42975

[palemon]

I think last move should be removed. Bxf1 exists two times. And Nd3, Kxf1 is not even an alternative move.

evaluations according to rybka:

10... Ne4 11. Na4 Nxd2 12. Nxc5 Nxf1
-13. Bxf1 e6 14.Qe1 b6 15. Na6 Bb7 16. Bb5 Nd8 17. Qe2 Rc8 18. Qd2 {2.44/14}
-13. Nd3 f6 14. exf6 Rxf6 15. Bxf1 e6 16. Qe1 Rf8 17.Nde5 Nd8 18. Rb1 a6 19. b3 {2.43/14}
-13. Kxf1 Rd8 14.Nd3 d4 15. Qe1 e6 16. c3 dxc3 17. Qxc3 Rd5 18. Rb1 Bh8 {2.43/14}
 

[richard]

The new generator fixes this bug.  I think it has been around for a while but had gone unnoticed due to users not having a way to see the evaluations.  I have a short term fix in place which marks alternatives that were missed in situations where toga output multiple lines for the same move choice, however this doesn't help in this position where there were there are so many winning moves than other winning moves exist outside the top 4 moves.  I can probably fix this without a generator verification run , but it requires a fair bit of new code on the server side so it might end up easier to do a verification run.  At the moment I'm generating new problems with the bug fixed generator, the next time I upload those I'll consider starting a new verification run, although I'd rather wait to see if any other bugs are going to surface.

Regards,
Richard.

[drahacikfm]

Richard,

There really should be a way that you can manually disable a problem so it is not served up.  Should not have to wait for a new generator to be written or for a new multi-week generator run.

People might actually quit Chess Tempo for getting marked wrong on Kxf1 or Qxf1, which are completely winning, if such a thing happens to them more than once.

Old problems that are disabled have the red "disabled" displayed when you view them.  Whatever flag is set in the database for those problems, can't you just set that flag manually?  I think it's worth the time, because you don't want people justifiably irritated by problems such as these.  There might be only 10 or 20 of these in the whole dataset, then the set will be almost perfect 

[richard]

it is fairly easy to disable them (as you point out I can just set the same flag as is set for other disabled problems).  The bit that isn't done yet is to make sure they are not "re-enabled" the next time I do another generator run , as if the issue is one that is not fixed by the generator it will happily re-enable them again.  This is not a major issue and I've done half the work to allow human intervention in the disabling process (adding some new flags to the database), I just need to get the other half done (making the code that loads the output of the generator honor the new "disabled by hand" flags).  I'd been putting off doing this until having a few more problems , with the plan of getting the rest of the "hand disabled" code done and then disabling a bunch of 1 star problems (in cases where multiple high rated users had given them one star). However, I take your point on the negative impact these problems have. Even ignoring the rarity issue, telling a paying subscriber that , "Yeah that is a bad problem, I'll disable that sometime in undefined future.", is probably not an acceptable answer.  I have some newly generated problems to upload in the next few days so I'll see if I can get the disabling code finished before then.

Regards,
Richard.

[drahacikfm]

disabling a bunch of 1 star problems (in cases where multiple high rated users had given them one star).

I give a 1-star rating very rarely and usually only in these two cases:

1) It's a hanging piece problem where the opponent just left his queen hanging.  Or the hanging piece is checking your king, so you have to take it.  I don't give 1-star to other hanging piece problems, where a bishop or knight is hanging.  I give 2 stars to those problems.

2) It's a one-move problem where the idea of the problem is not shown by that one move and the problem makes no sense as a one-move problem.

In that second case, it would be much better to extend the problem more moves than to disable it.  In most cases, the problem could actually be very interesting if it had more moves.

[newlook]

I give a 1-star rating very rarely and usually only in these two cases:

1) It's a hanging piece problem where the opponent just left his queen hanging.  Or the hanging piece is checking your king, so you have to take it.  I don't give 1-star to other hanging piece problems, where a bishop or knight is hanging.  I give 2 stars to those problems.

I don't give a lower-than-3 star rating because of a hanging piece. I figure that the skill rating system takes care of that. Some hanging piece problems have error rates that are surprisingly high. If a problem empirically proves to be easy, its low skill rating will direct it to those who can benefit from it.

I do agree about forced moves. I also give a low star rating when the piece moved by the computer before your first move is left hanging, unless there is some complication that would make you think. It's likely to be the first thing people look at.

Perhaps there should be a list of evaluative tags, separate from the tags for tactic category.