Author: Les Fernandez
Date: 14:42:03 07/11/03
Go up one level in this thread
On July 11, 2003 at 15:48:46, Victor Zakharov wrote: >Hi Les, > >On July 10, 2003 at 17:18:16, Les Fernandez wrote: > >>On July 10, 2003 at 06:58:43, Victor Zakharov wrote: >> >>>>Some time ago I remember reading that someone was able to take CAP data that has >>>>been collected and analyzed and wrote a program that generated 4 times the >>>>number of positions in the entire collection that were all unique positions! >>>>does anyone remember this and what has been done with it or is it still being >>>>worked on? I dont remember any of the details but remember it sounded very >>>>interesting. >>> >>>Sorry, I don't understand what youn mean. >>>Current CAP as a rule contains all positions that were at least in two games. >> >>Hi Victor, >> >>What I think he was referring to was some time ago I wrote an app that took the >>analyzed EPD's on Dann's site (at that time there was about 6,000,000 positions) >>and generated an additional 18,000,000 positions in about 30 minutes bringing >>the total up to 24,000,000. These extra positions were as sound and accurate as >>the original positions. What was nice about this is that it required no >>analysis engine and was done strictly based on symetries and current analyzed >>positions. In fact I have also thought of other ways of accomplishing a further >>increase for the number of positions but have not had time to continue it. I >>think at one point they were posted on Dann's site but I have been so busy I >>have been out of the loop for a while but hope to continue these studies in >>chess that I had started a while ago. I miss this stuff and hope to be jumping >>on board again shortly. Hey Dann if you are reading "HELLO" >> >>Les > >Thanks, I understood. Surely it can be usefull for some purposes. But for me Hi Victor, When you say useful for some purposes are you referring to the CAP data?? If anyone finds CAP data helpful then having 24,000,000 posiitions available instead of 6,000,000 can only be an asset. Of course thats assuming that the newly generated positions can contain evaluations no better then the original set they were built on. Anyway what is nice about generating new epd sets without using a chess engine is saving us alot of time. Think about this! If we have say 5,000 positions that have been "carefully" analyzed by a bunch of GM's and found to be good then using this method we can now build that data set up to 20,000 positions with no need to reanalyze those positions. As you know the CAP data is coninually being improved on and as time goes on and as Dann finds flaws in a particular position it gets filtered out. Chances are that a majority of the evaluations should hold up rather well. Thanks Victor for your comments and lets keep chess research going. This is a very exciting field and only time will tell how far we can take it. Take care, Les the >main problem of CAP is not the number of positions. Making evaluations more >reliable is the trouble. > >Victor
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.