Twice the time from start to 50% complete.

HA. Yeah right. It never works like that. Not even close.

I just read a really great (buy quite long) post about estimates and why we suck at them.

I will TLDR; this into a few bullet points here, but i really think you should read the whole thing.

• Our brain has 2 ways to estimate, quick gut feel based on patterns/experience and methodical expensive based on slow thinking
• Both suck
• Big estimates (e.g. 2 weeks, 2 months) have almost no chance of being right
• Small estimates (8 hours or less) have a better chance // I’ve seen this first hand.
• Small estimates let you get fast feedback on how bad you are
• You WILL still get your small estimates wrong
• But, short sprints place a hard limit on the cost of a horrifically bad estimate
• Agile rocks

http://blog.hut8labs.com/coding-fast-and-slow.html