Scoring problem with the Challenge problem (ANUMFS)

As you can see from the last submissions for the Challenge problem of MAY’14 long contest (ANUMFS), both me and @aawisong submitted solutions which scored an absolute score of 0 (yes, that means we got the answer exactly right for each test case without the need of asking any question), which is the minimum possible.

However, it seems that the system cannot handle such scores and it ended up assigning a score of 0 to everyone. Of course, the correct ranking would assign full score to me and @aawisong and a relative score of 0 for everyone else.

The issue is that the problem allowed the possibility to obtain 0 score per test case. My suggestion would be to fix this problem in one of two ways:

  1. fix the scoring algorithm in the judge so that the minimum score per test case is > 0 (e.g. add 1 to the absolute score of every test case, or something smaller than 1, if you want, like 0.00001) - should be very easy to do and wouldn’t change the relative order of the scores (it would simply add a fixed value to the score of each contestant)

  2. fix the general scoring problem in the system, so that scores of 0 for minimization problems are handled correctly - this is probably more difficult to fix

6 Likes

Hi @mugurelionut,

We were already discussing about the same issue from about 24Hours. Regarding the error with judge, we have a solution to fix it. It is not a problem.
We are only investigating on how the data is extracted by the users. There is a genuine way to do it by using binary search on answer (We were aware of this method before the contest but we were okay with it as this method was used by contestants in many previous contests and also did not really expect users to completely extract the data). How ever we noticed that @aawisong used less than 150 submissions to get the whole test data answers. We are looking into the issue more carefully. Please give us some time, we shall respond accordingly.

Let me tell you how I did it, without binary search. First, I computed a hash for each test case in order to verify which test case my code is running on (this took about 30-40 submissions, let’s say). I used total running time to know how many cases are below a certain hash value and to choose hash value limits which uniquely identify each test case. Then I identified all the test cases with answer (-1,-1) (there are 6 such test cases out of 20) - another 20 submissions. For the other 14 test cases I computed the representation of the answer (x-xmin,y-ymin) in base 6 (8 submissions per test).

7 Likes
//