What if the squirrel could see?
The most effective investment strategy is a highly un-diversified portfolio when you are right.” – Jim Collins, From Good to Great.
I recently appeared on TwentyMinute VC podcast and declared that highly concentrated portfolios are the best way to invest as a venture capitalist. This started a debate between concentrated and diverse portfolios in venture capital.
Matt Lerner at 500 Startups in cooperation with Yannick Roux recently produced a Monte Carlo simulation of various portfolio sizes and shared the results here.
First off - genuine thanks to Dave McClure, Matt Lerner, Leo Polovets, Josh Hannah, Paul Arnold, Harry Stebbings, and Yannick Roux for engaging in a healthy debate. Dave and Matt make a compelling case for diversification, especially at the very early seed stage. But, I have a different point of view.
What if the squirrel could see?
The Blind Squirrel theory (hereafter referred to as the BS theory) states effectively that all that matters is getting in 1 or more of the mega-winners in venture capital AND the best way to do that is to make LOTS of bets.
The Monte Carlo simulation results seem to prove that and I have no debate with the simulation. My issue is with the inputs and underlying assumptions in the model. Specifically, I feel the BS theory works in theory, but fails in practice because of the following 4 fallacies.
1) Access Fallacy - the BS theory can only work if an investor has access to invest in any and all private companies. This is fundamentally why Vanguard works in the public markets with unrestricted access and why there is no Vanguard of the private markets. (Just FYI - Vanguard now has 2T under management). I can’t simply look at a database of startups and write 200 checks to startups for 100K each for a 20M fund. First I need to source all of these investments and then I need to ensure that I can get access to invest. In many cases, great startups are oversubscribed and turn investors away. That said – for firms like 500 Startups and YC, this may be less of an issue because they are frequently the very first check into a company and companies have in fact applied to join them.
2) Selection Fallacy - the BS theory seems to suggest that we have no idea what we’re doing when we select companies. Honestly, sometimes I feel like this is true. But if it were really true, there would be no investors that consistently pick winners year after year for 20+ year careers like my idols Warren Buffett and Charlie Munger. Or, why does SV Angel have 16 unicorns out of 628? That’s a 2.5% chance to unicorndom which is WAY better than industry average. Surely it’s not just luck.
3) Stewardship Fallacy - the BS theory also assumes that investors have zero impact on the outcomes of companies. It rests on a pure mathematical model of picking stocks. This is way off base. I am currently on 6 boards, and I can tell you there is a very wide range of ability, advice, expertise, patience, and ideas surrounding the board room table (I’m also not saying I’m the best – just that there is a wide range). Since there is such a wide variety, and we all believe that investing in teams is critical, it seems logical that the “board team” is also a factor in the success of startups. However, in the BS theory, that doesn’t matter. My belief is that being involved in a smaller number of companies allows an investor (and sometimes board member) to contribute more to each individual company. There are no economies of scale by spreading your time and attention very thinly.
4) Capacity Fallacy - the BS theory basically assumes that the numbers and ratios are constant no matter how many companies you invest in. Taken to a logical extreme, you might say that you should invest in 100,000 startups. My guess is that if you ran the Monte Carlo simulation for that, you’d find that it performs well and probably has a lower standard deviation than the 500 startup portfolio. But the problem is that the universe is limited. There are not 100,000 available startups. And there is no way that 1-2% of them (1-2K) would deliver a 50X return. There are only a small handful of startups in the world each year that produce 50X returns.
As Paul Arnold points out:
I prefer the Seeing Squirrel theory (SS theory). The Seeing Squirrel tries everything in their power to change the inputs. The only way to drive meaningful alpha in the asset class is to have different inputs. Either increase your odds of getting a unicorn or decrease your odds of getting zeroes. And within each bucket, maximize your returns and minimize your losses.
These are the model inputs that Matt Lerner and Yannick suggest:
We really try to do much better than this at JSV. Over the last 14 years of professional investment activity, we’ve had a loss rate of just ~25% while we’ve had 10% of our investments reach over 1B in value. Note, these are not necessarily 50X investments because of dilution + follow-on investments at higher valuations, but we’re all quite happy with the returns. The high “hit rate” for the amazing outcomes is largely due to patience, selectivity, and stewardship. We invest at a bit later stage than a seed fund as well, but it’s still very early stage with median valuations around 10M-12M pre.
Matt Lerner declares:
We have an expression “Even a blind squirrel finds a nut every once in a while.”
My preferred expression is:
“Who do you think is going to find more nuts, a blind squirrel or one that can see?”
——-
PS - the concentrated portfolio approach works for us at JSV, just as I’m sure the diversified approach works well for Dave at 500. I have tremendous respect for what Dave has built at 500. But, they are fundamentally different approaches that require different skill sets, networks, stages of investment, and discipline. Much of the difference in opinion and approach stems from differences in stage of investment. In early seed stage investing, there are more startups, fewer breakouts, and less data. Availability and access are likely increased while selection is arguably more difficult. With every stage of investment after seed, the number of startups goes down and the data becomes richer.