New tag suggestion

[stormcrow]

New tag suggestion - 'Desparado' as per problem 13884. A piece that once it starts attacking, must continue attacking.

[tmr]

Is this truely desparado (which is defined slightly differently in wiki - "In chess, a desperado piece is a piece that seems determined to give itself up, typically either (1) to sell itself as dearly as possible in a situation where both sides have hanging pieces or (2) to bring about stalemate if it is captured (or in some instances, to force a draw by threefold repetition if it is not captured)")?

The problem appears to more of white's under protected pieces.  A white pre-move of Nxc5 (vs Ba3) would have avoided this and white's dark bishop is never lost.  This makes it not desparado, I think.

[newlook]

How about mating attack and/or mating threat as new tags? I see many problems that involve mating as a theme but don't involve any other tactic on the list. The king is obviously vulnerable, but not really exposed. If there is a sacrifice, I will put that down, but there isn't always. Sometimes it's just a simple case of getting your queen, with protection, right next to the enemy king.

Another theme is insufficient defense of material (maybe not a standard term). The piece isn't hanging en prise. It's defended, but not enough to counter the number of attackers bearing down on it. Or the attacker(s) is not enough compensation. Maybe another tag would be something like "winning the exchange."

Another tag-related suggestion is to provide a page with definitions of the tags. A diagrammed illustration of each would be a nice bonus.

[richard]

Hi Newlook,

I'm about 2/3rds of the way through putting together a page giving definitions of each of the tags.  I have about 2-3 examples displayed in the pgn viewer for each motif/tag.  While putting the page together I've noticed that a number of tags are probably duplicates and some others overlap, for example removing defender/undermining and double attack/fork.  One of the issues I've faced is that not everyone agrees on what 'tactical motif X' actually describes.  Some authors use a different term to describe the same thing, others use the same term to describe different things. I haven't decided how to deal with these issues yet, my current plan is to get the definitions page up and use that as way of generating discussions on the tags to get a feel for other's opinions on the tags before making changes.

I like your new tag suggestions, but I'll probably wait until the tag descriptions page is up for discussion to make any further tag changes. That way I can incorporate any other suggestions into the tag update.

I've had to defer further work on the tag descriptions page in order to get the new generator and alternative moves display finished.  The new generator is now running, but I still have a few bits and pieces to do on the move display.  When that is done I'll finish off the tags descriptions page.

Regards,
Richard.

[boromir]

I would also like to see 'double check' added to the tags list. It's actually relatively common in the problems. I also think you should rename 'fork' to 'knight fork' to make clearer the difference between the fork and the double attack (at the moment they seem to be treated pretty much as interchangable by the users).
At the moment the tags system is not working very well, as there are a lot of problems that are mislabeled. I have made a handful of searches on specific themes and every time I just gave up using those sets as I had to discard a lot (up to 25-50%) as mislabeled (and went back to Reinfeld, etc.  ).
A definitions page might help, but I don't know how many people will actually read and adhere to it.

[drahacikfm]

Yes, double check would be useful.  Right now I have to label those with Double attack and discovered attack.

When you see a wrong label, vote against it.  I do that a lot.  Each time you do it, it makes the label system better.

Maybe the system could keep track of what percentage of each user's labels got voted against.  When a user has more than some percent, such as 30%, then he is no longer allowed to add labels.  To remove the ban, maybe there could be a test with 10 problems and you have to label them correctly.

[richard]

Hi boromir,

The problem with the terms is that even established chess authors can't agree on terminology.  Fork and Double Attack is a good example.  Some authors are happy to call all double attacks a fork, some only like to call forks a double attack with a knight, others are happy to include pawn or king double attacks in "fork" group. So here users are not really wrong but reflecting which part of the chess literature they received their definitions from. It is not really feasible to rename fork to "knight fork" now as that will make that tag even more wrong for all the non-knight forks already in the system.  I think the best plan may be to merge fork and double attack and call them Fork/Double Attack as for lots of people they really are the same thing (it is generally the same theme the only argument is about which pieces they involved). I can then add new tags for "Knight Fork", "Pawn Fork", "Bishop Fork","Queen Fork" etc (or Fork for only the first two and "Bishop Double Attack" etc for the others.  All the old tags will remain (mostly) accurate (and allow for looking at the general case of double attacks/forks) and the new more specific tags will allow users to look for just one kind of fork/double attack.

Another problem is the difference between a motif that plays out and a motif that by the mere threat wins material without being played out,  I find it useful when users have tagged these, but sometimes at a quick glance they can appear like incorrect tags until you look more closely at the problem.

Then there is the simply wrong tags.  Some users have their own idea of what a tag means that makes sense in their own mind, but don't relate to any usual convention of identifying tactical motifs.  My hope is that eventually users will vote against the wrong tags and they will be cleaned up.  This assumes that overall there are more users that can accurately identify the tags than there is users who make wrong tags , I think this is probably the case so far, but there has been a very recent influx of very bad tags which hasn't had time to get voted down yet (smothers that are not smothers and hanging piece when the piece wasn't hanging have been particular hard hit [If you can give details on the ones that you found particularly bad I can follow up on those also] - Up until the recent influx I found that tags were generally over 80% accurate and get more accurate over time). Some users are excellent taggers but have a misconception about one or two tags, I'm going to put in a mass "vote against" for some of these tag votes in the next couple of days which will clean up a fair number of mistagged items.

Another completely different approach is to try and develop some code which automagically tags problems as they are  created.  This is easier for some tags than others , but overall would be considerable new effort, so it is on the long term wish list for now.

Regards,
Richard.

[richard]

Drahacik: I've been thinking about the issue of "bad taggers" and I like your idea of taking action when a user starts to get to certain "vote against" rate. Rather than banning them altogether I could just stop allowing their votes to count (at least when tags are displayed to users other than themselves). If they improve then their votes can start to be used again.

I'll add the double check tag at the same time I make the next update.

Regards,
Richard.

[boromir]

Of course I vote against a tag every time I see a wrong one, but I would kinda like to have them right from the beginning .

A lot of the 'not-really-wrong' tags are definitely down to disagreement on the definitions.
I agree that fork and double attack can be interchangable, and I am quite happy for them to be dealt as the same. Then they should definitely be merged into one category though, otherwise it's just confusing for everybody. Having a different tag for a double attack with each piece might be a bit overkill, but I guess it could be useful in some practice situations. Then again, you could also argue for the same with a lot of other tags (skewer by rook, skewer by queen, etc.), which would definitely be too much.

I am not fuzzed about the exact definition of each theme, just as long as something is agreed upon for this particular site.

How about a little pop-up in the add-tag menu, so that if you hover your mouse above a tag a short definition shows or something like that. That would maybe help agreeing on common definitions.

Would it work if a tag had to be voted for 2 or 3 times before the problem is marked? that way, if one person got it wrong, it wouldn't affect the problem.

Cheers, Boromir

[richard]

Hi boromir,

I agree that generally categorizing by piece type is probably overkill but I think fork/double attack is probably a special case due to the tendency for people to differentiate within these two categories based on piece type much more so than any other motif.

The popup is not a bad idea, I could also make sure people had at least visited the descriptions page before doing any tags with a reminder whenever they try to tag that they need to read that page first. Translation of the tags might also help, I suspect second-language issues are often behind mis-interpretation of tag names.

The 2-3 tags threshold is also a good idea.  One of the issues is that I know myself that if I see a tag already made on a problem then I don't tag it again, I'm guessing most users behave the same way. This unfortunately means that it only takes one user to vote against it to cancel the accurate vote.  If I implement this I'd probably make sure the user got to see their own tags immediately but only saw others after the critical number was reached.

I've added both of the above to the todo list.

Regards,
Richard.

[drahacikfm]

Requiring 2 or 3 votes for a tag to appear will run into critical mass issues.  Right now I guess less than 20% of the problems have tags.  If you require 2 votes for a tag to appear, almost no problems would have any tags right now. And it might be a very long time before many problems get the required 2 votes.

I think a high percentage of the current tags are correct, so the solution does not seem to be to make it harder for a correct tag to appear (requiring 2 votes).  The solution is to make it easier to get rid of bad tags or taggers.  So when a user is not accurate, as measured by votes against, remove all his tags or at least stop him from adding more.

If you really want to require 2 votes for a tag to appear, you could count some users (me! ) as two votes whenever they add a tag.  Maybe include any user with a minimum 100 tags and with votes-against less than 3%.

[richard]

There is certainly  a bootstrapping problem for the 2 to 3 votes idea, there is currently 8202 tags on active problems and 6218 problems which are tagged.  This means that it is likely that there will be almost 0 problems which would have tags if this idea was turned on now.  The problem at the moment is that the tag votes are not really treated like votes, if someone sees a tagged problem  where they agree with the tag, they are clearly not tagging it again. As mentioned previously this does make tags vulnerable to being wiped out by just one incorrect vote against.

If the 2 to 3 vote requirement was implemented it would be possible to keep the current tags for the purpose of creating sets but not display the tag to the user until the threshold is reached, this would encourage users to keep tagging problems, thus (hopefully) being more likely to overwhelm any bad 'against' votes and hide any bad incorrect 'for' votes.

Another approach may be to show the current votes and have a thumbs up/thumbs down next to each tag so that it is more obvious that users are agreeing/disagreeing to an existing vote, I think this is more likely to get votes than hoping users vote "for" a tag after already seeing it added.

Weighting high quality taggers is not something I'd considered, but it is similar to ideas I've had for the problem quality ratings.  Generally speaking higher rated users are more likely to accurately assess problem quality than lower rated players so I think extra weight might might be assigned for higher rated players (I have a non-public page which allows me to assess quality ratings by separating star ratings into high rated player and lower rated player ratings. For those who were wondering the higher rated players tend to give higher quality ratings to problems (but averaged across all problems only about a quarter of a star more) - I guess higher rated players are less likely to incorrectly blame mistakes on the problem instead of incorrect analysis).  Similar ideas could be extended to tagging where a combination of user rating and negative votes (to catch the high rated users who have an idiosyncratic definition of a certain tag).  The main thing I don't like about this is that I don't like the anti-democratic overtones , i.e. the feeling that it creates a "all users are equal, but some are more equal than others" situation.  I guess one response to that is that chess ratings are based on the idea of meritocracy and it is just a fact of life that higher rated users are able to make higher quality tagging/quality decisions and that it would be silly not to utilise that fact to improve the site :-) 

As you can see I'm still in the "roll this idea around" stage at the moment :-)

Regards,
Richard.

[newlook]

You could have training/calibration exercises for tags. For example, a user would be shown 10 problems from a pre-tagged-and-verified set and have to identify whether each is a pin, a fork, or a skewer. Getting a perfect score would qualify the person to use those three tags. Or you could make it a voluntary system, with the idea that most taggers are motivated to get it right and only need (and will welcome) information and feedback.

[dlester]

I try to tag as many problems as I can, and I am sure I don't do them all correctly.  But, when I look at the list there are some I just have no idea what they are.  Below is my understanding of the tags we currently have to choose from.  My understanding is far from absolute gospel.  Others will have a different understanding of several of these and I think it would be great if people replied openly with their comments.  Maybe it will give Richard some ideas for improving the tags we have to choose from and it might even develop into a framework he can add to the site that will define what each tag means here, thus we would all know for sure what was intended and use them accordingly.

Note:  I use 'piece' a lot to describe particular motifs, but an attack doesn't necessarily have to be on a piece.  It could be any kind of threat.


Attraction:  I honestly don't know what this is supposed to be.

Back Rank Mate: This one is pretty obvious, though I have used it on some problems where a back rank mate didn't actually exist, but some advantage was obtained due to the threat of a back rank mate.

Blocking:  I haven't seen this on any puzzles, probably because I understand it to be a defensive tactic and there aren't any defensive puzzles.

Clearance:  I don't know the distinction between clearance and removing the defender.  When I am tagging, and a piece is captured or forced to move to clear the path for an attack I will nearly always end up tagging it removing the defender.  If I have used clearance it has been rare.

Decoy:  Much like clearance, it is just another flavor of removing a defender to me.  Clearance, to me, means capturing the defender.  Decoy means forcing it to move.  Either way, it is removed.

Discovered Attack:  If a piece of mine moves and I can now attack something else in the process, I consider that a discovered attack.  Of course, if the piece I move out of the way subsequently attacks another piece then I have two different pieces attacking two different other pieces and I understand that to be a double attack (which is different than a fork).

Distraction:  I have no idea how this is different than Decoy.

Double Attack:  Already defined in Discovered Attack verbage.

Exposed King:  I use this tag a lot.  If an attack on the king, or a threat of an attack on the king is any part of the tactical motif I throw this tag on there.  It is very generic, and since many of the puzzles do involve some kind of tactic relevant to king safety this one is almost always accurate.

Fork:  Whenever a single piece is attacking two or more enemy pieces.  Not the same as Double Attack.

Hanging Piece:  Whenever a piece can be captured for free, but I will also use this tag for instances where a piece can be exchanged with a piece of lower value.

Interference:  I have no idea how this is different than Blocking.

Overloading:  Whenever a piece is performing two or more roles, usually defensive, but those pieces are attacked (or can be attacked).  Obviously when the first capture happens the subsequent recapture will hang the other piece.  I would imagine it is difficult, if not impossible, to really make a distinction between an overload and a double attack since the only real way to exploit an overload is with a double attack.

Passed Pawn:  I have tagged a few with this one, though it isn't terribly common.  I do remember some puzzles where a pawn became a promotion threat, or the exchange of pieces made it impossible to stop a passed pawn from promoting and I used this tag.  I have never used it just because a pawn became passed as a result of the exchange.

Pawn Promotion:  I don't think I have used this one but maybe once or twice.  Making the distinction of when to use this versus Passed Pawn is somewhat sketchy.

Pin:  Pretty straightforward.  Whenever there is an enemy piece that is trapped between your attacking piece and a more powerful piece of its own and moving would result in losing the more powerful piece.  Sometimes the threat of a pin is as effective as actually having a pin.

Removing Defender:  Already mentioned in other descriptions.  This is the one I use generically for all of them.  If a tactic involves any means by which defenders are eliminated to enable an attack, I will usually tag it as Removing Defender and largely ignore the other tags that mean the same thing but may be slightly more specific.

Sacrifice:  Although technically this is also a flavor of Removing Defender, if the removal of a defender involves giving up a piece in what would otherwise be an unfavorable exchange then I will generally include this tag.

Simplification:  I haven't used this much, but I understand it to be a sequence of exchanges that are materially equal but result in a positionally favorable position.  The tactics engine does not really produce puzzles like this so I don't see puzzles where I find it applicable.

Skewer:  Similar to the pin, but with the value of the enemy pieces reverse.  Instead of the more valuable one being behind, it is in front.  When it moves to safety it leaves the less valuable counterpart to be captured.  Many people do not make the distinction between a pin and a skewer, and generally I don't see why it matters.  They are basically the same from a pattern recognition perspective.  I don't know why it can't just be a tag for Pin/Skewer.

Smother:  Arguably the most aesthetically pleasing way to end a game, I consider this to be mating the enemy king by having it trapped with no place to physically move.  This is a mate with a knight.  My son actually did one of these in his very first kindergarten tournament last year.  It was a good day.

Trapped Piece:  This is a piece that has no safe squares where it can move without forcing it to be lost either totally for free or for an exchange that is less valuable than the lost piece.  Oddly enough, checkmate is technically a trapped piece.

Undermining:  Another flavor of Removing Defender.  As with some others I don't know what the distinction is between this and Distraction or Decoy.

Weak Back Rank:  I have tagged quite a few with this whenever the back rank ends up being exploited simply because there are more attacking pieces than there are defending ones.  Does not always result in a back rank checkmate.

X-Ray Attack:  I think, but have no idea for sure, this is just a synomym for a skewer or pin. 

Zugzwang:  Any position where your opponent has to make a move, but no matter what they do it is going to result in a weaker position.  I don't think I have tagged any problems with this.  If we end up with endgame scenarios this is likely to be more common.

Zwischenzug:  I don't think I have ever tagged a problem with this, but there are certainly several where it is applicable.  This is when you have a series of moves in an exchange and you inject some in-between move in the sequence that doesn't otherwise alter the exchange, but results in an advantage for you.  Frequently this will be a check move in the middle that results in one extra capture, though it doesn't have to be a check move.

[richard]

Great post dlester,

I'll address some of them, quoting from my (not yet finished tag description page), but hopefully this can kick off some discussion from other regular forum readers:

Attraction:"Attraction or Decoying occurs when a player entices a piece to a square (often using a sacrifice) where it will later come under attack."
I've noticed some authors see this as a synonym for decoy, while others (you!) don't agree.

Blocking: "Blocking (or a blockade) occurs where one pieces is forced to a position where it blocks the path of another."
I think some users might get this confused with interference (I know I have :-) ), some authors do use blocking to refer to interference, I think I might change this tags name to blockade which makes it clearer.

Clearance:"Clearance is when one player forces a piece away from a square, rank or file (often using a sacrifice) to make way for another piece to utilise or attack the cleared path."
This is slightly different from removing defender in that it is "clearing a path" rather than "clearing a piece".

Distraction:"Distraction or deflection involves forcing the opponent to move a piece that was previously guarding important squares or pieces."
Sounds a lot like removing the defender doesn't it :-) ?  I think that Distraction might be moving the defender elsewhere as opposed to just "taking the defender". But this is far from clear to me.

Fork: "Double attacks (also known as forks) occur when one player''s piece attacks two opponent pieces (or important squares). The opponent can't counter both threats so loses material."
Your differentiation of fork's and double attacks is not uncommon, but I've definitely seen a lot of authors user the terms equivalently with some authors restricting forks to only knight and pawn moves.

Pawn Promotion versus Passed Pawn:"The threat of pawn promotion is often seen at the heart of tactical possibilites. Resources required to prevent the promotion are not available to help elsewhere on the board and this can be exploited to create tactical opportunities."
I've (tentatively) decided to merge these two. I agree it is not that clear when to use one over the other at the moment and user choice here seems to be quite arbitrary. I think the original intention was that pawn promotion should only be used if a pawn WILL actually promote (similar to back rank mate versus 'weak bank rank'), however this is not consistently applied. There are currently 0 problems in the set where the user actually has to promote (I don't have the UI support for choosing the promotion piece yet - I do have a 1 or 2 hundred promotion problems ready to go when I do have the UI support), but there are some problems where you can see promotion on the users side in the computer analysis line.
 
Interference:"Interference refers to the situation where a player interposes (or forces the opponent to interpose) a piece between an opponents piece or sqaure and the piece defending it. "

Undermining/removing defender: "(Is this a subtype or same as deflection?) Removing the defender occurs when the defender of a piece or square is taken or distracted allowing the previously defended piece or square to be attacked."
As you can see by my description I'm a little confused as to the difference between undermining/removing defender/distraction :-)

X-Ray Attack: "An X-Ray attack occurs when one piece attacks a square or piece through another opponent piece."
My description is not useful yet but it is definitely different from skewer (despite what my description may lead you to believe), I've seen a few people refer to x-ray as synonymous with skewer  but I think this is incorrect.  I find x-ray's much easier to describe with diagrams (which is what my descriptions page will do - well actually interactive boards so you don't have to visualise the steps in the tactic yourself). Have a look at a few in the problem search, there are a few with that tag that are just skewers but there are enough good examples that it should give you a better idea on the distinction (while I wait to come up with a description that makes them sound less like a synonym for skewers :-)  ).

Anyway as you can see I've managed to slip a few nasty synonyms into the current tag list and there are a few others that even the best informed users are not going to agree on.  In areas where authors are just using inconsistent terms I think the best we can do is just say, "Ok some authors say X  means Y, other say Z. For the purposes of tagging here X means Y.".  Hopefully your post can generate some discussion from people better informed than myself :-)

Regards,
Richard.