I invest in three types of deals. The first group are founders I know well; in this business founders are everything. I am very comfortable backing superstars based on little more then their involvement (though that rarely happens as they almost always have a great idea.)
The second group is comprised of companies in sectors where I have a great deal of experience such as ad tech, marketing, email, enterprise saas, some consumer tech and data. I look hard for original, defensible ideas that have the potential to become huge. I would rather lose on a bunch of companies than to miss out on one of the big ones. This is a business of swinging for home runs and being comfortable with striking out. Cruise Automation is a great example of this strategy. When we invested in Cruise a lot of folks thought it was a weird investment; six months later it was sold to GM for a billion dollars. Also, one of the greatest advantages of AngelList is that it enables significant diversification to have a better chance of getting into one of the winners. I don't consider myself an index investor (in the public markets I had a 30.9% IRR over the seven years I was active), but I love the value of smart diversification.
The third type is where I can spend a great deal of time with the entrepreneur and build a longitudinal understanding of them as individuals and the dynamic of their markets. As Mark Suster correctly points out, it is much easier to invest in lines vs dots.
Here are a couple podcasts where I discuss my investment thesis and approach:
As a ten year startup CEO I built a very large network of founders, angels, VCs and friends. All of my deal flow comes from those relationships.