With high-tech companies needing less capital due to advancements in technology, startup development methodology and online marketing, we have seen a Renaissance in angel investing. While angel investors participate in part for the excitement of engaging with entrepreneurs and placing bets on the future, they also do it for the expectation of significant financial returns. Various studies of angel investing published in the last decade estimate aggregate returns to angels on the order of 18-37% per year, well above market. The catch is that 50-70% of angels make less than what they invest. Returns are very unevenly distributed and this begs the question to what extent is portfolio theory fundamental to angel returns.

The best data set with detailed investment & exit information comes from the Angel Investor Performance Project by the Kauffman Foundation. The data was collected by surveying angels who belong to angel groups. Cleaning the data and restricting to the domain I was interested in—first round investments in early-stage high-tech companies—yielded a data set about the returns of 56 angels with exits from 112 companies. The data show the type of skewed distribution one would expect from early stage investing:

• 75% of exits happened between 2001 and 2006. There is some reason to believe that the data may have a slight bias towards negative returns as 50% of investments happened between 1995 and 2000. Angels may have been buying high and selling low.
• 3.2x cash-on-cash return for all investments put together (total dollars out divided by total dollars in). However, returns are extremely sensitive to big hits. A lucky angel put $600K in a software company in three rounds from 1988 to 1994. In 1996 the company went public and the person got a nice 55x return. Removing this one company from the sample drops the aggregate cash-on-cash return for all angels nearly in half to 1.8x.
• Of the companies angels invested in, 63% were complete write-offs for the angels involved.
• 66% of angels made less than what they invested. 45% generated no return. The remaining 21% of angels received only 4% of the total returns (7% if you exclude the 55xer).
• 6% of the angels generated returns >10x that accounted for 68% of the total return (42% w/o the 55xer). The cash-on-cash return for that group was 36x with and 21x without the one big hit, in both cases more than ten times the average for all angels put together.
• The data includes only one super angel who had 29 exits generating 2x return. Most other angels had one or two exits and only a handful had three or four.
• Due to missing or overly granular investment and exit dates, it is practically impossible to calculate meaningful IRR numbers or to calculate returns in excess of financial markets.

The analysis suggests that angel investing as a whole can be quite profitable but, when dabbled into a deal or two at a time, it is more akin to gambling.

Without accurate data about angel investment portfolios, the next best option is to do Monte Carlo simulations of synthetic portfolios where thousands of hypothetical angels invest in thousands of hypothetical companies. The hardest part in setting up Monte Carlo studies is making good assumptions as they can pre-determine outcomes. Some have approached the problem by guessing probabilities of certain outcomes much in the same way VCs do basic portfolio presentations for LPs but with a bit more math in the mix. Rather than guessing, I chose to reverse-engineer a distribution of returns based on the data from the 112 companies. For the math-inclined amongst you, this involved piecing together a cumulative density function from three separate pieces: 60% chance of zero return, a logarithmic non-linear model for 0-10x returns and a combination power/exponential non-linear model for the long-tail of exits greater than 10x where not much data was available.

I ran a very simple Monte Carlo simulation evaluating the portfolios—ranging from 5 to more than 100 companies—of hypothetical angels. The average cash-on-cash return was right around 3.2x, exactly as with the Kauffman data, which is a good sanity check. Average returns don’t vary with portfolio size, which is to be expected.

Median returns vary substantially with portfolio size. Going from 5 investments to 10 investments increases median return by 68%, from right around 1x to nearly 1.7x. There are diminishing returns to growing portfolio size. Going from 10 to 15 increases median returns by another 40%. Doubling portfolio size from 15 to 30 adds another 50% but then in takes going all the way to a whopping 125 company portfolio to triple median returns compared to the 5 company portfolio. Similar conclusions apply with respect to other metrics. The probability of getting a return that’s greater than 2x doubles (from 34% to 69%) as one moves from a five company portfolio to a 50 company portfolio.

The data unequivocally suggest that playing like a super angel or an active seed fund as opposed to dabbling with the occasional angel investment is a key strategy to consider if financial returns are important. The data also call into question the behavior of some angel groups that do just a few investments per year.

This is not to say that volume investing—like throwing darts to pick stocks—should replace doing due diligence and the thoughtful development of investment theses. In fact, every Monte Carlo simulation of angel or venture investing I’ve seen, including mine, doesn’t take into account the various types of signaling that go on between entrepreneurs and investors and between investors themselves. For example, great entrepreneurs usually have over-subscribed investment rounds. A pure volume-oriented investor would find it difficult to compete for and win these hot deals, especially in a world where seed funds keep popping out like mushrooms after rain.

If you want to know more, let me know. I’m always curious to hear your thoughts. You can find me at @simeons or at FastIgnite.

Simeon Simeonov is founder and CEO of FastIgnite where he invests and helps entrepreneurs build great companies. Sim is also executive-in-residence at General Catalyst Partners and co-founder of Better Advertising and Thing Labs. Prior to that, he was a VC at Polaris Venture Partners and chief architect at Allaire/Macromedia (now Adobe). Sim blogs at blog.simeonov.com, tweets as @simeons and lives in the Greater Boston area with his wife, son and an adopted dog named Tye.