Due to my dislike of CodeChef discuss, I choose to discuss/rant about this topic here on Codeforces. Although non-Indian participants may not relate to the ICPC situation, I'm sure the CookOffs and Lunchtimes are a common issue.
First of all, why is the difficulty gradient so fucked up every single time? (Contest Link). Be it ICPC, or CookOffs, or LunchTimes, the same thing is seen over and over again.
Scoreboard for reference:
Secondly, why does online ICPC selection have only 5 questions for 3 hours? Not to mention, 3 of them are pretty hard, and the difficulty drop is huge, so it does not help filter out deserving teams from undeserving ones.
Is ICPC selection supposed to be based on a speed typing test? In unbalanced contests like this, it almost always happens that some undeserving teams qualify. Every penalty matters a lot too. For instance, we failed to qualify for one of the regionals — KGP — because of 1 penalty.
So, my question to CodeChef is: How hard is it to set a balanced contest? Codeforces seems to does it every week. CodeChef fails to do it even in its most important yearly contest, along with its monthly short contests, every single time.
I honestly believe that even I can set a better contest than this. Lastly, why was the leaderboard hidden as soon as the contest ended?
This is so true. The set was very similar to a Cook-Off, once you make 2 questions, the rest are freaking tough. Last to last year had a balanced set, with wide variety of questions. Sad to see so skewed balanced sets, that too in such a contest where you have only one chance. Had it been different contests for different regionals, the disappointment would have been slightly less.
The problem set actually reminded me of 12th board exams English comprehension!!
So it is not only me then. I see first 2 questions(in a Cook-off). Okay I have ideas. Other 3, what do I do after taking the input?
Is codechef really a non profit organisation?
I wonder who approves these kind of problem sets.
PraveenDhinwa i guess.
.
Is there any possibility of reconduction?
Another sucessful attempt by Codechef at shattering hopes of various deserving teams.
I personally think that the contest was well balanced.
Wall Grid was a straightforward bruteforce problem.
A Family of Bipartite Graphs was also a relatively standard DP.
I don't think you should blame codechef and the problemsetting panel for this since they would expect the above 2 problems to be easy and medium respectively.
The number of successful submissions indicate otherwise.
Your "relatively standard DP" question was solved by 3 teams out of 3000, so I guess it must be standard for Red+ people, but most Indians are not Red, so I'm sorry :)
That problem was solved by my teammate AnonymousBunny.
Good for him :)
That guy is doing CF from 4 years ,and you wont take any noob in your team XD
facebook hacker cup 2017... Petr attempted all questions of world finals.. and he was the only 1 to do it... so problems must be standard for him and others are...??
But even if you say that BIPFAMIL is a hard problem (it's fourth problem), the other three are relatively very standardish. Did you solve all of these three? If yes, I am not sure what's your complain?
Nice problem set, indeed. I think I accidentally registered for typing speed test.
How was it a typing test? I don't get it. Which problem except cakewalk was just typing test?
Second was unsolvable to many deserving teams. So the best chance to get selected was how lucky you chose cakewalk at first go. And then how fast you are able to type. Isn't it?
314 teams solved it. If you are not in 314 teams, I don't see how you call yourself a deserving team?
Are you going to select only those teams who solved at least "2 problems"?
I think there are going to be many teams who solved 1 question but due to "maximize distinct colleges" rule, they will qualify for regionals. Can you prove me wrong?
I think the maximize different colleges rule is justified. Had that not been in place and selection was only on the basis of merit, then in the world finals the proportion of teams would mostly have been from Russia and China, from universities like ITMO. There would have been hardly any teams from India. And since the world finals have some criteria to have distinct colleges, it should apply at the regional level as well. But I do think that there should some limit, around 50-60% of the slots for maximizing college teams and the rest based on merit.
How would you differentiate more than 2500 teams that solve only one problem? By just a single cakewalk problem.
You can't differentiate them because there is very little difference between them. There are a lot of people at lower skill level and there is nothing to really differentiate.
(However a team contest with only 5 problems seems a bit weird to me too...)
Thats what the point is...those who did the 2nd question too won't be reaching the regionals i guess... The gap was in the 2nd and 3rd question...
Ok the contest is now over and everyone was lazy/bored/ignorant/tensed/whatever to write the brute, its not like only 10 people faced this problem its was (3000-20).
Anyways now the situation is that teams will be selected for regionals mostly based on how fast they submitted the first solution.
Now a team who has been practising for more than a year will make sure that the solution is fool proof and try to prove it and then submit whereas a new team who doesn't have much experience will just directly submit without any proof or testing.And i know teams who has done the same. Now we didn't know the later problems will be tricky so we didn't rush to submit the first one! so can you actually say who is a better deserving team among the ones who only solved one problem (because it seems if you didn't do the second one your team is useless) so among the useless/trash-teams of the country can you say who is more useless than the other and select for the regionals??
A team that has practiced for more than a year should be strong enough to solve such easy problems very quickly with no need for serious testing.
(Disclaimer, I didn't take part in the contest and have no stake in this.)
This was the worst problem set ever.
Second was a really weird observation problem to be honest.
Frankly, No-one I know had a valid proof of their solution, and due to the lightning fast speed of codechef editorial releases, it is still a mystery for us.
Last year, the other team cheated to defeat us, and this year was our last chance at icpc as a team and this shit happens.
Thank you codechef for an amazing icpc experience.
And please let Codeforces host it next year onwards, that is the least codechef could do for the community. Thank You.
If you solve enough problems solving second one won't be hard enough for you. There has been similar problems in both Codechef and Codeforces. Atleast after a few penalties you could have guessed the intended solution/pattern. And then you call yourselves pro and has been practising for a year. Idk how much practise you did. :)
The second one was definitely a non-trivial problem.
Okay, yes. We have witnessed this situation several times, but unlike this shitty set, we most of the times could skip such problems (To come back to it later) and solved the next one, then come back to it.
But here we just had 5 problems, with the worst possible gradient.
And I never called myself a 'Pro', please keep those assumptions to yourself :)
It was a 5th problem in some previous codechef contest (division wasn't introduced then. ) Now you skip 5th problem of cc and then expect to get in icpc? :)
I really don't have time to waste it on people like you, who don't understand the point of what I'm trying to say.
Good for you, you could solve it and yes we Suck. :D
You are demotivating people.
First one was a cakewalk, agreed
Second one was based on a vague observation about which i still haven't seen a proof (coz Codechef editorials are shit fast).
Third one was a question that requires 0 logic, only tremendous implementation that assesses nothing apart from implementation.
Fourth and Fifth problems were "unsolvable" coz maybe Codechef was too lazy to come up with something good.
And what's with the immediate hiding of the ranklist and the contest page? If you require us to pay registration fees for giving an online contest, we deserve to atleast see our performance with respect to others.
MikeMirzayanov, please host ICPC for India on Codeforces next year, Thank you.
I dont think it should be called easy.But it is not even as difficult as only 3 people can solve it.
Also comparing to last year and previous year contest. This is more balanced because
In 2017, from my knowledge there were 3 straight forward. One standard dp, one basic fft, one wasy-medium level counting problem.
In 2018, apart from tree hashing rest all were easy. Even it was easy-med problem if you know how to hash a tree which was also a pretty standard thing.
So I you are stuck on an easy problem , you are lost in those years.
But this year, the problemset was good to separate people in first 100 of the list. But its just that here numbers do not signify difficultly.
I feel on someother day if many people submitted 3rd one, then everyone would have tried backtracking on it.
Also in general my opinion fares as always problems on codechef seems to be solved by lesser people than expected making problem look hard.
I agree with rajat1603 that the set was more balanced than the balance I expected with respect to last two contests.
We can agree on that, it was more balanced in comparison to the last one.
But main question is the contest really balanced with such falls in submissions on further problems??
As I said I dont think that fall on submissions rate justify difficultly.
Do you think people only 22 people could come up with bruteforce for 3rd. It was just people lazy to try it out expecting it not to pass. Also please dont say it was implementation heavy, it was not.
Coming to fourth, though I worked for close to 2 hours of the contest trying to apply some open close trick of zscoder and etc. I could solve half the problem using it. So I was further into it trying for other half. But towards the end I tried to change my approach then we found the solution in very short time and could code it in less than 10 mins though it was 3 minutes late. I think once you see editorial (incase there will be one) you will feel it was an easy dp to think and implement as well.
I have no clue of last problem. Also did not give time to it. So no comments on it.
Out of curiosity, do you have a source for this 'open close trick'? Or at least a quick summary? It sounds pretty interesting.
http://mirror.codeforces.com/blog/entry/47764
Thank you!
Could you share your approach for BIPFAMIL? It's highly unlikely we'll get an editorial anyway.
try to count not connected graphs and subtract from total. I tried to comment my code well. I think it should help understand easily.
Solution Outlines: https://discuss.codechef.com/questions/138614/acmind18-icpc-online-round-solution-outlines
teja349 and rajat1603
"Also in general my opinion fares as always problems on codechef seems to be solved by lesser people than expected making problem look hard."
I agree. Most users of codechef take it this way.
google code jam round 3 problems were alo standard fo errikto .. gennady .. rng58.. but why not for u... any answer
summing up the contest in few words "worst contest of my life
Next problem submission goes down by approx 1/10. Which kind of balanced contest do have this?
PraveenDhinwa which contest have such rise in difficulty?
I mean, ICPC isn't really meant for people with rating < 1400... There isn't really a point on doing a contest for people with low rating, that'll cause a typing competition (OBS: I don't think this contest was a typing competitions and also believe the problemset was fairly balanced).
Just because you are good, don't look down upon others.
I totally agree with RajatDe. May the best Indian teams who find the problemset balanced qualify for WF's and get a honorary mention.
online ICPC must be like (DIV1 + DIV2). but this contest is like (DIV0.5).
Born_Confused see this xD
dontknowcoder
Any idea what the team cap per institute is for amritapuri? Also,when do they release the lists?
Can somebody share the approach for 2nd Question(Reduction Game) for Codechef What a way to remove server problem.
People made too many submissions on it. Of course, it was not meant to reduce the server load. The first problem statement being long was meant for this purpose :P
So making the problem statement intentionally lengthy is the solution of handling server load: P
Haha not. We had 32 checkers this time. A 100% increase than our earlier max capacity of 16 checkers. The number of submissions were a lot this time than usual due to lower accuracy.
Can you please tell us total number submissions in the contest. I think this would be very less form 2016 and 2017 year contest(from accuracy and number of successful submission). Just put hard questions and average teams will never try to solve them and traditional server problem solved.
What kind of irresponsible answer is this? You increase the problem statement length unnecessarily and make it a "You understand English quickly? You get a regionals spot!" situation?
A lot of my friends had difficult in understanding the problem properly, simply because it had a lot of obsolte information and hid necessary information.
This is so accurate!!!!
Which necessary information did the problem statement hide?
"The robots have gained consciousness, and have begun questioning their purpose in life. So given a chance, they will collide with each other and end their misery."
Why put such a important information in the story?
You forgot to mention the next line" They can coordinate with each other as well and decide when they should start their journeys.". This entire paragraph describes the objectives of the robot. There are proper explanations of the examples which explain this even further. I am not sure what confusion did it create?
This part is confusing:
You are wondering whether it is possible to give the robots the initial directions in such a way that no robots collide with each other (ie. they'll all be safe), or if no matter what initial directions you give, some of them will end up colliding with each other (ie. unsafe).
Even if in the worst case robots were colliding answer was unsafe but the above statements makes it look like we need to find the best case to save them colliding from each other.
I think there's no problem with it.... sort of puts a smile on ur face in middle of a tensed contest .. i mean the way question was presented
You could do another thing to reduce the server load.
Please let codeforces host it. :)
Hold it right there sparky. Are you human?
They said the scoreboard will be given after plagiarism check, which doesn't make sense to me, but they might have a reason.
I think the question set for SnackDown got Exchange with the icpc online qualifier round!!
snackdown 1-A was balanced though,this icpc was ....
sanckdown 1-A was not balanced. It was easy af and atleast last year had better problemset.
ohh then why only ~1000teams solved last one and ~1200 solved second last out of ~16k teams that too worldwide.
because about 50% of teams are first timers participating without much knowledge of ds and algo. Many others didn't solve seeing that they would manage to qualify even without scoring full in last 2. Yes last was a bit tricky and more mistake prone. If you found them hard i don't know what is actually easy to you. Perhaps cakewalk? And yesterday's 3rd wasn't tough either. Most of us(including myself) didn't try that just seeing number of submissions and thought basic cakewalk wouldnt pass. And I don't get one thing, the problemset was same for everyone why are you guys complaining. Yes teams for colleges like IITs, IIITs and BITs had lots of competition and many good teams fail to qualify for that reason but for that you shouldn't blame the problemset. And suppose 2 cakewalk problems were given, still it would be a typing speed test wouldn't it? And if you are so much pro why couldn't you solve the simple range intersection one before your competitors? And isn't every round of cf a speedtest for first 2 problems? So i don't understand where the problem is?
But i do agree gradient was pretty high. But again you shouldn't expect a problem is easy if it is 2nd should you? Cf contests are kind of better because after solving 2 questions fast you get some boost so you are likely to perform better with rest. And getting penalties early is a disadvantage( which is what Ashish's team faced). But if you are arguing about the first problem, teams less pro er than your's did it fast then i guess you are not as pro as you think?
first problem statement was too long so we left it and saw it again after 500 submissions. thanks for your single sided analysis Mr Pro
Even I did leave it for the first time i saw it. But seeing submissions in its favour I again switched to it. Have you never participated in cookoffs and ltimes or what? I seriously am doubting that.
i left participating there long ago when they took my rating for no reasons,fast exponentiation gkg code and fastinput reader,i said i saw it again after 500 submissions.i was writing a backtracking for 3rd problem but gave up on it due to peer pressure
You left participating is your problem right? And it clearly shows you just participate for ratings. And if you just copied from geeksforgeeks, did you try mailing them about that? I know many people whose ratings were restored back. Anyways this time try participating in cookoffs and ltimes and next year don't make the same mistake. :)
yeah i mailed them but they said its late now. see if you have low rating that indicates you are not doing anything,rating is the thing you get after solving its not gifted to you by them. you seem to be pretty arrogant for no reason i see
"If you have a low rating it indicates you are not doing anything" — No it's not at all true, there are many coders who just solve last problems of any contest, obviously they have lower rating does that anyway tell you they are not doing anything? Or have you seen any incidence where a low rated coder has been insulted somehow in discussion forums? Anyways I am not being arrogant, I just dont feel its right to watch the problemsetters being attacked.. Even your codechef post was very harsh, though it got upvotes because lots of dreams was broken yesterday. But everyone should realise not everyday is our day.
last problems of cf rounds are more scoring they will be definetely high up in ranking,more rating that means.
ok see we all people here are noobs,you prove your worth at your level,goodluck for world finals bro and next codejam's final.
everything here looks easy and standard to you,you should move to much much upper level problems then.
and being arrogant,that's the thing a listener decides not the speaker
i guess it is utterly useless talking to antipr000!
he wont listen to anyone whatsoever and is a mere waste of time
oh wait is he PraveenDhinwa???
I dont think the problemset was bad. Sure, a couple of problems more would have been better. But i dont think giving the excuse that a problem is too long is okay in any contest. Evn if u do CF, there are plenty of times when a D or E is easier than C. I myself have done contests solving ABF and not CDE. So its ur duty to read all the problems atleast once. Also, do cp for fun. Many very good coders dont even participate at all and so rating is not a correct judge of skill anyway.
i got full in both rounds. thanks for analyzing things from your point of view only. cf rounds manage that in 3 and 4th problems. and you are a pro bro.
"because about 50% of teams are first timers participating without much knowledge of ds and algo"...lololololololololol
Yes it is true. I know many colleges in my locality having 50+ registrations and only 4-5 students from those colleges are actively participating in contests. And i am not pro but I don't like complaining just because I failed to do something. You failed because you weren't good enough yesterday, it's not on codechef or problemsetters, its on you. And getting full in both snackdown rounds doesnt make you a good coder. Yes you might be a good coder and yesterday was just not your day but you also need to learn to accept your own mistakes. Afterall questions were same for all. Anyways just out of curiosity what was your rank yesterday and how many did you manage to solve?
I think your teammate went just a bit too far. Even though your points are totally valid, raging because of an ICPC contest is just useless.
CodeChef's bad contests are the reason why I'm not doing it at all too. I know a friend who once got 27th place in div2 for getting fast AB done and that's just plain shit
Worst ICPC regionals ever
this year Moscow Subregional was the worst :rage:
Are there editorials?
You will find them after 16 years
I don't get this joke :(
Forget getting the editorials too. Or the problems for practice.
let alone editorials they didnt bother to put it in practice section yet!!!
And they take Rs 900 for such contests.
ICPC India regionals are so overrated. Can't even organize an online round with balanced questions and a functioning IDE. Combine this with the unreasonably high fees for an online preliminary round and the illogical "maximize distinct colleges" rule, and one can't help but wonder if this contest really deserves all the hype it gets.
I would seriously love to see Hackerearth or any other body taking over ICPC in India.
Codechef is simply too bad.
1) They prioritized Snackdown over ICPC. They decided to reserve two weekends for Snackdown, and push ICPC to a weekday. They claimed that "ICPC officials" stipulated this date, but I guess it's mostly just Codechef people and the contest directors whom Codechef could contact quite easily.
2) I don't understand why they removed the "each regionals has a separate online round" format. It was nice, people had a chance to brush away bad days and qualify for Regionals, and it was win-win for everyone. This format of hit or miss is already very shabby.
3) Horrible problemset. I don't know if they ask Reds to test it or something, but for the past two years the questions have been extremely skewed and unbalanced. Last year was a "Go math or go home" contest, this year was even worse. Pathetic.
Anyways, they didn't have any server issues this time around, thankfully.
And I'm curious why they increased the online round fees so much this year. WHERE IS THE MONEY GOING? WHY SHOULD WE PAY ALMOST 1000 BUCKS FOR ATTENDING AN ONLINE CONTEST?
Is ICPC trying to make profits out of participants? And who is colluding with them?
Server issue was their. In the end (5 mins) I couldn't submit a problem because the website was not responding. Also having a contest on weekday is something which is totally not acceptable, I am having mid-Semester exams currently and it makes things pretty difficult.
website is shit pure shit,i proved myself not a robot to check ranklist
1000 bucks like 1000 USD? Or what currency are we talking about here?
Also, please don't refer to some regional body by the name ICPC... every region organizes things completely differently.
Of course if you can only submit once every 30 seconds!!XD
Apart from the contest (in which we just failed to submit the 3rd in time and came 92nd, hence the 8th in our institute, out of contention), I'd also like to bring to light the bizarre difference in selection criteria of different regionals. There is a gross difference between different regionals, and we are informed of this right on the day of the final contest? In fact, Pune-Gwalior wants to take 116(!!) out of 127 teams from different colleges, whereas Amritapuri says that they're rank-based with 250 slots? Why can't there be uniform criteria implemented everywhere, i.e. say 50% reservation?
Why didn't you attempt WORDGRID over BIPFAMIL? From my perspective, once you know WORDGRID is bruteforce, one team mate can start writing code for it while another works out the math for BIPFAMIL. Did any teams try this strategy?
Bhai I guess when we see that only 10-15 teams have solved a particular problem it's very hard to have guts to even read it!
By that logic BIPFAMIL is harder than WORDGRID and my argument stands. Moreover, there are 3 people in a team. I think it is reasonable to expect that at least one reads all the problems :)
I agree we should have read the rest of the problem!
But when its a tricky second problem its better all three teammates try to get it accepted (that's what we did) and as Ashishgup said people had pressure because they couldn't solve the second one also and just by solving that we could get a better rank so then it was unlikely that we would have read very less solved problem!
anyways Bhai I or my team have no issue with you or anyone but see the submission to each problem and tell me if it looks like an balanced contest where 3000+ teams participated across the country.
Yeah true. We fucked up under pressure, and that's on us. After having a lot of penalty, we started working on the same problem in a hurry to get an AC, and trying the wrong problem to do so.
However, a better balanced problem-set would have definitely helped a lot
We started off slow but once we got AC on the first two problems, we did this. We didn't solve BIPFAMIL, but there was definitely enough time to implement WORDGRID regardless imo.
Oh Shit. WORDGRID works with a simple brute. :(.
On a lighter note, When among over 2000 teams of best Indian coders only 20-30 solve a problem you get a feel of DP + BITMASK + 2D Segment Tree + Blogewoosh#6 + Trie + Some weird trick. You don't think Bruteforce.
Stop crying b***h. You think you are a great coder because you were yellow. As PraveenDhinwa said, 3rd problem had a brute force solution.
Learn algorithms later, first learn to write brute force solutions.
tedmosbypimp I guess people didn't get your sarcasm!!
Codechef outdid themselves this time by consecutively screwing up the online contest. Last year they had an issue in handling floating point answers and checking them with the checker code, and this year, well... Just stop taking the responsibility for holding acm icpc contests in India for a year and give some other platform some chance for a change.
Indeed after rank 324 it's just the matter of who solves the cakewalk problem first!!!
I am ranked 325 :P
Congrats bro!! U were quicker to solve that one then.
You sure you are not Flash??!! XDXD jk
326 here :P
The actual difference in our level though :(
PraveenDhinwa Is it possible to reconduct the online prelims?
What the hell!!! Your teammate's rage was too much!
PraveenDhinwa ,Answer my question very honestly ,do u really think the teams who solved just the first problem should qualify for the regionals ?? Do u really think a problem solvable by a person who started competitive yesterday is enough for a team to qualify to the country's most prestigious sport programming competition ?
This is what ICPC is. They want to maximize the participation of different colleges. It's just their rules are, it's their priorities. What can I do about it? In some sense, I find the reasoning behind it justifying, it's to spread awareness about programming.
Oh you find this justifying! So spreading awareness > someone's dream? How do u justify that someone who worked hard day and night , who dreamed everyday of being there will fail to qualify and someone who gave it casually will qualify? Do u really think that this is the best way of spreading awareness ? How different is this from the Caste based quota that has eaten our country up?
Suppose World Finals doesn't have the maximizing number of colleges 'quota'. How many Indian teams do you think would qualify? And if this philosophy is followed at the world finals, why should it not be valid for regionals?
This Contest could be better with 7 questions, 5 remaining the same, 2 questions from medium difficulty(Average Coders) Because:-
If you are one to whom Codechef people will listen, Just a request, give up on conducting ICPC for India.
ICPC 2016 — Judging time too slow.
ICPC 2017 — Floating point error in Judgement.
Now, three times in a row, truly a consistent behaviour.
Maybe luck is not on Codechef's side, let other platforms such as Codeforces etc. conduct ICPC for India.
It wasn't floating point error in judgment last year, rather the inability of people to do what's required for the question. They didn't manipulate any programming language so that it would give a floating point error. Many people were having trouble displaying the answer within the accuracy required, which is something you are expected to know. Even a simple google search tells you how to do it.
However, there were issues last year as the second problem gave Internal Error.
I think this decision of maximizing number of distinct colleges (which leads to the issues you raised) is taken at the site level, so PraveenDhinwa might not be able to answer that (obviously one can ask for his opinion on the same)
For instance in the past two years, Chennai has always supported meritocracy over maximizing number of distinct colleges by reserving some X% of seats for top performers even though they might not be first in their colleges.
If Amrita and Kolkata follow the same this time, it might just help in marginally alleviating the events of this year's online round.
kolkata region does that http://icpckolkata.in/PrelimRules.pdf top 10 will get selected
It's a good thing for starters but IMHO, it's just namesake since I am guessing at least 6-7 distinct colleges in top 10 [so they would have anyway gotten by the top team criteria], which essentially means only 3-4 teams on merit barring the top rankers out of the 160 slots.
I think that Mr. jr is correct. This is one of the most prestigious coding competition in the world. And so the selection for the regionals must be fair. But PraveenDhinwa you have downgraded this schema. SAD...
It's interesting to see that a lot of people think that the registration fees comes to Codechef. I know for a fact that not a single paisa from that goes to Codechef. It goes to the regional sites for them to conduct the regionals. In fact Codechef pays the regionals as well. So maybe get your basic facts right before accusing others.
But eventually it is the regional site which hands over the problem setting responsibility to CodeChef. I wonder why the regional sites still trusts codechef even after lot of mishaps. I guess for them sponsorship matters more than anything.
"I wonder why the regional sites still trusts codechef even after lot of mishaps." What were the mishaps? Online participants unduly criticise the problems due to not getting selected due to the rules of the ICPC regionals. Why do you think that by changing the platform this attitude will change?
https://blog.codechef.com/2013/11/06/acm-icpc-iit-kgp-online-round-goof-up-we-are-sorry/
Every single word is Correct .
Taking 1000 rupees for an online contest, but don't have enough servers to handle load!! And what type of stupid logic is this to add unnecessary English statements, just because your site cannot handle load?? Is this a TOEFL reading examination or a coding contest??
I mean, ICPC is in English, so it makes sense to do the regional qualifiers in English as well.
My point was, including un-necessary redundant statements that are highly unrelated to the question.
I agree with your point regarding unbalanced Problem-set but everyone needs to admit that 2nd problem was really nice. I mean so simple it looks but has really nice corner cases.
.
Well said. The contest should be set such that most of the deserving teams get selected. Many deserving teams from our college couldn't qualify because of this type of contest.
Question from a person not familiar with background of preliminaries in India: why do they only have 5 problems every year? You can't really grade teams properly with such approach, when you need to cover range from red guys to absolute beginners. It feels like you are going to have hard time even with achieving your main goal of picking strongest teams to qualify, i.e. bisecting on "good enough to qualify" — results are going to be much more random (you have less options when stuck on easy problem or got penalty for WA, you have much more advantage from getting a problem which you've seen before etc.). For example, qual rounds in NEERC subregions usually have 10+ problems. You can easily fail to balance it too (like giving 5 trivial problems and 5 tasks that are way too hard), but at least you should have easier life when trying to make it balanced. Is there some specific reason for having only 5 problems in Indian quals? It is weird to see this kind of selection for ICPC, where you have much more problems to handle.
On a side note — it is funny to see half of the comments reading this post as "We failed qualification round, we aren't happy about it, so we're going to complain" and replying with "Tasks were trivial, you should work on your skills instead of whining; any strong team should be able to qualify to the next round no matter what kind of problemset they have to work with". Compared to some of the recent "I'm not whining, but..." posts, this one has 0.00% of unrelated whining in it :)
PraveenDhinwa
PraveenDhinwa :)
This is the thing with codechef, I mean just look at their last few cookoffs etc. The contests highly unbalanced. Just solve A and B quickly and get a top 30 rank. This is what pissed people the most, the uneven distribution in question difficulties. Also, people were pissed by the poorly constructed problem statements and the question to time ratio.
Also, i think that they only have 5 questions because -
1) Too lazy to think of many questions.
2) Afraid of the high load on server due to more submissions.
Do NEERC regions have qualifying rounds before the regionals? Just to be clear, the ICPC Indian round which was held today, was a 'qualifier', from which teams get selected to go to the different regional sites in India, and from which teams get selected for the World Finals.
NEERC has subregional contests for quite some time, and recently they also started to add additional round before subregionals (example). So in such case you need to go through the chain 3 contests in order to qualify for WF, and not 2 like in case of India. Also outside Asia there is no that "pick regionals that you prefer" thing; so for NEERC there is a single final competition — held simultaneously in several locations, but on the same set of problems and with a single standings / WF selection.
In SEERC (region which my native country belongs to) it is also chain of 3 contests before WF — you may read recent unrelated blog for some stories about how it works or doesn't work :) As far as I know it wasn't always like that too, but that chain-of-3 structure was there before it started to appear in NEERC — so I'm quite sure that while I was there around 2010-2015, it was like that for at least some of the seasons, if not for all of them.
Ah ok! Looks like I had gotten this structure totally wrong. Thanks for the info :)
aa gya chutiya isika intzar tha
Translation for non-Indians : here comes this douche again, we were waiting for him.
PraveenDhinwa
I have replied to this thread down in the dalex's post telling my reasons why it's being done like this.
ALL LOSSES IN THIS WORLD ARE DUE TO A LACK OF ABILITY. IF YOU WANT TO CURSE SOMETHING, CURSE YOUR OWN WEAKNESS!
Not really
Yes I know actually this quote is from tokyo ghoul anime.....I also hate todays contest....It is a worst contest i ever take!..very unbalance problemset
Worst shit ever ! Was it a coding Competition or typing test ? Codeforces please conduct contest from next time. Codechef is real shit .
https://discuss.codechef.com/questions/138347/acm-icpc-online-round
Yup, agreed. Some comments are saying "if your team is strong, it will qualify for regionals regardless of the problem set". What they are failing to understand is that an online prelim round is meant for filtering teams that are good enough for regionals not WF. Sure the contest filtered teams that can reach WF, very well but it failed in serving its purpose, and that too miserably.
Can codechef set a balanced contest?
Just started to practice on ratatype.com .
A MUST FOR ICPC.
where's the haha react button
1600(900+700) Rs for such a contest Wow! nice buisness ACM ICPC
The problem set was unbalanced so that people won't able to submit codes and hence there will less load on their servers.
Wow! Looking at the pictures, that is a dangerous teammate to have. Who knows it could be you next time instead of the monitor? (please take it as a joke, and no offense).
Talking about the problemset, I feel the more number of problems with a less steeper gradient could have helped comparatively in a better judgement than the one which happened today. A bit upsetting to see teams solving 1 cakewalk problem getting a chance for the regionals over those which solved 2. [Personal Team Name: bttr_chiz_aloo, Univ: BITS Pilani Goa Campus]
Nonetheless, this is part and parcel of the game, and ranting doesn't help. Maybe next time a better performance? ( though this is my final year :) :P )
Who knows it could be you next time instead of the monitor?
Don't you think you are blowing this a little out of proportion? They are good mates. He had actually kept his monitor in his room for more than an year, waiting for the perfect time to rage.
I don't know who his teammates are, nor am I interested in knowing. It was just meant to be a healthy joke. I know well enough that he and his teammates would be sane enough not to hurt each other.
Lets not doubt anyone's sanity because of my joke.
I don't think there is any way of me (or anyone) knowing that you were joking, because the tone was certainly not of a joking one. You even edited out your comment after my reply, further solidifying the fact.
Yeah, I guess, I should have mentioned that it was a joke. My bad. Apologies.
Is rajat1603 a member of UniverseIsDeterministic ?
Was that an official ICPC round? LOL. Why can't you hire some red/yellow guys to make a contest with 10-12 problems? Or even use some strong local teams for that, inviting them directly to the next round?
To get a real feel for the answer to that question, you have to meet sparky a couple of time first.
@dalex: No. It was not onsite round. There are two stages of the ICPC in India. The first is online round which is open to everyone, from which the teams are selected go to the various regionals sites where they compete onsite for slots in the world finals. In the online contest, we have 5 problems, whereas the onsite contests consist of typical 10-12 problems. I think you are confusing this round with onsite round.
I'm not stupid. Why can't you prepare 10+ problems for online round?
Higher the number of problems, higher the load on server. With 10-12 problems, servers will give up. First understand the real problem and then comment solutions.
I am sorry. I completely misinterpreted you.
It's been traditionally 2.5 or 3 hours typical cook off types round. The regional directors ask to prepare the contest according to this. We can potentially increase the number of problems, but the selection criteria and level of the participants will always be such that these kind of complaints can't be avoided. One factor is that ICPC always try to maximize the number of colleges for the participation. The other thing is that there are a lot of people on similar levels in India, so the ties between them will always be broken by the penalties rather than number of problems solved. Even in the onsite rounds, we have 10+ problems, there are 50% of the teams which get more than one problem correct, rest of the teams just get one problem. These are the reasons why I think having more problem doesn't solve the issue of qualifying to next round.
My understanding suggests more number of problems could differentiate the teams in better way. Anyway, if you don't agree,let's agree to disagree.
"It's been traditionally 2.5 or 3 hours typical cook off types round"
Yes definitely, the quality has been coming down over the years, just like the typical cook-off rounds :D
Why would you care about lower half of teams solving <=1 problem in qualifying to next round? I think more problems and 5 hours is definitely way to go.
Because ICPC cares about it. They want the number of teams from as many different colleges as possible, hence the first problem is supposed to be cakewalk. Regarding the suggestion of 5 hour, I agree with it, that it can help with a better gradient on top, but as I said that it might not help much at bottom. For the upcoming online rounds, we will try to do this if it would be feasible to make it 5 hours.
I would say in order to get as many teams as possible you should have 5-8 really easy problems with some gradient. The top will be decided on the rest 3-4 problems. Otherwise the contest is frustraiting for most of the participants.
"might not help at the bottom", there is very little to differentiate among the bottom teams. We are not complaining about the contest being unable to differentiate among the lower half, rather teams ranging from 90 to 99 percentile (more specifically, teams ranked from ~30 to ~300 as they all solved exactly 2 problems) as it is a "key range" because this range is as small it can get to accommodate both those teams who qualified easily and those who failed to qualify.
If you have a problem with 2500 AC and also problem with 300 AC, that means you can create 2 more 2500 AC level problems and probably 2 more 300 AC level problems. Then you leave 2 20 AC level problems and you get 8 problem good selectional competition, where many teams can solve 3 and more.
dalex life is short, don't waste it arguing with PraveenDhinwa,
whatever he doesn't listen to anyone
Simply because conflicting interests. Latest announcement was: ACM-ICPC participants are entitled to huge discounts on our certification exam! Members of any team that solves all 5 problems can take the Foundation, Advanced or Expert exam for just? Rs.500, that’s a maximum discount of? Rs.4500! Solved 3-4 problems? Take Rs.1000 off! Even participants who have solved 1-2 problems can get? Rs.500 off the price.
I am going to be really honest. Although I appreciate the efforts of Codechef to promote coding culture in India, the CCDSAP is turning to be one of the biggest scams in the history of Indian competitive programming. Rupees 5000 for a programming certificate? How is that justified? Isn't that a bit too much for an average college going Indian to afford?
Thank God this thread is happening on Codeforces, else codechef would have required additional servers to handle the traffic XD
I don't want to complain over the difficulty range of the problem set. I screwed up the 3rd question while implementing, butwhy not just keep a 5 hour contest with some more problems?
People might get stuck in some easy problem and lose time on penalty but they can cover up in a 5 hour contest and that would surely judge the teams much better. Since if some team is even second in their college, they might be in trouble in getting to the regionals given the rules (no complain in the rules since the key motive is to spread awareness which is fine).
I hope they will design a similar problem set for snackDown Pre-Elimination Round.
As they told it was pretty balanced contest.And most contestants like balanced contest. So why not????
Just out of curiosity, what was the laptop's worth? JK
It was a monitor that I had saved for a long time for I knew there would come a day in my life when I'd want to break things.
Reduction Game was a COPIED problem from problem I from this SET
How are they the same problem?
Also thank you for caps-locking the word "copied". Otherwise I might not have noticed that word.
I wouldn't call it copied, but both of them use the exact same idea. In fact, our team reduced it to that exact same problem and solved it. (Did not know it existed though)
As for how: Create a new vector V which contains A[i]-K for all A[i]>K. For obvious reasons, we take out the maximum value out of this vector (because all others have to reduce to 0).
The problem now becomes: Among the remaining elements of V, how can we perform operations such that the resultant element is minimum (one operation is selecting two non-zero elements and reducing both by 1).
This is essentially: Total sum — Maximum number of mismatched sock pairs you can form from the given vector V, where ith number corresponds to the number of socks of the ith color. :)
Maximum number of mismatched sock pairs you can form from the given vector V, where ith number corresponds to the number of socks of the ith color.Can you please tell me how to find that?
Binary search. If you can make k pairs, you can make k-1 pairs too and vice versa. To check if you can make k pairs :
no need to include bs , you could calculate most optimal by modified check() func..
What we Want link What we Got this shit
How to solve problem 2 REDCGAME?...
Anyone rajat1603 Ashishgup PraveenDhinwa , teja etc?
1st sort the array.. then i tried priority queue/heap for maintaining bottom and top value still got WA...
Thanks in advance :)
Greedily take <=k element.
try to save the largest element in the sorted array by nullifying among the other elements
Let arr be the elements >k except the LARGEST one having size n.
we remove all k elements from the arr to the array… arr[i]->arr[i]-k of i in [0,n-1] and add it in the answer
We find the sum of all elements from 0-arr[n-2] leaving behind the arr[n-1] element
Csase 1: sum<=arr[n-1] … best you can do is by saving arr[n-1]-sum from this…. so add LARGEST-arr[n-1]+sum to answer
Case 2: you can balance it optimally. so the answer will be 0 or 1. that depends on the parity. so add LARGEST-parity(sum+arr[n-1])
Kinda non-trivial….
@MikeMirzayanov host icpc on codeforces next year
Why does ICPC want to maximize unique colleges? I don't understand. I see mentions of "drawing awareness", but I'm not sure what the point of drawing awareness is.
This "maximize unique college" thing seems to look for diversity, but honestly it would be way more useful to not charge money for an online round if you want to incentivize more variety of team to participate .-.
Man I gotta admit.. that's some next level rage and tilting. Feel sorry for you bro
Well. I would like to say one thing. Instead of increasing awareness of programming, such contests are decreasing them.
Totally agreed.
No matter how much you practice, some newbies get chance due to this type of contest.
What I felt after the contest, is the mockery of hardwork of deserving candidates, out of five questions you need to solve just one or two problems for entering the auspicious regional sites, for which some people dream and code.In that case what is the point of team contest? You can alone solve one or 2 implementation problems in 3 hrs.Making a long statement does'nt really ensure the quality of the problem, for whatever reasons.
Where can I submit? I do not see submit tab on contest link page.
*codechef :- ahhh that's not something we do here
u cant because they are careless enough to not put it in practice section!!
wuw everybody has so many opinions
Edgy
Very similar problem to 2nd problem — REDCGAME.
Codechef 2013 Cookoff-----------------
link — https://www.codechef.com/COOK38/problems/RRGAME
I also feel that judging teams for onsite rounds based on a 3 hour contest containing an unbalanced set of 5 problems is not fair. We misinterpreted the first problem due to unnecessary comprehension. Our bad..!! So we started solving second problem and after solving it, solved the first one. Because of unbalanced contest , we were not able to solve any more problems and so time penalty kicked us badly. Basically, teams are not getting selected based on knowledge, but getting selected just based on speed and accuracy. Had there been more problems and balanced contest, teams could be judged based on no. of problems solved which is more indicative of a better team and time could be used as a tie breaker only. (around 300 teams are getting selected based on time and accuracy only which is huge..).
How can Codechef expect India winning gold if most of the deserving teams get eliminated in the online round itself?
"How can Codechef expect India winning gold if most of the deserving teams get eliminated in the online round itself?"
How the fuck is a team deserving if they can't even solve atleast one of the last 3 questions? You do realize that if this set had come in world's it would have probably got fully solved by atleast half the teams in a couple of hours (using 1 PC for 3 people)? Winning gold. LMAO.
I agree completely, but this was not supposed to be WF filtering.
In India, most people go to WF in their later years of college. Suppose a team in their 2nd year has been practicing for more than an year doesn't get to go to regionals whereas some newbies who started 2 months before, get to regionals, because there were no problems that could distinguish between the 2 teams. What do you think will keep them motivated to keep practising so they actually become a WF deserving team in the next 2 years?
You are right. I hadn't thought of it that way.
I'm not from India regional, so just want to ask:
1 problem will be enough due to teams per college limit
1 — Depends on the regional. There are 4 regionals this time, and they take 225, 160, 85 and 125 teams. Typically around 90% of the teams are selected such that the institutes are unique, and the remaining based on merit (like for ex, in one of the regionals, out of 125, 116 will be unique institutes and 9 with no such restriction). More details here — https://www.codechef.com/icpc/2019
2 — Some teams qualified by solving just 1 problem, some failed to qualify even after solving >= 2 problems. (I have seen cases where people solving 2 didn't qualify, not sure about > ).
I can answer that:
There are 4 regional sites each admitting around 100 — 200 teams each. There is a limit on the number of teams per university qualifying to each regional.
Since the regional sites try to maximize the number of teams from distinct universities, it depends on which University you are from. For some universities, solving 1 problem would be enough, for others solving 3 would not be enough...
So that translates to: "Random college, Solved 1 problem in the online round have a shot at the world finals, but a good college's second best team in the online round have no chance."
Add a funny problemset, with complex proofs and comprehensive meaningless problem statements to reduce the load. (Results in shuffling teams with high probability)
Well, no complains, being from India can't whine about reservations over merit anymore.
Icpc rule is worst ... Deserving and hard working teams will not qualify. [ "Awareness" wth those who have applied for the online contest, they were already aware of this. Sending them to regionals will not increase their algorithmic knowledge.]
Then there should be online rounds for the entire globe. Top Team will reach world finals. In that case, I think even top college from India will fail to even qualify for finals like code jam. Please tell me what do you think Why they are hosting Regionals??? Believe me or not Sending them to regionals increase their knowledge.
Yeah...the rules don't make much sense to me either...
"How hard is it to set a balanced contest? Codeforces seems to does it every week. CodeChef fails to do it even in its most important yearly contest, along with its monthly short contests, every single time."
Cannot agree more Ashishgup!
Sick of hearing things like
We are all tired of fake accounts commenting as well. If you cannot hold an objective discussion of the round, or read the above comments (that monitor was a spare non-working monitor), please refrain from commenting :)
that
snot
poss
ible
and just so you know I upvoted rajat1603's comment
This blog is too lengthy, I haven't read it all, but just looked at some posts. I need to say something..
I didn't see the problems, but I don't like that authors are getting too defensive. If you can't take criticism for your problems & try to improve the concerns, because perhaps you made a wrong judgement, then don't make contests and stick to solving them.
Our team solved two problems , our overall rank is 310 and college rank is 4 . We filled only amritapuri site (it was our first attempt ) is there even a little bit chance of being selected ?
Sadly No :(
Although Amrita is known to call around 50-ish more teams than they initially decide (225 this year I guess) by releasing subsequent lists of qualified teams, last year the last teams to be invited were somewhere around 140-150 ranks.
These ranks "140-150" Are they for the institutes with 1+ teams going to Amrita site?
Yes, last time they somehow fitted all distinct colleges in their initial promised number [increased it by 10 or so to accommodate all distinct colleges] and then released 4 odd lists of 15 teams each and this was done on merit basis with the 10 per college cap, the 4th list went till 140-150, so all teams <= 150 and <= 10 in their college qualified.
I want my money back :P
Is it possible to get money back?
What if the first ranked team from a college decides not to participate in an onsite regional. Does the second ranked team from the same college gets the chance? Is there some policy in this matter? Since ACM seems to be so considerate about bringing participation from maximum number of colleges, I hope this should happen!
Some kind of moderation is needed throughout India for ICPC contests.
ACM-ICPC participants are entitled to huge discounts on our certification exam! Members of any team that solves all 5 problems can take the Foundation, Advanced or Expert exam for just? Rs.500, that’s a maximum discount of? Rs.4500! Solved 3-4 problems? Take Rs.1000 off! Even participants who have solved 1-2 problems can get? Rs.500 off the price.
How do you reduce the number of discount coupons given away?
1) Hard Problems
2) Hold Right There Sparky!
Solving 2 problems really fast was almost equivalent to solving 3 problems.
Top 9 contributer -> Ashishgup.
Here is my summary of the online round:
Que 1: Very easy straightforward question as expected to boost the confidence.
Que 2: Easy-Medium and solvable, if you consider yourself deserving team then there should not be any excuse why you were unable to solve it.
Que 3: Question in which recursion is used, most teams thought that it will give a TLE and were trying some alternate approach, but if given some time and thought carefully it can be seen that it will be solved without TLE by removing some cases.
Que 4: DP problem, can only be solved by users with good DP practice. Its level is somewhat like "Rated Div 2 problem E" which is very hard for Div 2 candidates but for Div 1 candidates its there "problem C" which is solved by RED coders easily and left unsolved by many purple coders like me.
Que 5: Very hard, as expected one problem is kept hard so that red/orange coders should not finish the contest very fast.
My team was able to solve 3 questions.
I agree with many users on one point that online round should be kept similar to onsite regional round with 10 problems and 5 hours to solve them so that deserving teams can at least solve 3-4 problems and solving just one problem very fast should not be a criterion for judging any team for the onsite round.
As the author of the last problem, I'll disagree with you calling it very hard.
Required knowledge:
Basic knowledge of treaps (there are tons of tutorials online showing how to do array reversals using treaps. This question just involved a minor modification to the same.)
The rest was just simple observation of how the first deck behaves upon modifications. tldr: It always has the form [-1, -1, -1, .... 1, 1, 1] .
most IMO problems dont require knowledge of anything beyond basic high school mathematics and their solutions often rely on 'simple observations'. would you call those problems easy?
I never called it easy. However it definitely isn't super hard.
Okay, I never thought of treaps and was thinking different approach, maybe that's why it looked difficult to me, also time was the factor. Thanks for the info.
If you know treaps,5th question is a straight forward. So I rate 5th < 4th in terms of thinking.
looking at the success rate, it seems only 1 team out of 3000 knows about treaps.
I dont think so.
Maybe some people got stuck solving D and not looking at E at all (this should be less probable).
Maybe someone from their team looks at it who does not know treaps and suggests other members it involves convolutions and some range updates. So maybe some fft with sqrt decomp and they believe solving dp looks reasonable to expect.
Maybe one of their teammate who knows one of special properties of treaps is reversing segment yet missed something in the statement (like missing first deck has only 1 and -1) and felt it involves some convolutions and treaps(obviously) and suggests to try dp(this happened with me and luckily since it was online contest and not regional nothing to cry about though be careful).
These are some scenarios. But I believe an estimate of atleast 25-30 teams to atleast know treaps (atleast one of their members.)
Very, very few people know about treaps, it's quite an advanced topic. Even less know how they're used — for example, I only knew they existed until 2 years ago.
I was justifying that 2 was very very small than expected. I think 25 is also less when there are 3000 teams participating.
I agree that two years back it may be a advanced concept. But I think nowadays it is a well known because it has very good resource to learn Here. Also I am able to see close to 4 to 5 questions on a yearly rate in recent times.
99% of teams being unable to solve problems using treaps at all doesn't seem like much of a stretch to me. Then you may have some who kind of know how but haven't actively used it enough to notice it can be used in this problem, or aren't able to write a working treap in contest time. There aren't many ready implementations (I have one but I don't think it's well-known) and modifying one is a point of failure too.
Considering I've seen almost nobody asking about treaps, it's pretty much as much of an advanced topic as it was 2 years ago. Time doesn't equal understanding, use equals understanding. Look at the number of views on that Quora answer. How many of these views are different people? How many actually read through it all? How many only took the basic idea without giving a damn about implementation? How many actually competed in this ACM round? It may have taught one contestant enough to solve this problem for all we know.
Plus it's the Indian CP community.
In my opinion if someone knows how to use it but doesnt know implementation its not more than 5mins to get a implementation on the web. Also if someone has atleast once coded it before then I think he can easily grasp the code on web and adapt it to this particular question which doesnt need much different implementation than the standard lazy propagation treap.
Also in the question, the update operation is reverse a segment which atleast for me was the introduction point to treaps. I thought reverse segment shouts treaps if you have tried to study treaps before.
Also regarding views, it is 32K , Lets say 5K distinct read it. Since it is post by an Indian and community has a good percentage of Indians. I thought atleast 10% might be Indians who read the post which will be close to 500. Now my expectations were around 25 to know it so that they could recongnise the tag of treap on the problem.
Anyways either way I am happy that I know treaps, And if it is an advanced concept I am even more happy to feel it simple to understand.
Considering how many times I have failed on data structures I've written in the past even if I had past codes available, and I'm red, I can tell you it's harder than you think.
But Quora is a global site. Anyone can read it even by coincidence.
Anyway. Sure, let's say 500 random Indian CS students read the whole thing and 25 understood it well enough to solve the problem. How many participated in this contest? Don't claim that the people who understood it are likely to participate in it, we have no idea if that's true.
The simplest way would be to just ask the people reading this blog (or Codechef Discuss) if they knew that question, if they could solve the problem with treaps and if so, why they didn't.
BY community I meant programming community and not quora community.
Yes I missed to consider how many participated in the contest.
I agree with you.
The simplest way would be to just ask the people reading this blog (or Codechef Discuss) if they knew that question, if they could solve the problem with treaps and if so, why they didn't.
I hope both of us arent interested in doing this and also have many better things to discuss.
Codechef's solution for the online round problems In the problem Word Grid, their solution has an expected complexity of 10^8 per test case. So for 50 test cases, the complexity becomes 10^8 * 50, which is 5 * 10^9. So how can this solution get accepted within the time limit?
" Due to my dislike of CodeChef discuss "....Agreed
https://discuss.codechef.com/questions/138367/why-was-acmind18-so-unbalanced
3.2k teams * 1000 = 32lacs INR...Don't know why they are collecting so much hefty amount for organising an online contest and an onsite for just 500 participants
And after this they can't even set a balanced contest...One of the worst contest ever:(
Ashishgup Do you even code bruh?
He almost ranks less than 150 in codeforces rounds. So i think, you have never seen him on ranklist. He is almost, a pro coder :) Btw, do you even code bruh?
So you created this account just to post this shit. Great
Any other platform other than codeforces which has quality questions. Any recommendation!?
AtCoder and CSAcademy and CSES problemset