We need more of these - I find it so much easier to understand concepts when you see them in action rather than just reading instructions....
Does this still happen if you have multiple concurrent projects active or multiple roots in the same project?
In the current Stage when the number of tasks remaining was less than the number of engines available they just waited.
IDeA seems very interesting and groundbreaking, I am waiting to see a completed output of IDeA
> positions never completed.
what do you mean completed? Do you see the tree being created? Do you see the green boxes?
See the Tree window in the videos for example output.
> IDeA doesn't stop until it solves chess (=never).
So, if I give it a King, Bishop and Knight Vs. King ending it will never solve it?
> For relatively short values of never.
That makes sense.
In parallel I also started another project with the same position and limiting the black and white alternatives to 5 and the tree width to 10% and the prefered side as white. this is still analysing at the time of writing with Rd6 showing and idea value of +156 and Ra6 with and Idea value of +7.
Is this normal or have i missed a point somewhere ????
Thanks again and thanks in advance!
Note that multiple engines can also be useful for simultaneous infinite analysis and in that case you would often use several different engines where shared hash would make no sense at all. Of course you can also use different engines (simultaneously) in IDeA, but that would not be wise, except perhaps closely related engines such as the Rybka family (standard, human dynamic).
All this has nothing to do with Vas' and Gian-Carlo's interesting cluster implementations. However, if/when a cluster Rybka or Sjeng become available they could be used for analysis in Aquarium (as long as they are UCI engines), both infinite analysis and IDeA. Maybe the best practice for using a cluster engine will differ from what we are used to recommend for "normal" engines and perhaps IDeA needs to be modified to take full advantage of the new possibilities. We'll see.
It was already clear to me, that sharing hash in both directions over internet would cost more than pay, but I could imagine some kind of triple or multiple brain feature, that would collect eval and order of moves from engines connected by network into one master engine or maybe even GUI only, which would help choosing a single move or variant to think about deeper.
No cluster ersatz of course, but much cheaper and easier maybe with some more effect then simple remote control, which is possible in much more ways than by a feature of a chess GUI.
> I could imagine some kind of triple or multiple brain feature, that would collect eval and order of moves from engines connected by network into one master engine or maybe even GUI only, which would help choosing a single move or variant to think about deeper.
Aquarium can display the analysis of different engines in a single window (or multiple windows, if you prefer, or want more info about the analysis of each engine). However, you must make the final move choice yourself, as the engines scale their evaluations differently. An evaluation of, let's say, +1.58 has a different meaning depending on the engine that produced it.
Here is a quick example, showing four engines analyzing the same position:
This is especially useful if you have access to more than one computer and run some of the engines over a network. You can see some similar examples in this post.
In my version (4.0.2) I cannot change the number, the number is always 1.
I can only change this number if I choose Rybka 3 (instead of Rybka 3 1-cpu) as IDeA-engine.
However, running e.g. 4 instances of Rybka 3 (multi-cpu) does not make much sence on a computer with 4 cores.
I can also asume that R3 1-cpu has not been moved to another directory but check it in any case from the path variable in the Engines mode.
I am experimenting with the new IDEA and have the following question: Once you have set-up a project and starting position, how can you influence from what poisition in the tree the AN starts?
It used to be that IDEA started from the position from which you started IDEA, but it does not seem to work that way anymore.
Also you can also create a project that includes the whole starting line (create a new project from a game in sandbox or database to see what i am talking about)
is there any tutorial on this, or perhaps something Dadi wrote for chesscafe?
EDIT: hang on I think I found it. Very elegant :)
but question about comprehensive tutorial remains - perhaps make a sticky thread with all tutorials and dadi's posts?
When there is a problem ask and someone will share their experience i guess ;-)
if yes, can you provide it as a downloadable pdf? i would like to print it.
thnx for the great work!!
i would like to assign different engines to different projects, guess that's not possible (yet)?
Videos are the most pedagogic way
After three days of analysis there was still no tree and no evaluation!!
And one has to go through so many "f" (finished) positions that creates a pain.
So where is a mistake in my setup.
How to get a good result (position evaluation + tree) within a reasonable amount of time.
> After three days of analysis there was still no tree and no evaluation!!
the part where you press NEW to create a new configuration... you skipped that one.
But where are the advantages of IDEA in comparison with "Game Analysis" as this function is already extremely good.
The 2 methods are not compared.
I have found that clicking on so many "f" (finished) positions is time consuming.
Maybe one suggestion - after IDEA has done its work on multiple positions it would be nice to put it in ONE GAME with numerical assessment of each contiuation and sorting from the highest to the lowest assessment with regard to the side on the move - for white positive for black negative assessment.
Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill