______________ _______________ _____________ >> 290,619,897 1,630,354,780 1,920,974,677 135,665,399 >> >> Captures/Moves Per Second= 1,706,964 > > >What is interesting for me in your table is the "INVALID" number. Probably you >generate invalid move in move generator and only lately you see if it is legal. > This is indeed true. The postings here lean toward making move generation lean and mean and doing checking on the back end in hopes the alpha-beta cutoffs will eliminate the extra cpu overhead in validating moves. > >Must I understand Capture/Moves as simply "illegal moves generated by second"? >And is this number 1 707 964? And on what computer? > Captures/Moves is "LEGAL" moves/unmoves made per second. It is Total legal captures + Total legal moves / Elapsed time. The invalid moves made/unmade are not included. I ran this on a Pentium III Xeon 550Mhz machine. > >>The branching factor looks like it is between 35 and 36 using MinMax. >>This is what I have found to be documented for the average number of moves >>for a side. > >For this I have special fonction in my program. It do usual search by Brute >Force but only generate for each ply all its legal moves. This permit constant >saying of average moves for given position. Have this for 8 and 10 plys deep >search. > >Thanks for response, >Leonid. > >>Larry.
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.