Programming C, C++, Java, PHP, Ruby, Turing, VB
Computer Science Canada 
Programming C, C++, Java, PHP, Ruby, Turing, VB  

Username:   Password: 
 RegisterRegister   
 DWITE Issues
Index -> CompSci.ca, Contests -> DWITE
Goto page 1, 2, 3, 4, 5  Next
View previous topic Printable versionDownload TopicSubscribe to this topicPrivate MessagesRefresh page View next topic
Author Message
A.J




PostPosted: Thu Feb 19, 2009 7:03 pm   Post subject: DWITE Issues

The questions this DWITE were good, but the actual testcases weren't good.

For #5, for some reason, they said that there would be at most 10 edges, and none of the testcases even come close to that!!

And all of the 'x' functions were : 1 + x/100

And, I still had my second submission to hand in, but the server crashed! I couldn't submit my 2nd submission for #5.!!

The testcases should have incorporated testcases where there were up to 10 edges with multiple different equation using 'x' (not necessarily 1 + x/100). All the teams in my school (seeing that the worst case was 10 edges) didnt code it, since a testcase with 10 edges and say about 2 equations (using 'x's) on each path from 'A' -> 'Z' is impossible to solve in the given time limit.

And how does one go about solving #4 ? The highest anyone got was 3/5 (although that could be just us not getting the intended solution)

I am kind of frustrated that #5's testcases were actually SO small, and very doable (much easier than what our school thought).
Sponsor
Sponsor
Sponsor
sponsor
DanielG




PostPosted: Thu Feb 19, 2009 7:08 pm   Post subject: RE:DWITE Issues

very true, this is very similar to the final problem of the first dwite, where a simple heuristic guess was sufficient. for next year, dwite text case difficulty should reach that mentioned in the problems, and not allow those who attempt incorrect shortcuts (though correct for the sake of those test cases) to get full (or close to full) points.
chopperdudes




PostPosted: Thu Feb 19, 2009 7:16 pm   Post subject: RE:DWITE Issues

i've seen almost the exact question as number four, sadly, i didn't take note of the solution as at the time i probably didn't have the knowledge to understand it anyways.

i believe those who got 3/5 did so because of the time constraint. what i did, was find all possible permutations of the #'s, and find all the total distances for each permutation. but we also have a min variable to keep track of the minimum distance. so that we will not compute those permutations where during the process of permuting the distance is already greater than min.

i'm sure there is a "standard" way to do this type of problem (possibly dynamic programming), but i believe most who got 3/5 used brute force one way or the other.
Insectoid




PostPosted: Thu Feb 19, 2009 7:18 pm   Post subject: RE:DWITE Issues

And I wan't too please about being gimped a point in question 2...

nah, I'm kidding, I did read the notice.

lol, one of our 1-man teams submitted question 2, but had accidentally formatted the output with an extra space at the end of each word, resulting in 0/5, although he had the right answers.

It is rather annoying when the only real-time communication with Dan/Tony is via IRC, and my school has those ports blocked.
Dan




PostPosted: Thu Feb 19, 2009 7:19 pm   Post subject: Re: DWITE Issues

A.J @ 19th February 2009, 7:03 pm wrote:

For #5, for some reason, they said that there would be at most 10 edges, and none of the testcases even come close to that!!


The test cases where in the range of 1 to 10 edges as the question side they would be. Considering so few got theres working at all i don't think testing cases closer to 10 would do anything.


A.J @ 19th February 2009, 7:03 pm wrote:

And all of the 'x' functions were : 1 + x/100


Agian this is with in the sepcs of the question and again since so few got one working at all it did not effect the outcome.


A.J @ 19th February 2009, 7:03 pm wrote:

And, I still had my second submission to hand in, but the server crashed! I couldn't submit my 2nd submission for #5.!!


The server crashed exctaly when time ran out, it was due to the next contest not being found in the DB. So if you saw the error you where to late to submit it any way.

A.J @ 19th February 2009, 7:03 pm wrote:

The testcases should have incorporated testcases where there were up to 10 edges with multiple different equation using 'x' (not necessarily 1 + x/100). All the teams in my school (seeing that the worst case was 10 edges) didnt code it, since a testcase with 10 edges and say about 2 equations (using 'x's) on each path from 'A' -> 'Z' is impossible to solve in the given time limit.


Thats just silly. You just need one test case right to get points. Also this question was ment to be as hard as it souned, the test cases where just happened to be more lenient.


A.J @ 19th February 2009, 7:03 pm wrote:

And how does one go about solving #4 ? The highest anyone got was 3/5 (although that could be just us not getting the intended solution)


I still have to look in to this but it was desinged so it would be hard to get the last test cases as it would time out.


A.J @ 19th February 2009, 7:03 pm wrote:

I am kind of frustrated that #5's testcases were actually SO small, and very doable (much easier than what our school thought).


The idea of dwite is to wite a program that sovels the question not the test case, it's just hard to make the judge enforce that. Idealy it should have been has hard as your school thought however some teams got lucky in there asumption that they all where 1+x/100.
Computer Science Canada Help with programming in C, C++, Java, PHP, Ruby, Turing, VB and more!
Dan




PostPosted: Thu Feb 19, 2009 7:24 pm   Post subject: Re: RE:DWITE Issues

DanielG @ 19th February 2009, 7:08 pm wrote:
very true, this is very similar to the final problem of the first dwite, where a simple heuristic guess was sufficient. for next year, dwite text case difficulty should reach that mentioned in the problems, and not allow those who attempt incorrect shortcuts (though correct for the sake of those test cases) to get full (or close to full) points.


Your doing it wrong if you are using shortcuts. DWITE is ment to be partice, if you cheat you are only cheating your self. Since the judge is an AI and not a real person there are limiations in being able to catch all short cuts. For example it is very hard to stop hardcoding witch just outputs a random number or peoleop acidentaly geting questions right. In this round the test cases where aucatly quite good at stoping hard coding and for question 5 the parsing of the weight on the edges of the graph was only a small part of the problem.


insectoid wrote:

It is rather annoying when the only real-time communication with Dan/Tony is via IRC, and my school has those ports blocked.


I am constaly watching my e-mail and my MSN and jabber is on during contests.
Computer Science Canada Help with programming in C, C++, Java, PHP, Ruby, Turing, VB and more!
Insectoid




PostPosted: Thu Feb 19, 2009 7:28 pm   Post subject: RE:DWITE Issues

lol, I have to confess to submitting randomly generated answers. It is a contest, and my competitive nature forces me to eek out those few extra points at the last second.
SJ




PostPosted: Thu Feb 19, 2009 7:44 pm   Post subject: Re: DWITE Issues

for 5, i ended up randomly generating numbers too T.T

how difficult would you guys say #4 and #5 are? CCC stage1, stage2, or IOI?
Sponsor
Sponsor
Sponsor
sponsor
chopperdudes




PostPosted: Thu Feb 19, 2009 7:47 pm   Post subject: RE:DWITE Issues

#4 would be stage 1, but should be in senior i'd say, maybe S5... that's because i've yet to seen brute force failing in junior lol.

#5 i'd say would be at least the stage 2, but for me, i would probably not be able to tell between a hard stage 2 question and an IOI question lol...
ecookman




PostPosted: Thu Feb 19, 2009 7:55 pm   Post subject: RE:DWITE Issues

i wish i actually had the time to participate in DWITE...it sounds fun
A.J




PostPosted: Thu Feb 19, 2009 8:07 pm   Post subject: Re: DWITE Issues

chopperdudes wrote:

#4 would be stage 1, but should be in senior i'd say, maybe S5... that's because i've yet to seen brute force failing in junior lol.

#5 i'd say would be at least the stage 2, but for me, i would probably not be able to tell between a hard stage 2 question and an IOI question lol...


are you kidding me?

#4 and #5 were easily #4 on stage 1 senior. They were actually quite easy.

You must notice that on CCC, they won't EVER give questions where they expect you to get a few of the testcases and not all of them Confused

so you really can't compare

and Dan, I really disagree with you about the server down time

it was down at least 4 minutes before the contest (i remember seeing the time left)
Dan




PostPosted: Thu Feb 19, 2009 8:29 pm   Post subject: Re: DWITE Issues

A.J @ 19th February 2009, 8:07 pm wrote:

and Dan, I really disagree with you about the server down time

it was down at least 4 minutes before the contest (i remember seeing the time left)


The time shown is based on your computers time, if you don't blive me look at the html and javascript source.

The cause of the crash was the next contest not being in the database and it only trys to load that when the contest is over.

I do have plans for fixing the clock so it is based on the server time, witch i hope to get in for next school year.
Computer Science Canada Help with programming in C, C++, Java, PHP, Ruby, Turing, VB and more!
chopperdudes




PostPosted: Thu Feb 19, 2009 8:46 pm   Post subject: RE:DWITE Issues

how does one actually approaches #5?
A.J




PostPosted: Thu Feb 19, 2009 8:50 pm   Post subject: Re: DWITE Issues

chopperdudes wrote:

how does one actually approaches #5?


well, since dan's testcases were actually small, bruteforce would have worked. the people who got 5/5 on this assumed that the equations were always : a + x/b.....it was sooo rigged...oh well, at least 7th place overall isnt too bad...Sad
aoen




PostPosted: Thu Feb 19, 2009 10:00 pm   Post subject: RE:DWITE Issues

Question 4 was worded very poorly, our team only got 1/5 (when we likely would have gotten 5/5) because we thought that #..# had 2 units of wiring (a wire was placed on each of the 2 .'s), the first sample test case didn't clear up this ambiguity either (the answers assuming either the wires took up a square of space or wires are placed in between squares were identical). Lame :<. Our team wouldn't have moved up because of this, but still, how many people edit these questions -.-. Question 5 was flat-out too long, even the brute force approach required tons of code.

And as always bad test data is bad.
-The limits for the constraints question 5 weren't tested
-The question implied that the function for the time of each "path" could be pretty much anything (no brackets though), all of the text cases were in the form f(x)= a + b/x
-If such small constraints were going to be used for testing, how come the constraints on the actual question were so large, they indicated that brute force wouldn't work

I'm wondering, why is the test data always so... easy? Is it because larger test cases will lag the server XD?
Display posts from previous:   
   Index -> CompSci.ca, Contests -> DWITE
View previous topic Tell A FriendPrintable versionDownload TopicSubscribe to this topicPrivate MessagesRefresh page View next topic

Page 1 of 5  [ 73 Posts ]
Goto page 1, 2, 3, 4, 5  Next
Jump to:   


Style:  
Search: