Mate problems in the generator

[drahacikfm]

Richard, if I understand previous posts, you said you are working on a new generator with these changes for mate problems:

1) eliminate mate problems if a different move gives a large material advantage
2) eliminate mate problems if the starting position has a large material advantage

What about another frustrating type of mate problem:  the best move leads to mate-in-5, but other moves or lines lead to mate-in-6?  It's frustrating to work on a problem for many minutes, find a mate-in-6, and then get it marked wrong because there was a mate-in-5.

I think it's still ok to have these problems, but there needs to be a label saying that it is mate-in-5, so we know what we have to look for, and we don't do the mate-in-6.  You could also keep the problems in 1) above, as long as there is a mate-in-X label.  Should be easy for the problem generator to generate a label too, for mate problems.

[richard]

Hi drahacik,

The generator has both 1) and 2) but also
3)  longer mate elimination which kills most mate in N versus mate in N+1 moves.  I say "most" as I still let problems in where there is both a mate in 1 AND a mate in 2 as I believe that it's reasonable from a tactical vision point of view to expect a user to see the immediate mate. 

I considered allowing a partial credit implementation but decided overall it was probably cleaner to just reject all of the 1) , 2) and 3) problems.  I might consider revisiting this later if it looks like the new aggressive pruning  takes away too much diversity or difficulty from the problem set.

I'd also prefer to avoid having to label problems explicitly as while this does avoid some frustration from the users point of view it also makes the training that little bit more unrealistic. Hopefully with 1), 2) and 3) combined the need for explicit labeling will be greatly reduced.

Regards,
Richard.

[slacker00]

I  respect the concept of completely eliminating mate in X when mate in X+1 is available.   But, I'd be very happy if these problems simply stated what X was.  I mean, then I'd look  for that particular solution, and  if I found the mate in X+1, I'd know to keep looking.  I think the most frustrating part  is not knowing if there's a better mate possible, and whether I should look  for it or not.  Not to mention, finding a valid mate, only to find it's not the "best" mate.

[tmr]

I've experienced the frustration with these types of problems but I'd rather have this than a too easy or non-diverse problem set.  As an alternative, instead of eliminating the 3) problems, what about allowing more than one correct answer (such as both the mate in N and N+1)?  Might be easier to impliment than the partial credit approach.

If you have to selectively eliminate problems then I'd vote for 3) first as this is the most frustrating for me.  1) and 2) don't bother me as much since one should find the mate vs the material.  I suppose I'd eliminate the 2) problems before the 1).

I don't care for labels, they give too much away.

A few more feedback comments:
- I think it would help if the "Next Problem" button made a sound when you click it.  When the system is slow, I sometimes wonder if I've clicked it and on occasion have lost points because a problem is just being displayed when I click the button a second time.

- Similarly it be nice to have the column titles on the Stats page make a sound when you click it for confirmation (or some other indication that the system is refreshing the page).  Given that the update takes a while it is easy to think you haven't clicked properly.

- It would be nice if there was some indicator that your king is in check or that illegal moves were rejected (but didn't fail the problem).  While of course one should be aware of this, having such an indicator would be more life like.

- On illegal moves, not sure if these resut in a failed problem in all cases but I suspect they do.  I know I've slipped moving a piece and have failed a problem becuase of it (though I suppose it is possible that my slip resulted in a valid, but incorrect, move).  If illegal moves are indeed accepted (that is they result in automatic problem failure) rejecting such moves would be a way to avoid this and the previous issue (no indication of check).

- I think I saw one of your comments that a future enhancement would be adding row and column designations.  This would be a particularly helpful visual clue especially when doing a lot of consecutive problems and in positions with only a few pieces.  I know at least one occasion that I tried to move a pawn backward because I got mixed up about which side of the board I was playing from.  Again, my problem, but not very real life like (unless I was doing a buch of simultaneous games)

Thanks again.  Keep up the great work.

[richard]

Hi Slacker00,

Probably a decent compromise here is to add a non-rated mode where the number of moves are displayed, although at the moment it is likely that unrated modes will be subscription only.

It isn't that hard to cull all the X+1 mates, so after the next update there should be none of these in the set (other than the mate in 1's as discussed in another thread - and I could kill those also if required). The tricky ones are the material winning versus mate, I hope I've killed off most of these but I suspect a few may still be lurking.

I'm tempted to just release the new problem set now without the UI updates, however due to the huge number of long mates that have been killed, I fear dramatic rating inflation without the "don't show rating, tags, comments while problem is live" changes and I can't upload those changes without testing that and a number of other features.  I might release the non-subscriber updates first which will reduce my short term testing load and let me get the improved problem set out ASAP.

Regards,
Richard.

[richard]

I've experienced the frustration with these types of problems but I'd rather have this than a too easy or non-diverse problem set.  As an alternative, instead of eliminating the 3) problems, what about allowing more than one correct answer (such as both the mate in N and N+1)?  Might be easier to impliment than the partial credit approach.

If you have to selectively eliminate problems then I'd vote for 3) first as this is the most frustrating for me.  1) and 2) don't bother me as much since one should find the mate vs the material.  I suppose I'd eliminate the 2) problems before the 1).

I don't care for labels, they give too much away.

A few more feedback comments:
- I think it would help if the "Next Problem" button made a sound when you click it.  When the system is slow, I sometimes wonder if I've clicked it and on occasion have lost points because a problem is just being displayed when I click the button a second time.

- Similarly it be nice to have the column titles on the Stats page make a sound when you click it for confirmation (or some other indication that the system is refreshing the page).  Given that the update takes a while it is easy to think you haven't clicked properly.

- It would be nice if there was some indicator that your king is in check or that illegal moves were rejected (but didn't fail the problem).  While of course one should be aware of this, having such an indicator would be more life like.

- On illegal moves, not sure if these resut in a failed problem in all cases but I suspect they do.  I know I've slipped moving a piece and have failed a problem becuase of it (though I suppose it is possible that my slip resulted in a valid, but incorrect, move).  If illegal moves are indeed accepted (that is they result in automatic problem failure) rejecting such moves would be a way to avoid this and the previous issue (no indication of check).

- I think I saw one of your comments that a future enhancement would be adding row and column designations.  This would be a particularly helpful visual clue especially when doing a lot of consecutive problems and in positions with only a few pieces.  I know at least one occasion that I tried to move a pawn backward because I got mixed up about which side of the board I was playing from.  Again, my problem, but not very real life like (unless I was doing a buch of simultaneous games)

Thanks again.  Keep up the great work.

Hi Tmr,

Part of the problem with deciding how to improve the set is that different users have different things that annoy them :-) I might investigate allowing multiple correct answers in the future, but it does make the interface a bit more complicated, some of this complication is supported in the underlying board implementation (required for multiple lines in the PGN viewer), but there would still be some work in integrating multiple lines into the tactics and problem viewer.  I'll especially look at this if the current set ends up too easy due to lack of longer mates.


Next button:
- Good idea, I might try some visual notification (busy icon cursor) rather than sound, at least to start with.

Column Titles
- The new tables available in the next update do sorting etc on the server side which is much faster, but some kind of "still doing something" notification would still be useful as it still takes a second or two to respond.

Legal move indicator
- This has been done and will be available in the next few days. Illegal moves are no longer accepted (i.e. they don't fail the problem for you as they did in the past) and during piece movement, the current square the piece is hovering over is optionally coloured depending whether it is a valid square to move to in that position. Checks are not explicitly shown, but it will be obvious when you start dragging a piece that you have a check issue if you don't see the valid squares highlighted as you expect.

Row/Column coordinates
- This is also already added and will be available with the other updates in the next day or so.


Thanks for the feedback, don't hesitate to add any further suggestions you might have.

Regards,
Richard.

[bubak]

Richard wrote: "don't show rating, tags, comments while problem is live" changes ...

It's very useful to see the rating of the problem.  I hope that doesn't go away.  For standard-rating problems, if it's 1400 or 1500, I just solve it on the screen.  If it's over 2000, I think, "boy this is going to be a tough one", and I set it up on a real chess board on the table and get ready for a 5-minute think.  I have a hard time solving tough problems on the little board on the screen.

Without the rating displayed, that won't be possible any more 

A completely separate issue:  I get served a lot of problems where my rating gain is 0 if I get it correct.  I suggest only serving problems where you gain at least one point.

[drahacikfm]

That last post was from me, was logged into my son's account Bubak by mistake.

[richard]

Hi drahacik,

Unfortunately for rated problems I think you may be in the minority on wanting to see the ratings etc while the tactic is still live. I think most users see the ratings display as interfering  with the realism of the situation, i.e. you don't have a flashing sign when you play an opponent saying "there is a tactic on the board and its' a nice easy one" (if only there was, I might win more games! :-)  ).

Part of the new feature set is unrated modes where neither the user's or the problem's ratings are changed, as the user is playing in complete isolation here (their actions don't impact others via changes to problem ratings) I think it would be reasonable to allow the user to do whatever they liked including making the visibility of ratings/comments/tags during play a preferences option (although that change will probably have to wait a few weeks).

Unfortunately unrated modes are probably going to be a subscriber only feature.  I had seen the "not show ratings" change as a net benefit to all users so I'm sad you will be essentially losing a feature you found useful and can't get access without subscribing. I still think hiding those clues is for the better in terms of improving chess tempo as a training tool.

I also have larger board sizes on the todo list, so that is something that might make your life easier in this situation (and a feature I'm not planning to add to the subscriber-only list).

Regards,
Richard.

[richard]

Sorry , I forgot to address the zero rating change question.  This is an issue of rounding the rating change to an integer, the zero rating changes are probably all between 0 and 0.5.  I have fixing this on my todo list, but it's probably not going to make the next update.

Regards,
Richard.

[drahacikfm]

If I'm in the minority about wanting the rating to show, I guess I'll have to live with that

I agree the problem's rating gives you a good clue about what to look for, something simple or something very deep.  But it's not completely true that in real life you have no clues:  If I'm playing Grandmaster in a tournament, and I think I have a simple 2-move combination that wins a piece, I am very suspicious that it is a trap, and I spend a lot of time before I take that piece.  But if I'm playing a 1400 player, I know there's a very good chance that he just overlooked it and is losing a piece.  Of course I still check that I'm not falling into a trap, but maybe not as much as I would against a Grandmaster.

[richard]

If I'm in the minority about wanting the rating to show, I guess I'll have to live with that

I agree the problem's rating gives you a good clue about what to look for, something simple or something very deep.  But it's not completely true that in real life you have no clues:  If I'm playing Grandmaster in a tournament, and I think I have a simple 2-move combination that wins a piece, I am very suspicious that it is a trap, and I spend a lot of time before I take that piece.  But if I'm playing a 1400 player, I know there's a very good chance that he just overlooked it and is losing a piece.  Of course I still check that I'm not falling into a trap, but maybe not as much as I would against a Grandmaster.

That is a reasonable point.  The new problems list which will be available soon will allow you to page through the problems and see their ratings and success rate, you can choose to view  problems with ratings that you are in the mood for solving.  Obviously the problem viewer doesn't track your progress etc, but it does let you solve problems with known ratings.

Regards,
Richard.

[slacker00]

I figured I'd add that I use the problem's rating to gauge the problem also.  I  use the same reasoning as drahacikfm.  But I kinda like the idea of NOT knowing the rating.  It will force a more objective examination of the problem, and the problem's rating will be more true.

BTW, richard, thanks for all the hard work.  You really seem to be putting a ton of effort behind Chess Tempo.  I hope it turns out as you want and exceeds your expectations.

[richard]

I figured I'd add that I use the problem's rating to gauge the problem also.  I  use the same reasoning as drahacikfm.  But I kinda like the idea of NOT knowing the rating.  It will force a more objective examination of the problem, and the problem's rating will be more true.

I think the new zero-information display will also help combat the rating inflation that is likely to occur when a large number of longer mates will be removed.  Many of them were plagued  with issues, but there will be some collateral damage on difficult but legitimate mates.  Harder mates are still there in the new set but a good deal less common than before.  If the problem set gets too dumbed down I'll have to think about problem set generation strategies that are a bit more precise in which mates are rejected (and also consider supporting multiple correct answers, although there are some technical issues with both the UI and my problem generation approach I'd have to resolve before looking further at that). 

I've also implemented a slightly different time penalty method in blitz (suggested by  Roq).  The new method penalises time taken after the first move at a higher rate than time used before the first move. The intention is to encourage users to have grasped the essence of the position before moving rather than playing "hope chess" and guessing one move after another (a tactic that has reasonable success at CTS due to often not being required to reach the conclusion of a tactic).

BTW, richard, thanks for all the hard work.  You really seem to be putting a ton of effort behind Chess Tempo.  I hope it turns out as you want and exceeds your expectations.

Thanks, It makes up for my lack of action over the last 10 months or so :-)  I was starting to get embarrassed with the number of times I'd tell people, "Good suggestion, I'll add it to the TODO list." without getting any items onto the DONE list :-)   Just as I thought I'd have time to work on the site, my paying work would get in the way.  I'm hoping if enough people end up deciding to subscribe to the premium features , It will make it much easier to prioritize time for work on CT.

Regards,
Richard.