Author Topic: Statistical aspects of portfolio diversification  (Read 899 times)

Graham Osborn

  • Sr. Member
  • ****
  • Posts: 343
Statistical aspects of portfolio diversification
« on: May 15, 2018, 06:15:19 PM »
Hi folks, I thought some of you might have feedback on this.  The talk is on angel investing but I think the conclusions are applicable to "growth" investing as well.


rb

  • Hero Member
  • *****
  • Posts: 2614
Re: Statistical aspects of portfolio diversification
« Reply #1 on: May 15, 2018, 06:40:54 PM »
I went through it. The presentation i kinda shit. But the statistical aspect is actually correct. Basically in angel investing despite all the bs around of people who profess to see the future you don't know what each individual company will do. You will have a whole load of zeros and (hopefully) a few that will rocket up to the heavens. In the end you hope for a good result. Diversification is necessary because the level of confidence in each investment is low.

I guess you could make a case about growth investing if you try but I'm not sure it'll be right. Basically when you talk about public growth investing your uncertainty is not about the company/business. Your uncertainty is about the valuation. I don't think that diversification will save you here because the companies that rocket to the havens will already have a very rich valuation to start with. Whereas in the angel model you get them for regular valuation.

Graham Osborn

  • Sr. Member
  • ****
  • Posts: 343
Re: Statistical aspects of portfolio diversification
« Reply #2 on: May 16, 2018, 08:19:24 AM »
I guess you could make a case about growth investing if you try but I'm not sure it'll be right. Basically when you talk about public growth investing your uncertainty is not about the company/business. Your uncertainty is about the valuation. I don't think that diversification will save you here because the companies that rocket to the havens will already have a very rich valuation to start with. Whereas in the angel model you get them for regular valuation.

I agree with you in the sense that growth investing is more of a buy-at-intrinsic-value-and-wait-for-intrinsic-value-to-increase game vs a buy-at-a-discount-to-present-intrinsic-value-and-wait-for-the-differential-to-close game.  But still, your valuation in these cases heavily depends on how sustainable the company's growth is over the next 10-25 years.  So uncertainty about the business and uncertainty about the valuation are 2 sides of the same coin.

It's also true that such companies are more often overvalued than undervalued, or at least that has been the case since the 1970s.  But that is very different from saying they are never undervalued.  The early 90s (Microsoft) and 2008-2012 (Google, Apple, Netflix) were times when you could have bought businesses like these for very attractive prices.
« Last Edit: May 16, 2018, 08:22:41 AM by Graham Osborn »

rb

  • Hero Member
  • *****
  • Posts: 2614
Re: Statistical aspects of portfolio diversification
« Reply #3 on: May 16, 2018, 08:44:00 AM »
I didn't mean that growth stocks are never undervalued. What I was saying was that they won't be undervalued to the level of undervaluation you would encounter in an angel situation. You won't buy an undervalued stock that will end up going up 100x or 1000x.

LC

  • Hero Member
  • *****
  • Posts: 2583
Re: Statistical aspects of portfolio diversification
« Reply #4 on: May 16, 2018, 11:33:20 AM »
50-150 bets @ 50K each
2.5M-7.5M capital
Every company needs to have the potential to grow at 50%+ annualized for 10 years
Only asset-light businesses with high gross margins and extremely low startup and customer-acquisition costs can be considered

I think the insight here is that, there is a huge market out there and the business has the opportunity to gain a lot of customers easily, but so does everyone else. The trick is that the investor can buy stakes in ALL the major competitors to put in their angel-bucket, so eventually it will work out

Valuation:
For pre-revenue businesses: EV=𝑔/10∗𝑁𝑝
For post-revenue businesses: EV=𝑔/10∗𝑃
Variable definitions
EV = enterprise value (needs to be adjusted for debt, senior securities, etc)
𝑔 = growth rate (e.g. for 50% growth, 𝑔=50)
𝑁 = number of uses per year
𝑝 = gross profit per use
𝑃 = total gross profit, if available

Is this the modern version of net-net investing? Instead of buying a basket of net cash Crapco.'s, buy a basket of pre-revenue Dreamco.'s.
« Last Edit: May 16, 2018, 11:37:37 AM by LC »
"Lethargy bordering on sloth remains the cornerstone of our investment style."
----------------------------------------------------------------------------------------
brk.b | cash

Graham Osborn

  • Sr. Member
  • ****
  • Posts: 343
Re: Statistical aspects of portfolio diversification
« Reply #5 on: May 17, 2018, 01:21:27 PM »
I didn't mean that growth stocks are never undervalued. What I was saying was that they won't be undervalued to the level of undervaluation you would encounter in an angel situation. You won't buy an undervalued stock that will end up going up 100x or 1000x.

For Google and Facebook that has definitely been true.  Microsoft is close to a 1000-bagger since IPO I think, before adjusting for inflation.  There's definitely a bias for companies to go public later, which has taken some of the opportunity out of growth investing for the little guy.

Jurgis

  • Hero Member
  • *****
  • Posts: 4017
    • Porfolio
Re: Statistical aspects of portfolio diversification
« Reply #6 on: May 17, 2018, 01:28:27 PM »
I didn't mean that growth stocks are never undervalued. What I was saying was that they won't be undervalued to the level of undervaluation you would encounter in an angel situation. You won't buy an undervalued stock that will end up going up 100x or 1000x.

For Google and Facebook that has definitely been true.  Microsoft is close to a 1000-bagger since IPO I think, before adjusting for inflation.  There's definitely a bias for companies to go public later, which has taken some of the opportunity out of growth investing for the little guy.

The issue with 100x baggers - as with any investment - is really how long you hold and when you sell. The survival bias on success stories like MSFT, GOOGL, FB is huge. And there's a lot of growth tech companies where you would have made great returns if you sold at the right moment. And if you held past that, you didn't.

From what I understand, this might be a bit less of an issue in angel/private co investing. But even there the timing of exit may radically change the return.
"Before you can be rich, you must be poor." - Nef Anyo

Graham Osborn

  • Sr. Member
  • ****
  • Posts: 343
Re: Statistical aspects of portfolio diversification
« Reply #7 on: May 17, 2018, 01:30:45 PM »
Is this the modern version of net-net investing? Instead of buying a basket of net cash Crapco.'s, buy a basket of pre-revenue Dreamco.'s.

That's an interesting concept.  I remember at one point Buffett called Graham "quantitative" and Fisher "qualitative".  Of course, these are relative terms.  When you get into early stage situations the uncertainty goes up dramatically.  Buffett was very much opposed to angel/ venture investing, but as I thought about it I realized that it didn't fit the statistical construct of his portfolios at BPL and Berkshire.  Buffett liked positions he could max out to 20% of assets or so, and that just isn't something you can do in early stage.  Additionally, for most of his life Buffett has had so much capital that deploying in a low-capacity strategy like angel investing would have been a waste of time.

But to your question, yes I think there might be an analogy.  The difference is that with a net-net there is some kind of margin of safety (at least in the short term) to the downside, and you can be fairly quantitative in what that is.  In startup investing, the value of a pre-traction business is at best an order-of-magnitude calculation.  There is really no equivalent margin-of-safety concept at the individual-security level, except that obtained by keeping the pre's below 5M inflation-adjusted.