Partial Marks for AC Test Cases.

general
, , , , ,
#1

The current method of marking scheme only rewards points if you pass all the test cases. But I think it would be better if it provided partial marks for all the passed test cases.

For example if the judge has 20 test case files, then for every test case you should be awarded some marks like 100/20 = 5 points for passing each of the test case or something like that.

The point is - one should be awarded more points, or at least some points, if one was able to get AC in almost all the test cases as opposed to someone who couldn’t or, even better, someone who didn’t even try that question/subtask.

Also, the person who could pass all the test cases will be anyway ahead of those who couldn’t so its not unfair to anyone.

Just a thought. What do you guys think?

#2

I think that is not a good idea, because it may be possible of getting some points even if your code is completely wrong and also for your point of unfairness for those who atleast try to attempt the problem is invalid because we are getting score only if our solution is correct not for attempting.

3 Likes
#3

I am not asking for marks for just attempting the question. I am asking if you code could pass some test cases. If my code can pass 80% test cases then there is some credibility, its not just a random attempt.
And your point of getting points for wrong code is a completely different stories because it is not a problem with this system only, its a general problem of weak test cases, people get complete AC for wrong approaches all the time.

#4

In some problems you have to just print

" Yes " or " No "

" Alice " or " Bob "

At that time my code will be like this

#include <stdio.h>

int main(){

printf(“Yes\n”);

return 0;

}

Or print No

I will get more than 50 % score :slight_smile:

2 Likes
#5

For that purpose, there is many problems with “subtasks” which participant can awarded if he/she can solve in simple or specific case.

#6

I don’t think this is a good idea.Hackerearth does that but I personally feel this is not correct as this will make participants more oriented towards getting those partial points rather working on to get full AC.

#7

Ain’t a good idea. You should try and pass all tests in a subtask, not just a few ones that can be considered weak

#8

@ryan312 still ur soln is wrong, the aim of these contests is that u learn/to know how much u know. there are cases in a partially marked problem like (30,70), people get the tougher 70 but they might not get easier 30 due to runtime errors which cannot be solved at times so we need to accept if solns is not right then we don’t deserve it.
for partial recognition, u have lunchtime, long challenge where tell how much partial the will give u.U can’t expect in cookoff because it is aimed at ACM ICPC where u have only to verdicts WA ,AC.

2 Likes
#9

@abba5 what if there are t test cases where t>=10?