Just out of curiosity i want to know how do grader work for challenge problems?
As far i know (Correct me if i am wrong ) for normal questions grader compares the output file of our solution with standard set or result, but for challenge questions n possible combinations can be obtained by all submissions so how do grader check these results?
This discussion would be lot easier if we talk about some concrete example, so do that for last contest - Chef and Painting…
Your goal is to minimize number of steps. Let say you have a for two contestants the their number of steps for single test. Which one is better - obviously the one with lower number of steps.
The final thing is to assign score for both…score the better one is 1.000, score for worse one is calculated by proportionality or inverse proportionality.
no but they have to compare the actual solutions too… i.e. the path and the co ordinates.
So to check whether the solution is correct or not do they have n output files to compare? or how do they compare to check whether solution is correct or not?
One possibility is that for such questions they may have their own program and this program takes output of our program as input. If their output(taking our output as input) matches the required output(i.e. for chef and painting all the cells are painted) then it adds the number of steps to the sum. If all the cells are not painted(based on our output), it simply gives wrong answer. If our program is correct for all the test cases then sum(total number of steps in all test cases) is compared with the sum of best solution.
Of course first it is checked whether your solution is correct or not, after this step there is this score calculation I mentioned. I do not know what is the technical solution, but this is not really important, is it? Why are you asking?
No, just i was curious to know how do they compare such large number of different output… But thanks 
Yahh Sarthak might be…
Your output is taken as an input for a common program that attempts to fill the board according to the test case. If it fills the board correctly then you get AC. What score you obtain after getting AC is described by @betlista. The score calculator might be something like:
Your Score= Total no. of users / max(1,Users that took more steps than you)