![[us]](/mwf/v12/flags/us.png "United States")
I am using Aquarium 11. I entered a game into IDEA and set up four positions (Nodes) within the game to be analyzed. After reading Dadi's article I set the configuration to 120 secs OR 16 ply. I have 6 meg of Ram so I set up four Rybka 4.1 engines with a 750 meg hash each. I thought that with a configuration of 120 secs OR 16 ply the analysis would take awhile. But, I started seeing the tree fairly quickly. I let IDEA run all night and was quite pleased with how smoothly it ran.
I then went through the tree and marked some of the moves Green (also from Dadi's articles) and IDEA is still running while I have been at work.
Any thoughts, ideas or suggestions?
I then went through the tree and marked some of the moves Green (also from Dadi's articles) and IDEA is still running while I have been at work.
Any thoughts, ideas or suggestions?
![[in]](/mwf/v12/flags/in.png "India")
Since your computer has 6 MB RAM, I think it is a new one. Hence the depth is too low, you will achieve it within a few seconds (depends also on engine, and whether opening or ending, etc). The minimum time of 120 s will never kick in, because it is OR.
Try analyzing some well known positions, whether IdeA is able to find the best move, to get a feel.
Try analyzing some well known positions, whether IdeA is able to find the best move, to get a feel.
So I can set it for more than 16 ply? What is the ideal depth? Or at least a good depth for a deep analysis.
The computer is a little over 2 years old. Using Rybka 4.1. And the position is a middle game position from a game I recently played.
The computer is a little over 2 years old. Using Rybka 4.1. And the position is a middle game position from a game I recently played.
I do 60 seconds and 22 ply, with max time of 5 minutes...its not the speediest but I think it gets good results. What I tend to do is work through lines and where there are a lot of candidate moves I do deep IA on candidate moves to hone in on a few I think are most promising and, where applicable, use game database statistics to also identify candidate moves then I have IDEA focus on those candidate moves. It seems to be getting me what I want - but I am sure there may be better approaches.
That sounds like a good idea. I switched mine to 300 seconds and 32 ply. It has been running overnight and I am going to let it run all day.
> 32 ply.
This is too high. With Rybka, we have to add 3 to the reported ply, to be comparable with other engines. The point of having a minimum depth is that in case there are some tactics, the engine will not achieve high depths, thus taking more time for analysis, which is what we want. So 32 ply means this condition will never be used.
There are no ideal IdeA settings. I'll outline one idea. See what is the time available. If you have less than overnight, simple infinite analysis is best. If more time is available, IdeA is worth it. Decide on the number of positions you want added to the tree at the current position. Generally I like at least a total of 100 in the openings. Towards the endgame, I prefer 1000. Divide the available time by the number of positions you want added in the tree. That should be the approximate time per move. To fix depth, run the engine in infinite analysis mode for this time. See what depth is achieved. Max depth can be one ply more.
Sounds good. I will work with it more when I get home. Thank you for your help.
Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill