17

I sometimes do a programming contest in the same house with other people, in different rooms but behind the same NAT. That's not forbidden, shouldn't be forbidden, but are from the same IP.

Maybe they have trained together, and maybe they are using the same template and contest libraries. But still, one can imagine a superhuman being that can distinguish those cases.

Run your program with Valgrind with a small input, you'll probably find you're reading something out of bounds or uninitialized. Usually that happens to be a 0 in your PC but random data in CF, so it seems to work but doesn't

I'm not following at all. Rating difference of 0 means 0.5 win probability, rating difference of 200 means 0.76 win probability, rating difference of 400 means 0.91 win probability, rating difference of 600 means 0.97 win probability. How's that "X times better, +Y rating"?

Rating system explained: http://mirror.codeforces.com/blog/entry/20762

The rating system is not a logarithmic scale, it's a linear (or affine, to be precise) interval scale. The difference of ratings is what is meaningful, and the rating 0 doesn't have any specific meaning. Also the TopCoder page you link to explains just a weighted arithmetic mean.

One does not simply not care about t-shirts.

One doesn't simply walk into Mordor from Westeros.

The contest will be held as a combined Div.1+Div.2 round which will be rated for participants from both divisions.

Is this just a single event? If so, why is it called a marathon when it's just 2 hours?

Will Div 1 and Final Round results be combined for the rating calculations or are those calculated separately?

It's hard to help if we don't know what you are having problems with. Could you give an example of a problem where you "can't maintain the base case" and explain what you have done?

+1 This is very similar to the Glicko rating system used by many online chess sites.