Note that CG was a bit wishy-washy on that idea: chessgames.com chessforum.

Given the lukewarm response from CG, I have the same reservations as previously voiced by <sentriclecub> chessgames.com chessforum

Still, you have added much more flesh to the basic idea, so maybe they will see the merit of doing it right and here I disagree with <sentriclecub>'s idea of doing it independently from CG -- I'd vote to give them complete software specifications and let them do the implementation.

Minor coments:

1) The interface should include the ability to upload the entire line rather than branching the nodes manually, or even a small analysis tree in the same format as produced by ChessBase. The software can detect automatically where to branch.

2) I disagree with your idea <that the kibitzing area be located below, but on the same page as the diagram for that node>. This is the WORST feature of the CG's page layout. Give me the diagram on the left, and comments on the right any day. There are many websites that display a "floating" sub-window that stays on the same position in the browser window while the rest of the page can be scrolled up or down -- normally this technology is used to display some annoying ads, but it could be used in a constructive manner, too.

I fully support your idea. The current system is "clunky" because of the volume of analysis and the difficulty of navigating to where the sought after anallysis is. Unfortunately, CG.com's "Search" function is not up to the job of finding moves.

I understand <sentriclecub>'s reservations, and it's slightly ironic that he is now advocating slower rollout, but (and without invalidating <sentriclecub>'s concerns in the slightest) I'm inclined to go with <whatthefat> on this one now - the sooner the better.

Nevertheless, as MAJ pointed out, there are some serious technical hurdles to overcome, not the least being reconciling the different evaluations being produced by everyone's engines, some of which are in clear contradiction.

A good AT is something the Team will adapt to quickly and easily. However, it will need frequent reminders to all players in the Team, both on the main thread and in the forums, to encourage people to take it up.

------

Initial situation: There are 5 moves (10 plies) already played in the game.

First modification: in the game position, someone analyzes a move that nobody considered yet (ply 11) and posts additional 20 plies of analysis, advancing the (potential) game status to 30th ply.

Second modification: someone else follows the above line for 10 plies and branches there, continuing for another 10 plies.

------

You can visualize the resulting addition to the tree as a letter Y, with each arm/leg/branch 10 plies long.

I see two ways of presenting the analysis:

1) If each ply produces a new tree node, navigating the above would require 40 clicks: 20 to get to the end of one arm of the Y, 10 to backtrack, 10 to get to the end of the other one. Very annoying, IMO.

2) A more user friendly version would have one page at the bottom of the Y (ply 11) indicating that the next 10 plies are common to all resulting analyzed lines, and another page at the branch in the Y (ply 20), indicating that there are two variants from now on. The variants and the associated comments would be also included there. This should be it - any more pages, and the navigation takes too much time.

Where should be the comments to the analysis atttached? At the beginning of a line, or at the end?

My suggestion: at every node in between, indicating clearly the origin of the analysis.

Consider again the Y-shaped analysis starting at ply 11 and branching at ply 20, which I described earlier.

Looking at the node for ply 20, one would see the post added in the first modification, along with a sidebar comment, indicating that this post was inserted at ply 11. The same would happen with the post added in the second modification.

In the arms of the Y, only the posts associated with the lines containing these branches would be shown.

In essence, each node would show ALL posts that included that node in the analysis, perhaps ordered from the most obsolete to the newest. For example, looking at ply 30, one would pay less attention to a post inserted at ply 20 than to a post inserted at ply 28 (accompanying a fresh analysis).

1. Timing - I don't recall that <kutz> specified <when> he wanted the AT to be implemented. I have always assumed it would be after the GMT game. Like me, he doesn't like switching boats midstream. The team might embrace the AT quickly and transition from a forum-based system to AT-based, or both. But this is not within our control. I suspect that CG would take 3-4 months to program and debug the code if they gave it even modest priority.

2. Storage requirements - posts are just text and do not require much storage. I would worry more about the database that links everything together. They use MySQL which should be adequate, but I'm not sure.

3. "Anti-trolling" features - I mentioned this to <kutz> and in fact gave the exact example of Qf8+ given by <lonepsycho>. I suggested only Premium members be able to create nodes. As <whatthefat> points out, the sticky works fine. But I will point out that only Premies can edit it. I don't want to sound elitist or suggest that Premies cannot be trolls ... having an extra $25 does not guarantee intelligence or moral fiber. However, it might be a selling point for Premium membership from CG's perspective.

<reconciling the differences between different engines> No, that kills diversity. Include all engines and when you are at a node, you can "sort by rybka" "ascending" can be an option. Or you can sort by fritz, or you can sort by "ply-depth" "descending".

<I suggested only Premium members be able to create nodes> then we lose Rinus, who has been one of the greatest contributors in the last few days.

<lastly> I have read MAJ and Zanshin's posts and I agree with all that they say.

1) Regarding the suggestions <kibitzing area be located below, but on the same page as the diagram for that node> vs <floating window>: take the page layout suggestions out of the proposal. This is not a primary concern for the analysis tree implementation. I'll resubmit the floating window idea as an unrelated enhancement (btw, the layout schema should be selectable in user's preferences rather than being imposed on everyone).

2) Evaluations are quite important to have; they are not just bells/whistles. <sentriclecub> had a good idea in his last post about handling them. BTW, by "reconciling" I did not mean "averaging" or anything else that might be seen as a diversity killer, but rather some way of presenting the evaluations in a sensible and understandable manner. My bad for being unclear.

3) Entering entire lines is important, IMO, to provide ease of use and navigation. Would you rather copy/paste entire line or enter it move-by-move?

Let me first address the desired timing of implementation, if the idea is accepted by cg.com. I agree with <zanshin> that it would take some months to program and test the new feature. I seriously doubt that we will see the AT in the current game. My hope was that we might have it for the start of the next game.

<MAJ>, you made some interesting comments about navigation within the AT. I am sympathetic to your point about avoiding an excessive number of mouse clicks. Perhaps we could request a few reasonable navigation features as follows:

1) If I remember correctly, OE provides the ability to go backwards to any point in the line by simply clicking on the move. Perhaps AT could work the same way.

2) Moving forward, we could request a button to move forward to the next branching point which would help in certain situations.

I am intrigued by these ideas, but aside from potentially increasing programming complexity, think about this: Do we really want team members being able to easily enter an entire 180-ply game with a few mouse clicks? Let's think about this. I am not adamant about this, but I lean toward the one-node-at-a-time approach. I want team members who have something to say about the position 20 nodes out to be the ones creating the nodes. Someone who spent hours analyzing a line and has interesting conclusions to report should not mind a few extra minutes to create the nodes.

The inclusion of engine valuations as part of the AT has been mentioned a few times. Let me clarify what my intent was. I certainly think that engine evaluations should and will be among the most important kibitzes that will be posted at the various nodes of the tree. In fact, I would be very disappointed if team members would not post evaluations from all the different engines we use.

HOWEVER, it was not my intent to request any kind of special fields to dispay engine evaluations. It was also not my intent at nodes where there are multiple candidate moves entered on the AT, to attempt to somehow rank the moves.

I would hope and expect that one of the first kibitzes at every node once we are out of the opening would be an engine evaluation, which WOULD rank the moves. This would hopefully be followed by evaluations from other engines.

Let me address my last post of the afternoon to you. I am very pleased that despite your initial misgivings, that you are now in support of this proposal. As I mentioned to you over a month ago, I support the idea of an analysis tree. You are now clearly seeing the depth of my support and also that many others on the team support the idea, if properly implemented.

This is correct. When you are traversing down an opening line, the moves are created at the top and you can click on a past move to take you to the relevant point in the OE. However, once you do so, all moves after that are erased. I guess this is not a problem when you are exploring an opening, but in a game, the moves become set afer they have been made. So you should be able to traverse forward as well as back.