![[us]](/mwf/v12/flags/us.png "United States")
In a particular position there are some possible transpositions, eg 1...O-O 2.Qe2 Qc7 or 1...Qc7 2.Qe2 O-O.
IDeA treats these identical positions as separate positions requiring separate trees!
There are several problems with this:
1. When 'tree1' and 'tree2' are both explored, much analysis time is wasted.
2. When only 'tree1' is explored the user will see little or no info under 'tree2' when in reality alot of info is available.
Chess Opening Wizard (Bookup) handles transpositions gracefully, so there must be an engineering solution to this problem.
IDeA treats these identical positions as separate positions requiring separate trees!
There are several problems with this:
1. When 'tree1' and 'tree2' are both explored, much analysis time is wasted.
2. When only 'tree1' is explored the user will see little or no info under 'tree2' when in reality alot of info is available.
Chess Opening Wizard (Bookup) handles transpositions gracefully, so there must be an engineering solution to this problem.
Worse than that is that it doesn't recognize repitition of moves, I hope they can find a way to incorporate a repitition of moves checker into IDeA.
![[gr]](/mwf/v12/flags/gr.png "Greece")
It does recognize repetition of moves.
A position already in the tree is never analysed twice.
Now if you reach a position an with a move you are going back to the previous state, this is what a tree does
What do you mean it doesnt recognize repetition?
Can you be more specific?
A position already in the tree is never analysed twice.
Now if you reach a position an with a move you are going back to the previous state, this is what a tree does
What do you mean it doesnt recognize repetition?
Can you be more specific?
> Chess Opening Wizard (Bookup) handles transpositions gracefully, so there must be an engineering solution to this problem.
The main difference is "runtime" & while the engine is running.
There is certain "convention" , that if a move reaches a transposition, the number of positions displayed is 1.
This is done to avoid adding twice, and actually much more than twice , each position.
There is no algorithm to update the whole tree at O(N) time , meaning analogous to the number of nodes in the tree.
There is also no algorithm to eliminate cyclic paths (transpositions) other than brute force.
And i dont mean in the program. I mean in the graphs theory (which is a mathematics subsection).
The solution Bookup uses , is possibly what Vempele proposed, which works (does not update the number of positions on the whole tree but just those that you see), but i am not sure how fast it will be when there is an engine running and the tree grows considerably.
If we were talking about a tree that is not updated continuously then things will be different
If that procedure does not choke the whole IDeA process, then ok, but if i had to choose performance or the transpositions problem, i would definitely choose performance.
Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill