I'm talking about opening the visible dataset for those who have passed the tests already. Not a big deal if the participant could compare his solution for the hidden dataset with the results on tests for the visible dataset. As an alternative there could be an option to allow to submit the solution for the hidden dataset to test it on the visible dataset without a penalty.
On Thursday, April 12, 2018 at 8:12:06 PM UTC-4, Pablo Heiber wrote: > Hi, > > The reason to hide the test cases is because testing is an important part of > programming. Providing people with real test data, even without the expected > answer, would be doing a lot of the work for them. In the previous system > with I/O downloads, there was no alternative, but this is much more > significant test. In many problems, testing that your answer is correct is > significantly easier than solving it (at least if P != NP), so giving test > data is a huge help in debugging. Even when checking the answer is hard, > having the tests there helps debug runtime errors, time limits, and possibly > even hint at the reason for wrong answers if you inspect the data carefully. > > > In real life, you don't get to see how the users will use your code in > advance, so most of the time the programmer needs to come up with the > comprehensive set of scenarios. This policy mimics that, in a way. > > > Best, > Pablo > > > On Wed, Apr 11, 2018 at 10:16 PM Дмитрий Кузьминов <dmitry....@gmail.com> > wrote: > There could be a reasonable solution of the problem with testing Hidden after > you have solved the Visible. What if the platform would allow the participant > to download the whole visible dataset as soon as he has solved it? He has > earned the points for his submission of Visible, why not to open the tests? > > The separate submission of Visible/Hidden would also be a benefit, especially > with this improvement. > > > > On Tuesday, April 10, 2018 at 3:45:55 PM UTC-4, Bartholomew Furrow wrote: > > > That's a good point about not being able to test your new solution on the > > same input that you know you got right, Dmitriy. A few other thoughts: > > > > > > > > > - I don't know whether the UI makes this clear -- note to team, hopefully > > there's a really clear indication that you still have points from previous > > submissions -- but you can't lose points you have earned on a visible > > dataset, no matter what you submit later. > > > > > > > > > - I agree that it isn't without penalty, but I don't think 4 penalty > > minutes is a very significant cost. I also don't think there's anything > > wrong with that cost being present. > > > > > > > > > - When writing an analysis, it's often helpful to first consider a simpler > > version of a problem before you consider a more complex version. I believe > > that's what they're doing here. > > > > > > > > > - Although there are a couple of things that make it less appealing -- lack > > of ability to test on a Google-generated dataset, and the extra 4 penalty > > minutes -- it's still totally reasonable to code up a solution that solves > > the Visible and not the Invisible set. For example, what if you don't know > > how to solve the invisible one? Or, as we've discussed here, what if you're > > planning on submitting the Invisible set later if you get a chance? > > > > > > > > > Cheers, > > > Bartholomew > > > > > > > > > On Tue, Apr 10, 2018 at 1:21 PM Дмитрий Кузьминов <dmitry....@gmail.com> > > wrote: > > > Bartholomew: there is another concern. Let's imagine that I've solved the > > Visible dataset, and I know for sure that this doesn't solve the Hidden > > one. I submit the solution, I get the positive verdict, and I start the > > better implementation. In the old platform I can always compare the results > > that the old (less effective) solution produces with the results that the > > better solution gives me, and that comes for free: I don't need to spend an > > attempt with a penalty. Now it is not free. Moreover, I even don't know > > which attempt will be taken to score my visible dataset (shall I submit my > > correct solution for the second time?). > > > > > > > > > > > > Today we've got a Qual Round analysis, and the analyst still assumes that > > old rules are valid. Let's regard the "Saving The Universe Again": as usual > > we have a clue that if we know that "Since there is at most one C > > instruction in this test set, we can solve the two cases independently"! > > Who cares now of the solutions that doesn't solve both datasets? > > > > > > > > > > > > -- > > > > > > You received this message because you are subscribed to the Google Groups > > "Google Code Jam" group. > > > > > > To unsubscribe from this group and stop receiving emails from it, send an > > email to google-code...@googlegroups.com. > > > > > > To post to this group, send email to googl...@googlegroups.com. > > > > > > To view this discussion on the web visit > > https://groups.google.com/d/msgid/google-code/3684c25d-ffce-4f4e-ae74-b82113e69d54%40googlegroups.com. > > > > > > For more options, visit https://groups.google.com/d/optout. > > > > -- > > You received this message because you are subscribed to the Google Groups > "Google Code Jam" group. > > To unsubscribe from this group and stop receiving emails from it, send an > email to google-code...@googlegroups.com. > > To post to this group, send email to googl...@googlegroups.com. > > To view this discussion on the web visit > https://groups.google.com/d/msgid/google-code/77865ab0-33a6-460a-a369-76079a3c2bbd%40googlegroups.com. > > For more options, visit https://groups.google.com/d/optout. -- You received this message because you are subscribed to the Google Groups "Google Code Jam" group. To unsubscribe from this group and stop receiving emails from it, send an email to google-code+unsubscr...@googlegroups.com. To post to this group, send email to google-code@googlegroups.com. To view this discussion on the web visit https://groups.google.com/d/msgid/google-code/45c69f0f-671d-4866-82c2-1e9fa9c2b40d%40googlegroups.com. For more options, visit https://groups.google.com/d/optout.
