Latest Updates:
Page Index Toggle Pages: 1 2 3 [4] 
Topic Tools
Very Hot Topic (More than 25 Replies) IDEA (Read 33775 times)
Vass
God Member
*****
Offline


I Love ChessPublishing!

Posts: 1105
Joined: 06/22/11
Re: IDEA
Reply #11 - 10/17/13 at 18:44:42
Post Tools
Another method, used by one of the best correspondence chess players in the world (I won't reveal his name here):
He simply gives the root position to the IDeA with aggressive settings to generate almost every possible move. Leaves it for four days to calculate. It results in the widest possible tree several moves further starting from the root position (of course, the number depends on the computer power). This is a method I call a 'brute force'. Then he shapes the tree, so IDeA starts analysing only the promising lines, leaving it for 2-3 days. And at last, he enters moves by hand all along the (resulted) lines depending on his understanding of chess, giving these to IDeA to prolong them..
Not a promising line can be missed like that, believe me!  Roll Eyes
  
Back to top
 
IP Logged
 
Vass
God Member
*****
Offline


I Love ChessPublishing!

Posts: 1105
Joined: 06/22/11
Re: IDEA
Reply #10 - 10/17/13 at 18:26:27
Post Tools
Well, let me reveal something about the methods I use!
First, I run infinite analyses with as many different engines as I can (one by one), depending on the time I have. I save all lines in a pgn-file and then move slowly along the given lines while writing all the alternative lines that engines give in this same pgn. It results in a tree of infinite analysis lines made by hand (i.e. the same you can do using the deep analysis tool of Fritz 13 or other programs).
And as you can easily guess, all these lines I give to the IDeA (without all the evaluations). So IDeA takes these lines as high priority, evaluates them move by move, stepping on each one, one after another. Then creates alternatives to every move.. (even for the first ones which start the lines) Prolongs the promising lines.. deeper and deeper.. On and on.. Endlessly..
And it often turns out that the line which most of the engines gave in their infinite analyses is wrong after, say, 8 or 10 plys, so IDeA corrects this line turning back and choosing another direction/branch which was cut off by the 'big guns', developing it....and gets near the target.
  
Back to top
 
IP Logged
 
brabo
God Member
*****
Offline


Welcome chessfriend

Posts: 945
Joined: 02/02/07
Re: IDEA
Reply #9 - 10/17/13 at 16:37:17
Post Tools
Correspondence analysis tool , manual comparison of engine-evaluations, infinite analysis of specific moves deep in the branches,... are different ways to avoid the automatic cutting of the branches by the engines.
In the end it is about what gives you the best results in a limited amount of time. I fully agree that purely relying on infinite analysis won't be the best choice to hit the target but I have also strong doubts that relying solely on the 'idea' tool without anything extra will be the most efficient way of analysis. That the 'idea'tool can be another nice weapon in the arsenal of analysingtools, this I won't deny but I've now the impression that it is presented as the holy grail.
  
Back to top
WWW  
IP Logged
 
Vass
God Member
*****
Offline


I Love ChessPublishing!

Posts: 1105
Joined: 06/22/11
Re: IDEA
Reply #8 - 10/17/13 at 16:11:51
Post Tools
Well, the stress in IDeA's analysis lays upon a simple rule which I would describe like this:
Imagine you have to shoot at a target which is three hundred meters away or so (you never know how far it is and you never know where exactly is) with a gun and you don't want to miss it. Trying to shoot with Houdini's, Stockfish-es or Komodo rifles would never guarantee you a good shot. And this is all because these rifles can't see where the target is. (They simply cut off thousands of branches right at the start of the analysis, for example. Not to mention the millions of branches that they cut off in the analysis process).
Well, so take your old pistol (even your Rybka one) and start shooting endlessly to and fro while advancing further, a pace left, then a pace right, then a pace straight ahead...until you reach the goal (the target). Moving slowly further and further you can measure the right distance to all the objects, including the target. And this is what IDeA is all about.
It's useless to say that if you reach the target just a meter away (or 15 plys in depth), you can use your knife to spot it.  Grin

Edit: What IDeA is not allowing to the engine is..not to cut off the branches just like that.  Wink
  
Back to top
 
IP Logged
 
brabo
God Member
*****
Offline


Welcome chessfriend

Posts: 945
Joined: 02/02/07
Re: IDEA
Reply #7 - 10/17/13 at 15:41:47
Post Tools
In the idea-manual is mentioned the following which attracted my interest and generated my previous comment, see chapter 11: Trees
" This means that here we have the evaluations of two different chess engines side by side. You can even color the evaluations where the difference between the two engines exceeds a user specified threshold. Such differences might indicate a position that needs further analysis."
Now rereading this statement, i believe to understand better what is meant:
1) comparison is indeed shown per move between engines so as I stated (not only for infinite mode)
2) the difference might indicate that more analysis is needed which I wrongly interpreted as an automatic process while in reality it is still up to the user to decide if something needs to be done.

I believe the strength of good analysis lays for an important part in comparing analysis done between different engines (no matter which method was used) and adapt when differences in evaluations between engines are shown. As it seems little attention is given to comparing analysis between engines in IDEA, this sounds to me a clear weakness (compared to the method which I use today).
  
Back to top
WWW  
IP Logged
 
Vass
God Member
*****
Offline


I Love ChessPublishing!

Posts: 1105
Joined: 06/22/11
Re: IDEA
Reply #6 - 10/17/13 at 14:59:57
Post Tools
@ TonyRo
Using 4, 6 or 8 cores independently (calculating 4, 6 or 8 moves / branches) is much more efficient than using all the cores calculating a single move in a position.

@ brabo
No, brabo! Comparison between two or more engines' evaluations is for the infinite analysis. IDeA works on hard rock values. If you have 4, 6 or 8 cores on your computer you'll have to set 4, 6 or 8 SP threads of a single engine (for example, you install your Houdini 3 in the Aquarium GUI on one thread/core, then repeat this several times, naming them _1, _2, _3, _4 and so on, for example. And you can continue as soon as  the number of the installed Houdini engines becomes equal to the number of the cores/threads in your computer). Starting an IDeA analysis means that you run simultaneously your chosen number of Houdini's, each one of them using one core/thread.  Wink
  
Back to top
 
IP Logged
 
MNb
God Member
*****
Offline


Rudolf Spielmann forever

Posts: 10478
Location: Moengo
Joined: 01/05/04
Gender: Male
Re: IDEA
Reply #5 - 10/17/13 at 13:58:23
Post Tools
Moved by MNb.
  

The book had the effect good books usually have: it made the stupids more stupid, the intelligent more intelligent and the other thousands of readers remained unchanged.
GC Lichtenberg
Back to top
 
IP Logged
 
brabo
God Member
*****
Offline


Welcome chessfriend

Posts: 945
Joined: 02/02/07
Re: IDEA
Reply #4 - 10/17/13 at 13:36:41
Post Tools
Reading the manual, I realize that I am already doing (and more) what IDEA does but then manually. I see 1 major difference and that is in the comparison between the evaluations of different engines. IDEA seems to check for a certain move the difference in evaluation between different engines against a threshold value if more analysis is needed. I don't do that as I realize that single evaluations of an engine often tell you nothing about the quality of the analysis.  I prefer to compare the difference of evaluations of 2 branches with 2 engines. I try to explain:
Example 1:
A) Method of IDEA:
move x with engine A: +/=
move x with engine B: +/-
Threshold past so more analysis on move x is needed.
B) My method:
move x with engine A:+/=
move y with engine A:+/-
move x with engine B: +/-
move y with engine B: +-
No extra analysis needed as both engines give same type of difference between the moves so most likely the evaluationdifference is linked to the way how engines are evaluating positions.
Example 2:
A) Method of IDEA:
move x with engine A: +/=
move x with engine B: +/=
Threshold not past so no more analysis is needed
B) My method:
move x with engine A:+/=
move y with engine A:+/=
move x with engine B: +/=
move y with engine B: =
It is unlikely that the difference in evaluation of move y has to do with differences in how the engines are evaluating.
- Move x by engine A is too low evaluated
- Move y by engine A is too high evaluated
- Move x by engine B is too high evaluated
- Move y by engine B is too low evaluated
We need more analysis to get a better evaluation.
  
Back to top
WWW  
IP Logged
 
TonyRo
God Member
*****
Offline


I'm gonna crack your skull!

Posts: 1765
Location: Cleveland, OH
Joined: 11/26/07
Gender: Male
Re: IDEA
Reply #3 - 10/17/13 at 13:25:03
Post Tools
People also seem to be excited about the fact that IDeA allows you to use multiple single-core versions of engines 'n' times for added efficiency.

For example, my home computer has a hex-core i7, and can run MP versions of, let's say, Houdini, and I'd get performance (I'm guessing here) something like 4-5 times better than if I just ran a SP version. You don't get the full 6x performance (presumably) because there are some inefficiencies in the master-slave algorithms used in spreading out the analysis of one position to multiple processors and compiling it back together. Instead, in IDeA you can use 6 copies of Houdini SP, though I don't know all the details of why this is somehow better, nor have I tested if this is actually any better.

As an enthusiast of CC and analysis in general, I've always wanted to become very proficient in IDeA, though I've never totally understood it or got into it. I have a hard time letting anything sit for 4 days without just trying things and moving around the pieces.

Grin
  
Back to top
 
IP Logged
 
Vass
God Member
*****
Offline


I Love ChessPublishing!

Posts: 1105
Joined: 06/22/11
Re: IDEA
Reply #2 - 10/17/13 at 10:27:59
Post Tools
Nothing is near the pure calculating force of IDeA (the exact name - Interactive Deep Analysis). Believe me!

The latest manual for Aquarium is:
http://chessok.com/download/Aquarium/5/doc/aquarium/Aquarium5.pdf
Check out the appropriate section!

Some good articles you can find on the ChessOK site. For example this one:
http://chessok.com/?p=26620

Anyway, I'll try to explain how IDeA works.
It generates a move, the engine evaluates it in a certain depth (you can put whatever depth you want) and stores the number and the move (say, +0.35) so you can see them anytime you want. Then generates another one...and so on. Then steps on the first move and goes further.. It creates a spider web, but that's not all! The difference between this tool and all the others is that it is "backsolving" (it shows the new evaluation of this line on the start of the line), compares it with the stored evaluations of the other previously generated moves and starts with the next promising alternative). And if, after 20 or even 50 plys in the tree it finds this line is not promising (a tablebase hit with a draw conclusion in our pawn endgame case here for example, or the evaluation just goes down after move 'n' in the line) it turns out to the next generated move as a promising one and goes further. It compares all the evaluations of all the generated moves over and over again. And picks up the most promising one, evaluates it and if promising goes further on this line. As soon as it finds this line is not promising, it does the process again with the next generated move. It works in two stages - creating alternatives and prolonging the promising lines. After each one of them it 'minimax'-es the tree (i.e. does 'backsolving), compares all the evaluations and continues endlessly.
Don't let all the manuals and articles you'd find on the subject mislead you it is easy to analyze with this tool! It's not! It needs all the abilities one has to comprehend how exactly this tool works. It took me two years to train up myself on this subject. I've seen more than 50.000 moves/positions generated from the start position and calculated at very great depth and still nothing near the truth (for example, in a test position where I know where it is). So, this tool can be unproductive and very dangerous for the one who uses it unprepared. It needs a deep comprehension and constant hard-working to achieve the truth.
Anyway, it went to off-topic here, so if you want some help in this matter you can create a new discussion on the appropriate computer section. And I'll be there..  Wink
  
Back to top
 
IP Logged
 
brabo
God Member
*****
Offline


Welcome chessfriend

Posts: 945
Joined: 02/02/07
Re: IDEA
Reply #1 - 10/17/13 at 09:20:04
Post Tools
Vass wrote on 10/17/13 at 06:59:11:
Well, guys! I think you don't have to suffer in a pawn endgame, no matter how complex it is.
There are some modern ways for analysis that can solve things easily. For example, there is a tool called "IDea" in the ChessOK Aquarium GUI. It is a fantastic tool for analysis, that can solve problems by pure force calculating. Ir generates a move, steps on it and goes deeper, creating a web of moves like a spider. 150.000 - 200.000 positions/moves can be created for 4 to 10 days depending on the processors' power of your computer, every one of them calculated at 30 plys depth or more depending on the settings you use. It creates a tree which can be very 'heavy' in GBs, but nothing exceptional. Every top correspondence chess player uses this tool. The so-called "infinite analysis", even at 45 plys depth, can't beat it. When using this IDea tool for analysis one can easily see how incompetent can be the usual "infinite" one that everyone uses.
Anyway, the only problem is that this IDea tool is very hard to master in order to become skilled in analysing with it. I for one spent two years of my leisure time to  gain command of it. Not to mention the processors' time which went for it..
As for the opponent guy - he probably used this tool to solve this pawn endgame. Not a hard task I suppose - only a matter of time.  Wink
When a correspondence chess player uses this tool properly for analysing his games he practically can't lose. All he needs is entering a decent opening line which evaluation is close to equal. The only problem is the time. Right now, I play more than 45 correspondence chess games and in fact I can use this tool for no more than 10 of these.  Embarrassed
And I'm really sorry I can't help you in this endgame. I'll start playing 10 more games in the 1st of November... Anyway, not helping you may be for good. After all, what's the beauty in analysing a pawn endgame by pure force and mechanical skills!?  Roll Eyes

I don't have "IDea" in the ChessOK Aquarium GUI but I do believe my method of analyzing also produces good results. It is a far more complex method than purely relying on infinite analysis, described in http://schaken-brabo.blogspot.be/2012/05/analyseren-met-de-computer.html. (maybe I will translate it once to English as it is rather technical and likely difficult to read via googletranslate)
B.t.w. I've the impression that there is little difference between this idea-tool and the correspondence analysis mode available in engines running under the standard chessbase GUI.
  
Back to top
WWW  
IP Logged
 
Vass
God Member
*****
Offline


I Love ChessPublishing!

Posts: 1105
Joined: 06/22/11
IDEA
10/17/13 at 06:59:11
Post Tools
Well, guys! I think you don't have to suffer in a pawn endgame, no matter how complex it is.
There are some modern ways for analysis that can solve things easily. For example, there is a tool called "IDea" in the ChessOK Aquarium GUI. It is a fantastic tool for analysis, that can solve problems by pure force calculating. Ir generates a move, steps on it and goes deeper, creating a web of moves like a spider. 150.000 - 200.000 positions/moves can be created for 4 to 10 days depending on the processors' power of your computer, every one of them calculated at 30 plys depth or more depending on the settings you use. It creates a tree which can be very 'heavy' in GBs, but nothing exceptional. Every top correspondence chess player uses this tool. The so-called "infinite analysis", even at 45 plys depth, can't beat it. When using this IDea tool for analysis one can easily see how incompetent can be the usual "infinite" one that everyone uses.
Anyway, the only problem is that this IDea tool is very hard to master in order to become skilled in analysing with it. I for one spent two years of my leisure time to  gain command of it. Not to mention the processors' time which went for it..
As for the opponent guy - he probably used this tool to solve this pawn endgame. Not a hard task I suppose - only a matter of time.  Wink
When a correspondence chess player uses this tool properly for analysing his games he practically can't lose. All he needs is entering a decent opening line which evaluation is close to equal. The only problem is the time. Right now, I play more than 45 correspondence chess games and in fact I can use this tool for no more than 10 of these.  Embarrassed
And I'm really sorry I can't help you in this endgame. I'll start playing 10 more games in the 1st of November... Anyway, not helping you may be for good. After all, what's the beauty in analysing a pawn endgame by pure force and mechanical skills!?  Roll Eyes
  
Back to top
 
IP Logged
 
Page Index Toggle Pages: 1 2 3 [4] 
Topic Tools
Bookmarks: del.icio.us Digg Facebook Google Google+ Linked in reddit StumbleUpon Twitter Yahoo