I thought portfolio theory referred to holding a diverse bunch of assets, no? (I could be wrong on this.) For example, an investment portfolio might be comprised of 90% T-bills, and 10% "high risk" assets (yes, I realise the risk of this statement becoming quite false in the not-too-distant future, given what has been said about US debt recently).
It's not classic, but it is a diversified portfolio. YC invests in startups across all industries--automotive, mobile, health, financial, entertainment, etc.--which presumably is less risky (lower variance in returns) vis-a-vis just investing in one sector.
So, it's not really a diverse portfolio, but it's definitely a portfolio play, which is substantially less risky than having all your assets tied up in the stock of a single startup.
Think about it this way, if 2001 hit again and all tech stocks plummeted, all the companies in the YC portfolio would be substantially negatively impacted, so there isn't a portfolio effect to protect against this correlation in YC's assets.
Portfolio Theory is about keeping a set expected return but reducing risk (volitation) through diversification. Diversification is about choosing investments which aren't positively correlated. Are web startups positively correlated?
As a industry, probably/maybe. Was certainly the case in the dotcom crash.
Performance-wise, not really. Most startups fail, some make it big. Why? Low investment costs and high possible leverage with internet technology (to little marginal cost). This could change in the future.
In this case the second way is probably more fruitful to think about, since one big winner will even out all the losers. I'm sure this is possible to "prove" by some Black-Swan-ish statistic model theory.
NOTE: I could be wrong.
EDIT: I agree that web startups today aren't one industry. Social web startups is a candidate though.
I think it's a mistake to categorize all web startups as belonging to the "web startup" industry.
Even though we (CarWoo! YCS09) are a web startup, we see ourselves primarily as a company that plays in the automotive industry. All of our key metrics are highly-correlated with the automotive industry, not with what Techcrunch writes on any given day.
While a lot of YC startups' metrics ebb and flow with the goings on of the echo chamber, many do not.
http://seekingalpha.com/article/43008-portfolio-allocation-n...
In YC's case, it's portfolio is entirely made up of high-risk startups, so to the best of my knowledge, it's not classic portfolio theory.
http://en.wikipedia.org/wiki/Modern_portfolio_theory