Since the long challenges are 10 days long and hence there is enough time to solve the problems completely rather than partially solving them. Partial grading effects the sole purpose of a long challenge. people should be encouraged to go for full solution instead of particular test cases.
Actually this is a highly debatable topic. It has pros and cons both.
Some of the times i feel it is not worth but some of the times constraints of sub-tasks gives a lot of hint for the approach to solve problem…
And the problem is it is not worthy to solve problems for partial grade (if you really want to learn), because even brute-force algo passes for minimum points.
this is a really crucial topic that has been brought up on this forum once again. My views are that partial grading is good in some cases, and not in other cases.
In cake-walk or easy or easy-medium level problems, partial grading should be removed, and there should only be a 100pt task. The reason for this is that the smaller subtasks are simply ridiculous in most cases in such subtasks and simply do not test the user in any way whatsoever. I see many coders not aiming for 100 pts, but just coding a solution for smaller subtask and getting it accepted, and stopping midway once they get partial points, whereas they could’ve easily solved the problem for 100 pts, had they not been distracted by the “easy” points they were getting through smaller subtask.
In problems which are very challenging, like MARCH15’s RNG , partial grading is good. But only if the smaller subtask itself is challenging enough, and user has to tough it out to even pass the smaller subtask. RNG’s smaller subtask itself was pretty challenging, and hence made it exciting for me to think for hours on end on this question, which i wouldn’t have done if there was only one 100pt task, with full constraints.
On the other hand, on medium level problems, sometimes the smaller subtasks are really very easy and and really dont serve any purpose.
Hence, my point is that, even the partial graded subtasks must be challenging and not so easy that they can be passed by almost brute-force solutions. The purpose of subtask is to push the users to think more and more on the problem, which he/she would’t even have considered to think about in case of no partial grading.
Above mentioned thoughts are purely my veiwpoints and might clash with others, but i request @admin to consider these.
Partial scoring allows us to better rank people with same number of solved tasks. In lot of cases there is a difference between “brute-force”, “better solution” (which gives more points), “even better solution” (which gives even more points) and “correct solution”. Why so low number of people solved RNG from last long challenge at least for 10 points? Just because most of contestants were lazy to write a brute-force?
The way you described your point of view is indirectly harming people who are not good enough right now to solve even relatively easy tasks. You say “very challenging” about RNG? And I would say that this task wasn’t challenging, only problem was to implement solution carefully enough to fit into TL; and partial grading was helping those who are lazy to implement solution with good asymptotic. I don’t find any problem from last Long challenge challenging, but it doesn’t means that it is true for all other contestants. For very beginners even direct implementing of brute-force is challenging.
the reason why RNG’s 10 pt subtask had so less submissions, is because normal brute forced approach wont pass even the smaller subtask, and even normal matrix exponentiation would pass the 10 pt subtask…
well, as i said, these are purely my veiwpoints, but dont you think that at least the subtasks on cakewalk or easy level questions make no sense?