Author: Sune Fischer
Date: 04:13:58 03/03/02
Go up one level in this thread
On March 03, 2002 at 06:03:47, Uri Blass wrote: I think is is more interesting to ask these question at the higher plies. If a 20 ply searcher has a rating of 3000, and perfect play is 3300, then there must be real heavy DR going on after ply 20. A mere 300 elo in going 20 plies to 80 plies, when you get 2000 elo from 1 to 20 plies? But on the other hand, if the 20 ply searcher does the correct move in 99% of the cases, then it seems there is not much to gain. These are the two seemingly contradictory observations that bothers me. It would be interesting to compare the elo's of 20, 21 and 22 ply searchers, just to see if there is any DR setting in or not. >suppose you have the following 1000 programs: > >program A0:play random moves with probability of 100% >program A1:does a search of 3 plies to choose the move in 0.1% of the cases >and plays a random move in the rest of the cases > >program A2:does a search of 3 plies to choose the move in 0.2% of the cases >and plays a random move in the rest of the cases > >... > > >program A999 does a search of 3 plies to choose the move in 99.9% of the cases >and plays a random move in the rest of the cases. > >programA1000 does a 3 ply search in every move. > >Questions: >1)suppose you play 1000 matches of 1000 games between programs >Ai and Ai+1(i=0...999) and suppose the rating of A0 is 0. > >suppose that you caluclate for every i rating based on the performance > >What is going to be the expected rating of program A1000? > >2)Suppose you play 1000 games between program Ai and Ai+2(i=0,2,4...998) >and suppose the rating of program A0 is 0 > >What is going to be the expected rating of program A1000 based on similiar >calculations? >In which case the expected that the rating is going to be higher? > >Uri
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.