< Earlier Kibitzing · PAGE 65 OF 65 ·
|Aug-28-14|| ||jepflast: <YouRang: > (In reply to Hugin)|
<I don't follow your reasoning.
For one thing, the AT is merely a much more convenient place to store analysis. I can do analysis on positions that we might not reach for months. If we do get to that position months from now, that analysis will be readily available. If I had stored that analysis in some chessforum, nobody would go back and find it. Even I would have forgotten it was there!
The AT doesn't stop people from injecting human moves, and it doesn't force people to dryly follow engines.
If you don't like infinity analysis, then the AT should be your friend, because it encourages deep sliding analysis. I never vote for a move based on infinity analysis -- I vote based on what I've seen at the conclusion of several deep lines in the AT, with consideration also given to the arguments given by the folks I consider to be top-level human analysts.>
|Aug-28-14|| ||jepflast: <AylerKupp: <<Hugin> Based on years of experience i have to conclude AT is collective counter productive for the team to reach its potential strongest play>|
Interesting that you say that. I think that it is based on the AT <replacing> the Forum System, not when both the AT and Forum System are actively used. The Forum System is, as you indicated, very valuable for exploring ideas and discussing them. But the AT is very valuable in terms of collecting all the ideas in one place, something that you would need to repeatedly go to each Forum to get the information, and then it would be hard to compare different lines. So, ideally, both the Forum System and the AT should be used, as they serve different purposes. I think that this is one case where the whole is greater than the sum of its parts.
Maybe one problem is the way we're using the AT. My current peeve is what I consider the large amount of totally useless junk on the AT. To give you an example: In the recent discussions as to whether some would vote for 15.Bc2 over 15.Nxe4 based on the greater amount of analysis in the AT for 15.Bc2, I counted the number of "lines" both moves and 15.Bc2 had 151 "lines" while 15.Nxe4 had 129 "lines" (yes, I am a glutton for punishment).
But of those 151 "lines" for 15.Bc2, 67 of them (44%) consisted of single moves not qualified in any way (no comments, no evaluations, no move categorization other than the default "Candidate"). For 15.Nxe4 of the 129 "lines", 52 of them (40%) were also unqualified single moves. I didn't look at all of them (I am <not> that much of a glutton for punishment!) but for the dozens that I looked at not a single one had any additional information.
What purpose do these single move "lines" serve other than cluttering up the AT? The only "information" they provide is indicating that the single move is legal in that position. I think that all of us (well, at least most of us :-) ) know enough about chess to differentiate between a legal move and an illegal one, although given the typos that I often make I might be an exception.
So, <jepflast>, I would suggest that you think about modifying the AT so that no information is accepted (or at least not displayed) unless they contain some added value. That would require that each line contain either (a) one move with non-blank comments or (b) one move categorized with other than the default category, whatever that might be.>
|Aug-29-14|| ||jepflast: And now, from the Department of Tree Apologetics:
I feel like I've done a good job responding to the way the community wants to use the Tree, making improvements while keeping it flexible for different kinds of use, even if the process is slow. And I try to keep it out of the way of those who don't wish to partake.
So I'd like to respond to every perceived shortcoming, not only to make the Tree an ever-sharper weapon, but to maintain synergy throughout the Team in general.
That said, I don't quite understand some of the criticisms, but I'll try to answer them anyway. If I'm guessing wrong, please let me know.
(Hugin) <Based on years of experience i have to conclude AT is collective counter productive for the team to reach its potential strongest play.....>
I really doubt this. It's true our record without the Tree is better, but it was an era when we had weaker and less well-prepared opponents, more participation and enthusiasm, and probably a greater number of really strong/dedicated players on our team. A lot of us on the World Team got weary of all the hard work, I think, and just quit playing. And there is generally less excitement now, and fewer people are actively participating. To me, this is obviously why the forum system broke down. We just don't have the manpower. But we should try to drum up some enthusiasm and get it back.
(Hugin) <Now it turned into a dry boring and faulty way of analysis and don't you dare opposing anything in the AT attitude.>
I don't get this one. First, I don't sense this attitude; and second, the Tree is merely a reflection of the Team's analysis. It would be the same, Tree or not. I can only guess that Hugin is tired of being reminded of the "main line", but this would happen either way.
(kwid) <The problem I see with the AT is that the team will accept the data without looking over the shown horizon unless the lines have been run out as some analysts are trying to do.>
Again, I do not see how the Tree makes a difference. Analysis may be posted anywhere. People can accept it, extend it, or look for improvements no matter where it is.
(kwid) <But we need a main line indicator or a reference line right from the first move on. In other words show the theoretical best line with its historical projected end result.>
That's a good thing to post on the Tree. With appropriate comments, preferably at beginning & end.
(AylerKupp) <My current peeve is what I consider the large amount of totally useless junk on the AT. ...
What purpose do these single move "lines" serve other than cluttering up the AT?>
These are just places where someone has recorded a move along the way that they need to go back and try. For me, this is one major point of using the Tree: keeping track of the moves you have not tried yet.
|Aug-29-14|| ||jepflast: I'm thinking of changing the move ratings to better reflect how we use them. We probably don't need, for example, both "Probably Inferior" and "Inferior". So how about this:|
3. Probably Best
4. Prime Candidate
6. Fringe Candidate
8. Probably Inferior
And the default for every move would be "Suggestion". Thoughts?
|Aug-29-14|| ||YouRang: <jepflast><(Hugin) <Based on years of experience i have to conclude AT is collective counter productive for the team to reach its potential strongest play.....>|
I really doubt this. It's true our record without the Tree is better, but it was an era when we had weaker and less well-prepared opponents, more participation and enthusiasm, and probably a greater number of really strong/dedicated players on our team.>
I recall that the AT wasn't used as much (or at least not as effectively) in the first few games in which it was available. It's one thing for people to put analysis out there, but it's another to get people to look there and make use of the analysis there. It's a bit annoying when the anti-ATists post lines for which refutations are known in the AT.
However, AT usage has steadily increased, and I think that this has contributed to our recent win streak.
I can't imagine why it wouldn't contribute to success. It stores analysis in a more accessible way, and it reduces duplication of effort. Overall a great tool for collaborative chess analysis. Some of the arguments I've seen against it really make no sense. As you've pointed out, they refer to problems that have nothing to do with the AT itself.
BTW, I like how my new AT link automatically finds the latest move. Thanks for that. :-)
|Aug-29-14|| ||YouRang: BTW, I'm happy with the current selection of categories. But your proposed changes are okay. |
I might suggest that the bottom-most category be "delete". Clicking it will delete the move (equivalent to entering the "delete_node" URL, but easier).
|Aug-29-14|| ||Nickster: Great job <jepflast>. As I've mentioned before I'm a very big fan of the AT. For me it's indispensable. The only improvement I could think of would be a way to detect transpositions. But I understand that this may be impossible to implement.|
<YouRang: I might suggest that the bottom-most category be "delete". Clicking it will delete the move (equivalent to entering the "delete_node" URL, but easier).>
One of the rare disagreements with <YouRang> here. I believe that the 'delete' option must be kept deliberately obscure as to avoid some mischief. I'm sure that the majority of the <Team> members are trustworthy but there's always that one bad egg... No need to make it easy for them and erase hours (days) of work.
|Aug-29-14|| ||YouRang: <The only improvement I could think of would be a way to detect transpositions. But I understand that this may be impossible to implement.>|
The thought I had there was this: When adding a new move, the app would search the tree to see if the ensuing position is already in the AT somewhere. If so, then those matching nodes (and the new one just added) will be updated to include a list of all matching nodes.
The matching nodes could be shown somewhere on the page (perhaps below the board diagram) as a clickable link.
I realize that there are different ideas about what "matching" means. I think it would be fine to say two nodes match if their FEN strings are the same.
No idea how feasible or difficult that would be, but it wouldn't seem to be a huge structural departure from the current workings.
It would be very useful and illuminating to know if the new position you've arrived at has already be reached in some other line(s).
|Aug-29-14|| ||jepflast: Thanks for the support.
I think that the deleting nodes thing is not a huge problem; I only have used it twice, I think, this game. I will probably wait on that one.
Detecting transpositions would be nice. And it's not impossible, but the Tree does not currently store the FEN or position for each node. I'm actually calculating it each time by playing all the moves up to that point when you load a node page.
So this could not really be changed mid-game, but I may investigate the difficulty at a later date.
Any more thoughts on revamping the move ratings? This is much easier to change.
|Aug-29-14|| ||jepflast: <YouRang><I realize that there are different ideas about what "matching" means. I think it would be fine to say two nodes match if their FEN strings are the same.>|
Yeah, this is what I have in mind. Of course, some positions will transpose at a different move number, and some with a different move number but same number of reversible moves. But this can all be accounted for.
Maybe I can use the FENs to build the current position and save time when loading the page too.
|Aug-29-14|| ||Boomie: <jepflast: I'm thinking of changing the move ratings... |
6. Fringe Candidate>
Perhaps "Lesser Candidate" would be clearer.
I live on the Fringe and don't see that as a bad thing...heh.
|Aug-29-14|| ||mistermac: I have an idea that may prove useful. A new type of "fen" or what I will call a position skeleton, PS from now on. Now, because of its design, an individual PS cpild be ambiguous, as will become evident, but that can be solved. |
The original position of the board would have a PS: /11111111/11111111/00000000/00000000/00000000/000-
which could be denoted as 255,255,000,000,000,000,255,255 the decimal value of the eight ranks, separated by a comma. The ones denote occupied by a piece, (that is any chess entity from pawn to king is denoted by a one) the zeroes are empty squares.
Being numbers, even though each number my be ambiguous, but I suggest seldom, these PS values would be sortable. Thus any node in an AT would have a PS value, and thus would be sortable. Any two nodes with the same PS value would thus probably be a transposition. Amgiguities would probably be a problem with reduced material on the board, as they would be more likely, but at that stage of the game, probaly the endgame, the need for the PS would be reduced anyway.
I will not write any further at this stage, but any intelligent reader will probably grasp my idea, and then set about ripping it to shredsas useless ot impracticable.
But, in my days as a Systems Analyst, and go to man when impossible problems,(the simple problems I could never solve), were presented for a solution, they were often taken up, even if considerably modified.
|Aug-29-14|| ||YouRang: <jepflast> One other thought I had: What do you think about sorting the list of selectable moves alphabetically? |
I think they are currently "sorted" based on some sort of board location, but usually I come thinking "I want Nf4" and it takes me a few moments to find it in the list (especially if we have two N's on the board). I think if sorted alphabetically, I'd find it much faster.
I know that I can click on the dropdown list and start typing "Nf4", but an all-mouse interface is more pleasant than mouse+keyboard.
BTW, I mentioned the "easier delete" feature because I have, on occasions, made the mistake of selecting and adding the wrong move. Typically, I realize my mistake a few moves later, and I want to clean up my mess, but the current method of editing the url, goback, refresh, etc. is cumbersome (intentionally so, I suppose). However, perhaps sorting the move list will cut down on mistakes to begin with. :-)
|Aug-29-14|| ||AylerKupp: <jepflast> If you think that these single move lines are the single most helpful aspect of the Tree all I can say is that you and I have some fundamental differences on the best use of the Tree. But that's OK, it's your software package and you're entitled to have it work the way you think it should work.
But perhaps you'll agree that in the example I gave earlier, 151 total lines for 15.Bc2 with 67 single move lines (44%) and 129 total lines for 15.Nxe4 with 52 single move lines (40%) means that the concept is simply not working. To have that many single move lines so close to the voting deadline means that they would likely not be analyzed at all.|
Perhaps it's just too easy to add this kind of information to the AT. Maybe requiring the poster to enter <two> kinds of information for a single move line, the move itself and a categorization, would force the poster to be more discriminating about the single move lines that are added and only enter what they consider to be the more important ones.
Or maybe another entry in the pull-down menu that says something like "Look into this" (and a suitable highlighting color) might draw more attention to the single line move and highlight the fact that it should be looked into further. And perhaps whoever is looking into it further can include a comment to that effect so that there is no duplication of effort. I'm sure that there will be lots of other single move lines to look into!
|Aug-29-14|| ||AylerKupp: <MuzioFan> I think that you are being a little harsh on single moves.>
That's possible. And, as you see above, <jepflast> educated me. And he clearly indicates the intent of having a single move in the AT Instruction page. But, supposedly being knowledgeable about the AT, I had not looked at the Instruction page recently.|
But don't you think that if the poster of the move/line meant to say was that the that he should so indicate in the Comments field? Does that seem like too much to ask? (but see my suggestion about another entry in the move categorization pull down menu).
I'll bet that if these single move lines without any amplifying information were banned that there would be a more conscious effort to only enter meaningful moves and lines and request that they be looked into it further.
|Aug-29-14|| ||YouRang: <AylerKupp> I'll often run a multi-pv run on a position in order to get a feel for how many likely replies that our opponent has at that position. Then, I'll add a node for each position. |
As time permits, I'll do further analysis on each of those added moves.
However, if even I don't get around to exploring all those moves (and they remain single move lines), I figure that it's a benefit for viewers of the tree to have a feel for which moves are likely (i.e. no alternatives) and which are unclear (two or more alternatives shown -- even if no deeper analysis given).
Perhaps it's a worse crime to show a long line with no alternatives, as if the whole thing is likely, when in fact our opponent has so many ways to deviate that the line is almost a fairy tale.
But I recognize that there are different styles, which is to be expected in a situation like this.
|Aug-30-14|| ||Boomie: <YouRang: <AylerKupp> Perhaps it's a worse crime to show a long line with no alternatives...>|
My pet peeve is a long line with no comments. What are we supposed to think about that? Don't you have an opinion?? At least sign it, you gutless wonder! Geez!
|Aug-30-14|| ||Boomie: <YouRang: <AylerKupp> Then, I'll add a node for each position. As time permits, I'll do further analysis on each of those added moves.>|
Perhaps a Q and D way would be to copy the eval and line to the comment box. Then when you get back to that node, you will hit the ground running.
|Aug-30-14|| ||YouRang: <jepflast: I'm thinking of changing the move ratings to better reflect how we use them. We probably don't need, for example, both "Probably Inferior" and "Inferior". So how about this:|
3. Probably Best
4. Prime Candidate
6. Fringe Candidate
8. Probably Inferior
And the default for every move would be "Suggestion". Thoughts?>
Personally, I rather prefer having the default of "Candidate". I think it's safe to say that nobody deliberately adds moves that they don't think are legitimate candidates. Also, the distinction between a Candidate and a Suggestion seems unclear.
With the proposed changes, I think what will happen is:
(1) Some lazy users will add a bunch of candidate moves to the AT as Suggestions, and
(2) Other users will be annoyed that they have to take the extra step to upgrade the move to Candidate every time they add a move.
I agree that we probably don't need too many sub-Candidate options. Really, I think just two may be sufficient:
-- "Lesser Candidate" - A former Candidate move for which deeper analysis shows it to be worse than other viable Candidate moves.
-- "Blunder" - A Candidate that was later determined to be unplayable, or maybe a move that was added in error (and the user didn't know how or didn't bother to delete it).
3. Probably Best
4. Prime Candidate
6. Lesser Candidate
But again, IMO the categories are okay as they are.
|Aug-30-14|| ||jepflast: <Boomie: <YouRang: <AylerKupp> Perhaps it's a worse crime to show a long line with no alternatives...>|
My pet peeve is a long line with no comments.>
Yeah, we need comments at the end at least, so we know which side needs to improve.
|Aug-30-14|| ||jepflast: <YouRang> Thanks.|
|Aug-30-14|| ||YouRang: <Boomie><jepflast> A long line without comments isn't so bad if the moves are labeled "Best", "Probably best", or at least "Prime Candidate". After all, the status label is sort of a quick "comment".|
A long string of comment-free "Candidate" moves certainly has dubious value.
|Aug-30-14|| ||YouRang: <jepflast> You may regret soliciting comments about the AT, but I have another little suggestion for the wish-list:|
There have been times that I wished the AT had a "Search Comments" feature. For instance, I could search for "yourang" and it would give a list of all nodes (perhaps in most-recent to oldest order) that include "yourang" in the comments.
Of course, I could search for contributions from other specific analysts, or even dates or other keywords (e.g. "mainline", "hocus").
Ideally IMO, with 3 options for search scope:
1. Only tree at current node (default).
2. Only tree last-played node.
3. Entire tree, including past nodes and for moves not played, etc.
|Aug-30-14|| ||Chatu Ranga: Hi <jepflast>,
the analysis tree is great !
Not only that we can find out where more analysis is needed - it is also excellent for comparing similar lines:
One types in Ctrl + F "Bxe3" (for example) and can find every node (and that is: every line) in which this move is played.
|Aug-31-14|| ||jepflast: <YouRang: <jepflast> You may regret soliciting comments about the AT>|
Sounds like you just want to make me to go digging through the MySQL syntax documentation....
<Chatu Ranga> That's an interesting observation!
< Earlier Kibitzing · PAGE 65 OF 65 ·
A free online guide presented by Chessgames.com