Most of the work on the Ceri line, which started with 17...Qf7, now falls into this category, since 17...Be7 was played. Ideally, we should copy it into the transposition that is still alive.

Even more work was "lost" via 17...Qf7 18. Nf1 Be7 19. <Ne3>.

I made myself a little utility to "transplant" entire branches whenever we need to do it. This can only be used after a move has been played that eliminates some things by transposition.

As long as we keep all of the work in one place, we can now transplant it if necessary.

Whenever a transposition comes up, we should make sure to place a comment that says something like "***Transposition: do not post analysis here; use node xxxx instead***".

For the current situation, I transplanted node 3242 to take the place of 4899. The latter is now unattached to the Tree, but you can still see it by specifying the node number in the URL. Anything novel that was in node 4899 is now "lost", but I don't think that was very much if anything.

Seriously, dealing with transposition is not an easy thing to do as I learned when I was trying to do so during my maintenance of our AT spreadsheet in other games. The main problem is that just because one position could theoretically transpose into another by a different sequence of moves, that alternate move sequence may not be either legal or make any sense. For example, sometimes the alternate move sequence may involve ignoring a check (illegal) or reaching an intermediate position that would leave a piece hanging (nonsensical). So this alternate move sequence, while it would theoretically be possible, would not (could not) happen during an actual game, and any resulting analysis associated with this move sequence would not be valid.

What I had to do whenever I detected a possible move transposition was to review the alternate move sequence to see if it made sense. Most of the time it did, but a sufficient number of times it didn't, and on several occasions there were alternate (better) moves by the opponent made possible by the different move sequence that would have invalidated the analysis. These situations were not always easy to see, and on several occasions I erroneously posted alternate but inaccurate lines.

The only practical suggestions I can make is to be conservative when transplanting branches and, unless it involves a series of forced moves by both sides, don't transplant a branch that is more than a few moves long. And always add a note at the beginning of each transplanted branch indicating this the branch is a result of a move transposition and list the original move order. That way at least whoever is following the branch is alerted to look for possible anomalies. And, if in doubt, leave the branch where it was. I felt that it would be better to lose the analysis than to introduce misleading lines into the tree that might make us vote for bad moves.

I'm skeptical as to the usefulness of the following:

http://www.jepflast.com/chesstree/a... (transposes to 24...a6 25. Ra2 Bh4 26. Rf3 Bg5)

http://www.jepflast.com/chesstree/a... (transposes to 24...a6 25. Ra2 Bh4 26. Rf3 Qe8 or 25...Qe8 26. Rf3 Bh4)

If you are still considering making enhancements to the AT, I would suggest the following:

~~~

(1) Detect Transpositions

Have a mechanism in place so that whenever a new move is added, it searches the Tree to see if the ensuing position (board setup and turn to move) already exists elsewhere in the Tree.

If the position is found, the newly added move move could show (perhaps right of the "Update" button) a link to the lowest-numbered node that has the same position.

Of course, this enhancement would help us reduce redundant analysis. However, I think the benefit might go deeper than that, and reveal subtle (and sometimes surprising) things about the game that we could not detect any other way. IMO, this would be the single greatest enhancement.

~~~

(2) Delete Move feature

There have been times when I mistakenly added the wrong move (e.g. Kg8 when I meant Rg8, etc). I would like to be able to remove the mistake.

So, I would propose that if a node has no subsequent nodes or comments, then a "Delete" option be added to its status pull-down list (after "Inferior"). Selecting "Delete" and clicking "Update" would then remove that node.

I understand that it might cause a technical problems to delete, say, node 1000 after someone added node 1001 elsewhere in the Tree. If so, perhaps this "Delete" button would really just hide the move from view, but keep the node number. If by chance that hidden move is re-added later on, it could really just "unhide" that node.

~~~

(3) Reserve Tenth move

Perhaps you've already addressed this, but I recall that each position has a maximum of 10 moves that can be added. This can be a problem if 10 moves are added, and the move played isn't one of the ten.

If this is still the case, I suggest that you only allow 9 moves to be added, and only allow move #10 to be added if its status is "Played".

~~~

(4) View comments from Tree View

<YouRang> Those are very good suggestions. During the last game I think I did figure out how to detect transpositions. When a node is added that is identical to an existing position, I was going to have the code add a comment to that node saying that it might be a transposition. Further, I could add a button to "swap" all of the continuing analysis from one of those nodes to the other. (If there's more than one identical position, well, I'm not sure what to do yet!)

I reviewed <YouRang>'s post and also like (1) and (4).

With regard to transpositions: this is clearly the thorniest issue, and the reason is that a chess analysis is actually not a tree at all but a directed graph:

http://en.wikipedia.org/wiki/Graph_(mathematics)

A singly connected directed graph is a special case, which may be displayed as a tree. A k-connected graph is a graph in which at least k nodes must be deleted to disconnect the graph into disjoint parts. A tree has k=1, but if there are transpositions k>1. Now we run into the problem that if k>1 the graph cannot be displayed in 2 dimensions without crossing connections. Pretty significant issue.

<hms123>, my friend who is very constructive>