TonyRo wrote on 02/02/15 at 18:45:11:
Revisiting this old thread to ask Vass or anyone else with IDeA experience some questions (if he/they have the time to answer me). I have been using IDeA exlusively in an ICCF Master Class tournament and have been having good results so far, I'll post my finished games below. But as I use it, I still have some thoughts and questions:
1. Do you have any preference on analyzing deeply or not deeply per task? I have a year or two old i7 6-core processor, so my computer is fairly powerful, but I can't help but think that based on how IDeA works and how the analysis is structured, it might be better to use shorter depth/time controls per task to simply run through a lot more tasks per time. Because IDeA continues to prolong lines and minimax the tree, it seems like shorter depth/task isn't a huge deal - it might be better to simply have a tree with two times as many positions at shorter depths. Maybe I'm wrong here.
2. Have you had any luck using IDeA in the Mar Del Plata? Computers have a hard time with this line in general because of the heavy sacrificial attacking resources Black has that frequently feature quiet moves, combined with the fact that the positions are very closed, but IDeA might be struggling even worse. It just generates so many alternatives, and in each one usually fails to reach the depth and find all of the necessary moves for Black, even given large amounts of time. There are just too many branches. I tried using the "aggressive" tree option, but that might be just making it worse. The use of the thematic moves perhaps helps some, but I'm not sure what format those strings need to be in for Black - sometimes I notice it's not always checking them. Maybe I need to move to longer time/depth and limit the White alternatives to 4-5 just to make sure it's focusing on a smaller subset of the positions. I make progress by linking the projects and inserting analysis of my own and games from databases, but it's really not enough to make good progress.
3. I am currently using 2014 - it crashes on me overnight probably 70% of the time. I wake up and it's not responding. Does this happen to you? Do you know if the upgrade to 2015 is worth it?
4. What is your approach with the tree shapes? I have been using low time/depth and "aggressive" at first to fill up the tree with a ton of a moves for a few days, and only then moving to thinner shapes like Default or Moderate, occasionally Longest to focus on the most important moves, always moving the roots around as needed and highlighting moves green, etc. Just wondering what other people were doing as far as IDeA strategy was concerned.
Thanks!
...
Hi, Tony!
I'm, glad you are deep enough in IDeA, so we can discuss pluses and minuses of this outstanding tool for analysis!
1. "...it might be better to use shorter depth/time controls per task to simply run through a lot more tasks per time." Yes! 10 seconds (max 15) "and" 21-23 moves must be enough for most tasks.
2. Of course, Mar Del Plata is a real test for the good correspondence chess players! "How to" do it?
Well, first, you'll have to import all the moves from the played correspondence and high quality OTB games in the concrete variation you're interested in. [Create an IDeA project with a "root" (say 14...Nh5) - the starting point for your analysis job. Copy all the quality games you could find after this move, make a pgn out of them, paste it in your Sandbox, put the cursor after the initial (12...Nh5) move, right click with your mouse, "copy rest of moves" and paste them in your created IDeA project (after 12...Nh5) using "join moves" option. Then, stepping on every move in the variations you can make it task for analyzing or create an alternate task with the appropriate button on the top of the program window.] Secondly, you can create additional "roots" on the main moves you're interested in (say after 14...Rf7, 16...Bf8, 18.b5, 20.Nxb5 and so on..). Thus, IDeA, when creating tasks, will turn special attention to the positions you gave as "roots" and will create much more tasks for these positions. You can make thousands of "roots" if you want, but usually it's not necessary. The best method is to create additional "roots" after moves that can be countered with many variations. Of course, a root after, say, 19.cxd6 would not be necessary if you see that 19...cxd6 is the only answer. As for Mar Del Plata and the tree shapes, first use "aggressive" shape, then "tactical", next can be "wider", then "default", then "moderate" and the "longest" one in the very end.. And, please, do not stop entering "your" moves pressing the buttons on the top of the IDeA interface for analysis or alternatives! The engines are often blind for the best moves, especially in the Mar Del Plata variation. If you doubt an IDeA evaluation after a specific move, you can start an "infinite analysis" (put the mouse cursor after it and press the appropriate "infinite analysis" button on the top of the IDeA interface) with one of your six IDeA engines right into the project, leave it for some time and you'll see the line and the evaluation of the engine (while the other 5 engines will continue their work on the project). You can stop it by clicking with your mouse on the task (usually a white one) and while seeing the line the engine is counting on, press the underlined "finish task" below the tasks row. Then your engine will stop its infinite analysis and will take its usual job..
3. I used all the Aquarium versions since 2012 - now I'm on 2015 which is more stable than the 2014 version. My advices: Make a backup copy of the Aquarium "Data" folder every day (after stopping the analysis and exiting the program)! Aquarium is a buggy program, so use it simply - do not make connections between IDeA and Aquarium's database games. Copy and paste moves between Sandbox and IDeA interfaces is better than making connections between them and run an infinite analysis from the Sandbox while sending the analysis lines in the IDeA to be expanded. Most of the bugs come from the complexity of this software.. Next, if you have too many projects and too many positions are already analyzed, your IDeA will become slow. And if you experience this issue, then make a backup of the Aquarium's "Data" folder, uninstall your Aquarium, delete the folder it leaves and install it again. Then use the "Import" option to import your projects from your backup Data folder! This will eliminate all the evaluations of the IA column, but will import all the evaluations from the IDeA analysis column, which are necessary for the program to start from where you've been before. And IDeA will be faster in minimaxing and creating tasks than it was before. What is more... There are tools which you can run from time to time - better handled in the 2015 version.. For example, go to the Sandbox, find the appropriate button "Verify" and use it to verify your IDeA and "infinite analysis" trees as well as the "configuration folder". If a tree turns red after verifying, then it is broken. The only way to save your work is to backup your Data folder, uninstall Aquarium and then install it again, configure it as it was before and import the broken project with the "Import" button. This will import only the safe lines out of your big broken tree, cutting the bad ones. Of course, the IA column with its evaluations will be empty, but the IDeA evaluation column will be all there - and this is enough to continue from where you've been.. Another tool for maintaining the system is "compress trees" which can be used by right clicking on the status bar (on the very bottom on the program while in Sandbox or in IDeA window) and choosing it from the menu that is opening.
4. Highlighting moves green (or red, or whatever..) is useless. The only real marks for the program are the "roots" - using them is the key! As I already mentioned, your good work with the "roots" will bring you the results you're after.. Don't stop using them, because they are guiding the otherwise blind program what to analyse. Aaah, and one more thing, my usual observation is that you will need more than 50.000 completed tasks analyzed after every move, in order to decide that your analysis is done.. And sometimes, in complex positions like in the Mar Del Plata variation, even 100.000 completed tasks are not enough to decide upon the best move.. Tough task, isn't it?
As for the "shapes" of the analysis tree, two or three years ago one of the best correspondence chess players in the world said that in the complex positions he usually leaves the software to analyse with 100 % wide tree for 4-5 days, then 75 % wide tree for 3 days, then 50 % wide tree for 2 days and so on... Well, as I suppose he must have had very powerful hardware then... And no, it's not necessary! 33 % wide tree is enough for most of the positions you analyse... 50 % wide tree is too much already and can be used only in very complex positions.
So far...so good! Feel free to ask if more questions arise!
By the way, a helpful link for an appropriate thread in a well-known forum:
http://rybkaforum.net/cgi-bin/rybkaforum/board_show.pl?bid=9Your truly,
Vass