If you start with the hard problems and then do the easy problems, your score will be less than when you started with the easy problems. So the system is rewarding you to do the safe, boring problems.
Today, this worked out really badly for me because I failed the hard problem, problem D. So I only got points for A and B. So my score was less than some of the participants who only solved problem B. http://codeforces.me/contest/672/standings/page/124
If I solved D, I would have got less than 2000 points. That's less than the participants who solved A, B quickly and then took the rest of the 2 hours to solve C. D is a much harder problem than C. http://codeforces.me/contest/672/standings/page/20
And why did I fail D ? I used cin instead of scanf() and got TLE !
Auto comment: topic has been updated by nroets (previous revision, new revision, compare).
Hmmm , No one cares !
In case you are the one who only solved A,B but your rank is higher than people who failed solving D, I think you would be grateful to this scoring system. Anyway, this is what makes CF the best OJ.
That doesn't actually make any sense at all.
If you follow path like this
and are actually able to write all problems correctly, you will be usually higher than anybody who started on problem A even if he manages to write all four problems a little bit faster, that's simply how the point decay works.
Usually people who start from problem A are not sure whether they are capable of solving four problems, or they just don't feel like solving easy problems after hard ones, that is pretty much it, not because of points. There is no any kind of reward for doing this.
And actually considering your timing, it took more than 1:30 to solve problem D for you, I'm not really sure if that was correct decision, because you've actually lost THAT much points, that it would have even been better for you to solve A, B, C
It's not always optimal to start with tougher problems because
There is a chance of failing in tough ones
In some cases it is not the best way.
Let us assume a person solves
A in 5 minutes, B in 10 minutes and C in 40 minutes.
Assuming score distribution is 500-1000-1500.
Then C -> B -> A will give 2450 points.
A -> B -> C will give 2600 points.
It's well known that if the scores decay linearly, you should pick problems by order of greater point/time ratio first.
So assuming a 500-1000-1500-2000 split, it's advantageous to start with D if you're pretty sure you can do it in less than 4 times the time you took for A. This is not the case for basically all contestants, except for some very high rated ones who are sure that they can solve a problem like C or D very fast.
Auto comment: topic has been updated by nroets (previous revision, new revision, compare).
I didn't too well today either. Let's continue to improve, together.