Thursday, May 17, 2012

World finals - more coverage

I've just discovered a YouTube channel with many nice little videos the ICPC team releases about the finals: Those can be very educational :)

For example this one talks about problem preparation:
Everything sounds logical, the only thing I'm quite surprised about is that they still consider two independent reference solutions to be enough - many other competitions like Google Code Jam or TopCoder usually require at least four as far as I know.

I've discovered this channel via Oleg Hristenko and his World Finals website, which you should also check out: (list of references to other data sources here: (or maybe the automated translation at, although the latter seems a little buggy)


  1. As of lately, TC has 3 reference solutions. Although probably 2.5, because the only testers are admins and they are sometimes overloaded with other contests so...

  2. I know only couple of exceptions when we had less than 3 solutions for an SRM problem, so thing about 2.5 looks exaggerated.

    Also, if we speak about onsite level, it is 5 independent solutions for the last two years.

  3. Yeah.

    So what's the secret of the finals running so smoothly each time? Is it that when you have several months to prepare a solution, you do a better job? Or do they just avoid problems that have un-obvious catches?

  4. Well, 2 is just the "official" bare minimum.  In reality, there are several of us (including me, Per, and Onufry) who test-solve the complete set.  I don't think there's been a problem with fewer than 3 judge solutions in quite some time.

    Also, the judges all have a chance to nitpick every problem statement.  15 judges means a LOT of nitpicking. :)

  5. Thanks for the explanation, everything makes sense now.

    In my experience it was usually the opposite: it's difficult to get somebody else to review the problem statement even if there are many judges, so it's great that the finals' judges are not like that :)