In my personal experience(Meritorious both years I participated), one of the more obvious but important aspects is overlooked by many teams:
Our team began both competitions by selecting a "Coder, Writer, and Researcher". This should play off your strengths, but essentially, the code was substantial enough that our coder did only that. The Researcher spent the first half of the competition finding out what had been done and communicating with the Writer about the introduction and direction of the report. The writer spent the first half of the competition helping write the algorithm with the Coder, and working out the mathematical models to be used. Also the Writer attempted to synthesize the work of the Researcher into the start of the report. The second half of the project was a bit more hectic, the researcher obtained as much information as possible that might be available and useful. This was particularly important for important real world data into our models. The writer spend the second half... writing. Literally just writing non-stop.
This worked fairly well for us. I also recommend spending the first 1-3 hours letting the members think individually about the problem, then spending about 30-60 minutes explaining to one another their attack. Sometimes people aren't forthcoming with really great ideas.
Another recommendation is SLEEP! The first year we hardly slept at all for the entire competition. We drank insane amounts of caffeine and worked through the weekend. The second competition, we slept about 6 hours each night. We got the same ranking both times, but we did better the second year. Looking back, the quality of the writing was better. The code had less bugs, and we managed to produce more mathematics. Don't underestimate the power of getting a good nights rest.
Another general reference is consider a spherical cow which is a fairly standard reference. It is fun to read, and somewhat enlightening.
Perhaps I will think of more later, but for now this is all.