Author: Dann Corbit
Date: 14:08:49 01/27/00
Go up one level in this thread
On January 27, 2000 at 15:33:44, Tim Mirabile wrote: >On January 27, 2000 at 02:45:39, Jeremiah Penery wrote: > >>First, a postulate: Any increase in *search depth* will correlate to a ratings >>increase. I believe this to be a fairly linear curve. This will never give >>diminishing returns. > >This is a very big assumption right there. Hard to believe that the rating >difference between a pair of programs searching one ply vs two ply will give the >same rating difference as programs searching nine vs ten ply and as programs >searching 30 vs 31 ply. Yes. Intutitively, a program searching one ply deeper would be many times better, because it has many times as much information. Consider the starting board position and the number of possibilities that spring forth from there: White(1): perft 1 total moves=20 To examine our immediate move only, we must consider 20 board positions. White(1): perft 2 total moves=400 To examine our all possible replies, we must consider 400 board positions. White(1): perft 3 total moves=8902 To examine our all possible replies to those replies, we must consider 8,902 board positions. White(1): perft 4 total moves=197281 To examine our all possible replies to those replies to those replies, we must consider 197,281 board positions. White(1): perft 5 total moves=4865609 To examine our all possible replies to those replies to those replies to those replies, we must consider 4,865,609 board positions. White(1): perft 6 total moves=119060324 After just one more level, we have 119,060,324 possibilities. So as we read and examine a successive ply, we are gathering a large multiple of the previous information (along with all that it contains). Hence, by intuition, an additional ply should make us play many times better than before. It does not, so intuition is foiled.
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.