Oh no, more problems!

[drahacikfm]

Ok, now I'm mad at Richard.  Today I finally finished all problems in my personal problem set "Standard 0-1350" which was a total of more than 8000 problems in that rating range and took many weeks to finish.  Now Richard adds more problems to the site and I have 616 more problems to do in that set!

[richard]

:-)

I should keep a careful eye on how quickly those 616 get done, it would be a shame to time the next update as perfectly :-P

Richard.

[andreacoda]

I am sure he didn't do it intentionally!

Or...did he? 

Cheers,

Andrea

[drahacikfm]

Well I already started a new problem set "Standard 0-1400", so I won't go back to the old 0-1350.   The new one had about 1400 problems, so it will take a while.  I just have to beat Richard before he adds more new problems!  Then I can make another new one 0-1450 

All new problems that have only one move are rated 1281 initially.  Many are stupidly simple, just hanging pieces.  But there are some sleepers in there which will eventually become 1800 or 1900 standard.  Those are the ones I miss because I move fast when doing the low-rated problem sets!

[drahacikfm]

It's interesting that each day when I come back to Chess Tempo, there are more problems in my problem set "Standard 0-1400".  For example when I quit yesterday I was almost down to only 1000 problems remaining.  Then today when I logged in  there were 1132 problems in the set.

I guess a lot of problems rated above 1400 are getting solved correctly by people, and that knocks the ratings of those problems down below 1400, and they get added to my personal set.  I wonder if I will ever finish this set

[drahacikfm]

I am sure he didn't do it intentionally!

Or...did he? 

Cheers,

Andrea

Ok, "Problem set is empty.  Choose another problem set" !!  Finally finished all problems rated under 1400 Standard.  Richard better wait a while before adding more problems so I can enjoy being finished!

10,288 of them total on the site!  With 17,894 total problems on CT, I guess that means the median Standard rating is below 1400.

A lot of those under-1400 problems were very nice combinations. Not extremely hard, but very nice.  I'm hoping the next set, 1400-1450, will be even better, not so many hanging pieces which seemed to be about 20% of the under-1400.

[uri blass]

I am sure he didn't do it intentionally!

Or...did he? 

Cheers,

Andrea

Ok, "Problem set is empty.  Choose another problem set" !!  Finally finished all problems rated under 1400 Standard.  Richard better wait a while before adding more problems so I can enjoy being finished!

10,288 of them total on the site!  With 17,894 total problems on CT, I guess that means the median Standard rating is below 1400.

A lot of those under-1400 problems were very nice combinations. Not extremely hard, but very nice.  I'm hoping the next set, 1400-1450, will be even better, not so many hanging pieces which seemed to be about 20% of the under-1400.

My experience is that rating is often misleading(the same problems is also with CTS)

Here is an example
I solved both problems but
the problem with higher standard rating and higher blitz rating is clearly the easier problem considering the fact that the simple fork is not accepted as a solution.

http://chesstempo.com/chess-problems/22938
http://chesstempo.com/chess-problems/49065

I will repeat what I said in the past.

My opinion is that it is better to decide that problems get rating based on something else than success of humans in them.

success of humans can be clearly misleading because a problem can get low rating not because it is easy but because many players read about the problem in some chess book so they remember the solution.

The fact that you found more problem that their rating got down suggest that there may be a deflation in the rating of problems and I wonder if somebody compared the rating of problems to see if the rating tend to go down with time.

Edit:Note that you can expect to see hanging pieces even at 1600 standard level
The following problem(that I got in blitz) has standard rating above 1600 that again proves my point that result of humans are not a reliable source to decide if a problem is easy

http://chesstempo.com/chess-problems/23308


Uri

[drahacikfm]

I do not agree that there is a better way to rate problems than how humans perform against the problem.  Your post doesn't prove your opinion.  If a problem looks easy but humans get it wrong a lot for some reason, then the problem is not easy for humans.  If a problem is difficult in an absolute sense, but it's very famous and it's in lots of books and lots of people get it right because they know the problem, then it should have a low rating.  It's that simple.

Regarding deflation:

When new problems are introduced, they have ratings based only on the number of moves.  One-move problems have ratings of 1381, I believe.  After a while, some of those 1381-rated problems will go up, and some will go down.  At the start none of those problems would be under 1300, for example, but after a few weeks, many will be under 1300.  And many will be over 1500.  But the average will still be close to 1381, because that is the average of all the other one-move problems on CT.  This is all normal behavior, and does not indicate deflation.  It only indicates problems moving towards their correct rating.  Richard has checked and there is no general movement of the average rating.

[drahacikfm]

Keep in mind that the only purpose of a rating system is to predict performance when two humans or a human and a problem play against each other.  The purpose is not to measure strength on an absolute scale.  This is explained in Dr. Arpad Elo's famous book "The Rating of Chessplayers", 1978,  in which he invented the Elo rating system for chess.

When you realize this, then the performance of a problem against humans is the perfect way to give the problem a rating.  Because the purpose of the problem's rating is to predict its performance against humans!

[uri blass]

Another thought about the same subject.

I think that the performance of a problem against humans may be dependent on the humans who solve it.

The same problem can get 60% against people with rating of 1500 and
40% against people with rating of 2000 because there may be some wrong idea that only strong players can see and the weak players solve the problem correctly more often because they do not see the wrong idea and the move that they guess is simply good enough when later it is easy to win against the computer stupid blunder that prefers losing the queen in one move and not
allow some mate in 5.

What should be the rating of the problem?
If it starts with low rating then the rating is going to go down because most solvers are going to be weak players who get 60% against the problem.

If it starts with high rating then the rating may go up.

Adding do not know option when users lose less rating points can improve the rating because hopefully more weak players will choose do not know instead of guessing.

It is not a full solution to the problem because there may be problems when
the following pattern happens:


1500 player:see winning move 1.XX and does not consider the defence 1...YY so choose XX and believes XX wins.
2000 player:see winning move 1.XX and decide that it is not good because of defence 1...YY so do not choose 1.XX and lose rating points.

2400 player :see winning move 1.XX see defence 1...YY and calculate refutation of YY that starts by 2.ZZ so choose the correct move but may lose rating points in blitz because of using too much time.

Computer:see that 1.XX YY lead to forced mate so it choose 1.XX TT that only lead to winning a queen by white that is easy to find

Uri

[drahacikfm]

The beauty of the rating system is that it takes into account all those things automatically and fairly!  It takes into account how many 1500 players are guessing, how many 2000 players are getting it wrong, and how many 2400 are getting it too slowly.  All those factors, and many more such as "the problem is famous and many people know it already".  The resulting problem rating averages all those factors together and produces a rating for the problem that is the best possible.  Exactly because it will predict correctly the future performance of all different levels of players averaged together!  That's the only purpose of a rating system.  Not to assign some absolute strength to a problem.

Anyway, this forum topic was about Richard adding new problems to the site and how many problems are left in personal problem sets.  Now it's changed to something completely different.  I suggest discussion of the rating system should be in the rating topic, not invading other topics.

[drahacikfm]

Oh no he did it again!  The day after I finished a major personal problem set, Richard adds thousands of more problems.  Now that set is full again.  I swear he does that intentionally.  Next time I'm not telling when I finish a set! 

[revenant]

Hmm, I *wondered* where all these new problems with "Attempts: 0" and "Rating change: +0.7" were coming from  :-)

[richard]

<snicker> , time to start generating problems in time for your next personal problem set completion :-)