Nothing, everyone can participate in rounds 1-3. In each round, top 30 get some points, and in the end, people with the most points go to the finals.

В некотором смысле ДА. Но это не мешает бороться за финал (сумма ГП баллов) и футболки (лучшее место из 4 отборочных раундов).

Пожалуйста, задавайте вопросы через тестирующую систему.

Авторские решения на java есть по всем задачам, где это существенно.

If two participants have the same amount of points earned(0), lowest rank in all 4 rounds matters. I dont actually understand what happens if somebody skips one round but in others he is like top50, he loses his chance to win a t-shirt?

It has a nice solution, need to make sure that everyone has seen this problem!

That's a great pity for us, but nobody among contest authors and testers have seen any of these problems.

Both me and GlebsHP tried to Google this problem before the contest but we didn't succeed. As you can see, all three versions of this problem have almost nothing common in statements, so it was almost impossible to find them without seeing any of them before.

Okay, and how to solve it? :)

For every vertex, in any order: look at numbers of components in neighbours, choose minimal that is absent.

Proof: In the end, for component number K, there are edges to components K-1, K-2, K-3,... because otherwise we wouldn't be able to assign this number to some vertex.

Будем считать, что деньги мы получаем всегда, а за проигрыш платим a_i + b_i. Зафиксируем, сколько мы заплатим при проигрыше (X). Тогда мы хотим в каждом из исходов просто набрать максимальную сумму, если мы платим не больше X. Напишем 3 независимых рюкзака, затем переберем X.

Скажем, что взять ставку стоит a, и нам дают a+b, если она сыграла. Для каждого типа ставки насчитаем рюкзаки "максимальная сумма, которую можно получить, потратив не больше а". Теперь нам надо в каждом рюкзаке выбрать 1 вариант, мы заплатим за все три и получим профит от двух худших.

Посортим все возможные варианты из рюкзаков и переберем тот, который будет вторым худшим в нашем выборе. Теперь осталось среди вариантов хуже взять тот, у которого оптимальный баланс (за него мы получим + баланс),а среди вариантов лучше — тот, у которого минимальная цена (он не сыграет, и за него заплатим зря). Это насчитывается максимумами на префиксах/суффиксах. Ну и еще не забыть, что надо взять ровно один вариант из каждого рюкзака — можно перебрать, из какого рюкзака брать вариант хуже, а из какого — лучше.

Ну так

По Java по-моему нормально лимит выставлен был, TL не приходилось пробивать.

A,B,C,F под твой критерий подходят. D,E не сдал, не знаю :)

Second pretest is a counterexample. In first CD chooses 4, but you need to choose 3.

Did you use Java7 x32? It worked for my F solution.

And yes, I remember we both had problems with Java in last opencup ;)

Well, it is a bit offtopic but...

How to solve F in O(n) time? Both AtCoder editorial and my solution (slightly different) work in time.

Object creation is expensive in Java 7 x64 / Java 8 x64. The only reliable version is Java 7 x32 (why not add Java 8 x32? I wanna use streams and lambdas sometimes)

FWIW, exactly this issue was discussed during the recent marathon round one week ago. My measurements back then showed that even Java 7 32 on server is several times slower than Java 8 locally.

I don't understand why is it so hard just to roll back to the state several months ago, when everything was working fine... Also, so far I've never seen any Yandex representative even acknowledging that the issue exists, which is mildly frustrating (given that many different Java-users are complaining about it after every single contest).

For the sake of transparency, it would be much nicer if someone from Yandex publishes smth like "we changed <this and this> several months ago, which broke Java, sorry about that, we're going to do <this and this> to fix it".

Hi. Thanks for taking part in Round 1 and sorry for the trouble caused, there is indeed a significant slowing down in java execution on large input data cases.

We prepared the problems in such way that they would be less affected by the java issue, made sure to add problem X for server side testing and notified all contestants about java 7 recommendation at the contest start.

Nevertheless I understand that wasting time on problem X testing and not being sure in your language of choice is annoying.

We put our best effort to this problem right after the opencup round, where the problem was reported, and will do so in two upcoming weeks before the next algorithm round.

I've added assertion to my AC code that times are greater than 0 and it still passes. Probably you have some other problem

Go to passport.yandex.com/profile and change your country (although I'm not sure if it works with FB login).

T-shirt winners and finalists will be determined based on 4 rounds (Marathon and Round 1, 2, 3).

Almost constant I guess (independent of N)

        for(int i = 0; i <= 1000; i++)
	{
		ll x = i + A;
		ll y = N - x;

		if(x < A || y < A || x > N || y > N)
		{
			break;
		}

		x = 5000 - (x % 500);
		y = 5000 - (y % 500);

		ans = max(ans, (x % 500) + (y % 500) );
	}

Replace ai and bi with xi=bi and yi=ai+bi . You can restate problem in the following way: for each bet you take you receive xi and in case of proper outcome you pay yi. Now notice that for the chosen subset of bets, you receive sum of their xi and may lose maximal of three sums of yi for W,D,L respectively. Iterate over maximal loss X and pick best subsets of bets with loss <=X. Your dp suits it quite perfectly if you replace a and b with x and y.