The first few rounds are usually automatic for me. I take the best value and try to stay ahead of scarcity, but I'm a fan of using several techniques in making decisions in the later rounds.
Often times I'll consider what statistical advantage the player offers over another player.
For instance, you may be deciding between a power 3B & a 2B steals specialist. You have to value how much steals & homers are worth at that time and what the next best option is if you give up on the player.
If 2B1 produces 60 steals, & 2B2 produces 20 (and there is a scarcity of steals specialists in ALL positions), while 3B1 produces 40 homers, & 3B2 produces 30 (and there is no scarcity of HR specialists in ALL positions) then you are giving up 40 steals for 10 HR and on top of that the steals value is inflated due to scarcity.
If you continue to do this, you'll notice that you'll start a lot of trends and cause people to grab at positions earlier than they'd have liked. This means better talent falls to you in later rounds AND you get the best of the scarce talent. I wouldn't recommend doing this in the first few rounds though-- Carl Crawford over Pujols is just something you shouldn't do.
In the later rounds you should use "expected value" to fill in bench spots (but not TOO many-- these are "investments").
Expected value = normal statistical value * probability of attaining that value
It's like risk-reward --many of the other posters touched on it when they talked about long shots. A 30HR, 100RBI season out of Huff might be would be worth the 12th round cost despite his risk.
I appreciate any and all suggestions on my roster.