Author: Robert Hyatt
Date: 16:56:31 10/15/98
Go up one level in this thread
On October 15, 1998 at 14:08:24, ed wrote: >On October 15, 1998 at 10:12:45, Robert Hyatt wrote: >>No it can't. Alpha/beta doesn't produce a score for *every* move searched. It >>produces a score for the first (hopefully best) move and then only proves that >>the score for every other move is lower, without determining how much lower. > >Yes, I do understand Alpha/Beta. Many a times, a pgm will have a tough time >deciding/alternating between move A or B or C, and, and it is these move A and B >and C that I want to see in the written analysis. Also, if the top 3 variations >is displayed on the screen, it gotta be also possible to save to a file, won't >you say? :) > One point: repeat after me slowly: "there is no way for alpha/beta to produce a best move, a second-best move and a third-best move, without tripling the time it takes to complete the search..." There are approximations. IE you get the best move at ply=10, but then it is replaced later by a better move at ply=10... you can show both PVs and say they are the best two. Or you can save the ply=9 pvs too, but that is much less accurate. Or you can do it the correct way: search all moves, save the best as "#1", then remove that move from the move list, and search all the moves except that one again. That gets #2. Repeat as often as you want, knowing that each search will take about as long as the previous one, since the best move takes 90% of the total time... > >>It takes a second search to get the real score/PV for a second-best move... > >Hmmm... now that u mention it, it does make sense.
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.