It's a very difficult contest. Often times a fun sounding problem is hiding a state-of-the-art unsolved problem (or NP-complete problem) in computer science. It

It's a very difficult contest. Often times a fun sounding problem is hiding a state-of-the-art unsolved problem (or NP-complete problem) in computer science. It's easy to spend a day working on a feature that could be done in an hour if you have the library on hand or recognize the appropriate algorithm. At the very least, like a startp, haveing a partner can save you from wasting time on dead-ends and little bugs. Teams that do solve much of the task problem, or fare well against the bots (for competitive tasks), tend to be either researchers or graduate students studying the same field as the task's core problem, or teams of 5-15 people, or both! Sometimes (on the more puzzle-oriented tasks like CBV or Save Endo) have hidden tricks which you might just not notice. (To their credit, those tasks had a lot of breadth, so there was plenty to do even if you missed the tricks, and if you didn't care that the trick was worth an outsize amount of points (like the Windmill in Save Endo). I have a bad habit of using ICFP as an opportunity to learn a new language, and then waste a lot of time just learning the basics and system configuration. To make progress on the task, you need to work in a language that you already know like a part of your body. You can easily spend several weekends playing with a task post-contest.