Not logged inRybka Chess Community Forum
Up Topic Rybka Support & Discussion / Aquarium / 2000 fold repetition
- - By Ghengis-Kann (***) [us] Date 2014-06-26 02:19
Hi.

One of my projects has approximately 6000 positions downstream of a particular node, which are 2000 fold repetitions of the same 3 positions even after an "accurate" position count has been done.

Is there any way to deal with this?
I know there is some kind of crop function but I am heasitant to use it without understanding how or what it actually does.
Parent - - By Dhanish (***) [in] Date 2014-06-26 03:33

> 2000 fold repetitions of the same 3 positions


What do you mean? The IdeA tree stores one position only once.
Parent - - By Ghengis-Kann (***) [us] Date 2014-06-26 16:39
Hi Dhanish.

One of my projects had a position with a Black knight on d7 and white queen on a4.
The following moves are in the idea tree ...Nc6 Qb3 ...ND7 Qa4  ...Nc6 Qb3 ...Nd7   Qa4 and so on for almost 6,000 positions

I guess there are actually 4 positions now that I wrote it down, but it is clearly not the case that the tree stores each position only once.
Parent - - By Dadi Jonsson (Silver) [is] Date 2014-06-26 16:45

> it is clearly not the case that the tree stores each position only once.


Dhanish is correct. Why do you think that your example proves otherwise?
Parent - - By Ghengis-Kann (***) [us] Date 2014-06-26 17:59
Hi Dadi.

I'm not trying to prove anything.
I just want to understand how I can combine multiple IDEA projects into a single tree that is minmaxed to the starting position with an accurate count of non-identical positions.

When I see a long linear sequence of moves where N = 2 but there is only one move downstream of each node I take that to mean that the line is duplicated in the tree.
Both you and Dhanish are saying that is not correct.

When I export the same project multiple times to the same main tree the number of positions downstream of the root increases without and unique positions being added.
To me that seems like duplicate positions are being stored.

Would you be willing to elucidate me on what is actually going on here?
Parent - - By saurus_ (**) [de] Date 2014-06-26 19:02
I think the correct question would be, how to get the correct number count.
I think there is no simple answer on that.

Start with
1.Sf3 Sf6

now the next move could be

2. Sg1 Sg8 converting back to starting position.

Now all possible openings can be reached from 1 Sf3 Sf6

I guess this is correct from the roles of chess, but not what most want to see.
Parent - - By Ghengis-Kann (***) [us] Date 2014-06-26 20:08
Hi Saurus.

I think you are correct here.
My misunderstanding was that N represents the number of positions downstream from a particular node.
I guess it is actually the number of moves, which can be repeated among identical positions.

So....
Is there any way to remove multiple copies of identical moves that connect 2 positions in a .hsh tree?
Parent - By buffos (Silver) [gr] Date 2014-06-30 14:39

>y misunderstanding was that N represents the number of positions downstream from a particular node.


thats correct.

N represents (roughly, because its not always an exact count) the number of positions in the subtree.
Parent - - By buffos (Silver) [gr] Date 2014-06-26 20:06
you are navigating a tree and making circles.

When you do circles in real life and you pass from the same building 200 times, that does not mean that there are 200 instances of the building.
The same happens here. You run in circles visiting the same positions again and again.

>but it is clearly not the case that the tree stores each position only once.


You are confusing the tree and the navigation, which in our case is the notation. The notation is the log you do while navigating the tree.
Parent - - By Ghengis-Kann (***) [us] Date 2014-06-26 20:15
Got it.

Could you tell me how to crop the circular moves out of the tree when IDEA prolongates a 2000 fold repetition of the position?
I tried clicking crop in the tree menu but nothing apparent happened.
Parent - - By buffos (Silver) [gr] Date 2014-06-30 14:37

>how to crop the circular moves


there is nothing to crop.

imagine there are 3 nodes in the tree A->B->C->A

3 nodes in a circle. You cannot break the circle because  A===A.

Cropping is removing. In this case, if you remove node C you will have A->B

but why would you do that?

Idea DOES NOT prolongate any repetition. YOU are prolonging it when choose to run over the circle again and again.

>Got it.


from your question it seems you still got that confused.
Parent - - By Ghengis-Kann (***) [us] Date 2014-07-12 16:22
I am not prolonging it.

I just started out completely from scratch with Aquarium 2014 and have attached 4 screen shots from my first IDEA project.

The first 3 are consecutive moves of an infinite repetition and the 4th is what the tree looks like a few moves earlier.

I DID NOT prolongate these lines thousands of times by hand.
Attachment: IR1.jpg (499k)
Attachment: IR2.jpg (508k)
Attachment: IR3.jpg (490k)
Attachment: Bloatedtree.jpg (527k)
Parent - - By buffos (Silver) [gr] Date 2014-07-13 07:26
I really do not undertand what is your problem.

IDEA creates a chess tree. In it there are cycles (repetitions). Every cycle is contained in there once and can be traced infinite times

I really do not understand your problem. You probably do not understand what a tree is or something else is confusing you.

Are you confused by the Numbers next to evaluations. If yes ignore them till you understand what is going on. They are just an indication of the population of the subtree of the move.
Parent - - By Ghengis-Kann (***) [us] Date 2014-07-13 13:47
Hi Buffos.

The N is useful for seeing how much analysis IDEA has done on a particular variation, but it becomes useless if one of these crazy repetitions is downstream of it. 

Maybe it's not a bug and I am the only person who thinks this makes the N a pointless value, but why does it even happen?
Parent - By buffos (Silver) [gr] Date 2014-07-13 20:59
you just cannot quantify in your had the numbers

if from a move, the subtree is 10 moves long and every move has 2 alternatives (a binary tree of length 10) then you get 1024 = 2^10 positions

3000 positions for a subtree are not much. A lot of times moves share the same subtree (transpositions)

If you cannot visualize that i cannot help more. You just have to either ignore N or undestand the numbers.

Thats all i can help. N is usefull but you should understand what N actually is. Draw a tree in paper and try to understand it.
Up Topic Rybka Support & Discussion / Aquarium / 2000 fold repetition

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill