I regularly compete in so called "Programming Contests", where you solve difficult algorithmic problems with your own code and problem solving skills during a limited time-frame. For referential examples of what those might look like, search for competitions like e.g. Google Code Jam, or ACM-ICPC.
(If you know what programming contests are, you can skip the following paragraph)
In these contests, you can compete either individually or in teams, at a physical site or online, and the goal is to solve as many problems as possible before the contest is over. Each problem is associated with certain constraints which your solution must fulfill, such as running time, utilized memory, etc. The problem difficulty can range from "obvious how to solve" to "demanding great knowledge or heavy ingenuity to crack". The main goal is of course to have fun, but success may in some cases lead to cash prizes, honor, and perchance even an interview with recruiters of top-notch companies like Google.
(If you skipped the above paragraph, stop skipping, because here comes my question)
Through programming contests I have been able to develop a genuine interest for algorithmics, computer science in general, and an overall sense of meaning when it comes to coding and developing my problem-solving skills. I would like to do so in scientific computing as well.
Question: Do there exist any competitions in the same vein as the purely algorithmic kind, but with a focus on numerical analysis, optimization et cetera?
I tried Googling, but couldn't find any at first glance…
Addendum: I got the suggestion to check out Project Euler, but that isn't really what I was looking for. What I want is more an arena to practice the more "dirty" kind of numerics, like getting creative with solution schemes for pde's, solving numerical optimization problems, et cetera. Just for comparison, Kaggle is a site where you can practice data mining, machine learning etc. in various contests held regularly. I want something like that, but for implementing and getting creative with numerical schemes and the likes.