How To Screw Up

1. Promise a prominent professor that you'll do some numerical experiments for him.
   
2. Use local supercomputer to do said experiments.
   
3. Do not at any point verify that the experiments you think you're doing are the experiments that are actually being performed. 
   
4. Announce that the experiments are completed.
5. Perform a rudimentary analysis and observe that most of the data is useless.

When I was an undergrad my professors always told me to analyze data as it was being generated so that I could avoid precisely the above situation.  Luckily the math department has plenty of time on the supercomputer so I can run the experiments I need, but it's going to take a while.