Challenge Your Assumptions
I've been on a small data analysis kick lately. It was great news when a recent race I ran freely shared the race data in a nice plain-text format (http://www.flrrt.com/results/sea11.txt). I converted this into a CSV file and read it into R. Now, just looking at the crowd, I would have guessed that there were a 'few hundred' people, the bulk of which were men. In fact, there were 969 runners overall with nearly an even split between men and women (480 female and 489 male - that's a pretty insignificant margin).
I would have also guessed that the 20-30 age range would be the most well represented. Not true - In fact, the 40-60 age bracket made up half of the runners:
The youngest runner was 7 years old, the oldest was 92, with a median age of 40.
So, overall, I was fairly far off in my assumptions. However, I know that with practice, I'll improve. This goes for anything, including estimating tech projects, playing musical instruments or finding bugs in programs: the more you do it, the better you become. Also, don't forget to try new things and challenge your assumptions.