For those who might want to apply Neural Networks to a multiplayer game such as CSB, here is a set of 100k datapoints from which you may learn.
The set was generated using my main (top 3 at the end of the contest) bot, under the asumption that there is no collision with the opponent.
My Neural Network bot based on this exact dataset is currently 28th / 60 in legend league.
Other than that, here are two very interesting resources on Q-learning (unsupervised reinforcement learning) :
https://webdocs.cs.ualberta.ca/~sutton/book/ebook/the-book.html
http://ml.informatik.uni-freiburg.de/_media/publications/rieecml05.pdf
I have yet to make anything work in the unsupervised learning category, and would be very interested to know if anybody succeeds !