Matt Maroon - But Seriously

The second curve is wrong -- because it was stretched graphically, it has more total area beneath the curve. That's important, because if you assume that the mean is the same, but the std. dev. is larger, the probability for the most expected outcomes are actually <i>lower</i> (i.e. the peak of the second bell curve should be below the first one).<p>This is more than just a pedantic observation -- it's an illustration that the VC way leads to a lower probability of success at the most common valuations, but a small chance at a much higher valuation.

Just imagine this comment thread for a moment as a series of videos you have to click on and listen to the author slowly waffle.<p>Scary isn't it.<p>Edit: Also, funny comment from their crunchbase profile: "The twitter killer as I have blogged few minutes ago. It has video, video is web2.0, it will kill twitter."<p>(Sorry for being slightly offtopic from the main points of matts post which seem to make a lot of sense)

<i>Founders, on the other hand, have the competing interest of reducing variation. They’d generally be happy to make a $30m mean return with zero variance. In fact, they’d often be happy with a much lower return and a much lower variance.</i><p>Ding Ding Ding. Somebody gets it.

Matt's mostly right, but the mathematical conclusions are, depending on how generous you want to be, minorly misleading or flat out wrong.<p>Leaving aside the incorrect graphs and potentially wrongly chosen distributions mentioned elsewhere, Matt missed the real reason VC's and entrepreneurs have different risk profiles.<p>Entrepreneurs have one company they're rooting for--their own. VC firms have tens or, over the course of their lifetime, perhaps hundreds. The law of large numbers says that as you sample more (with more companies), variance around the mean reduces, and you're quite likely to end up with your expected value at the mean. The entrepreneur has no such safety blanket and so would clearly choose to reduce variance.

All probability density functions must integrate to 1.

I swear I have the worst timing ever. I go to college just as the .com bubble crashes, and now that I'm planning my jump into startupdom everything crashes again.<p>Oh well, there's nowhere to go but up, right?<p>Right?

Maybe matt should look at <a href="http://en.wikipedia.org/wiki/Maxwell-Boltzmann_distribution" rel="nofollow">http://en.wikipedia.org/wiki/Maxwell-Boltzmann_distribution</a>.

From my limited understanding of how VC works, I would think that both distributions (curves) should be bimodal with one mode at zero rather than being normal distributions.<p>For the higher variance distribution, the mode at zero should have a higher probability than the mode at zero for the lower variance distribution. You might also expect that the second mode would occur at a higher valuation in the high variance distribution.

Poignant observations as usual, Matt. Thanks for the link as well.

Of course you can build an online business that depends on paying customers and be profitable from day 1 like TicketStumbler, but since when is it any easier?

&#62; when you count the big corporations that aren’t going anywhere (Google, Yahoo, MSN, Myspace, Facebook, YouTube, Amazon, eBay, etc.) and the porn sites (with which you cannot compete) you have a lot of dogs fighting for very few scraps.<p>WTF? Facebook and YouTube didn't even exist 5 years ago. Good to know they weren't thinking the same way.

A (truncated or folded) normal distribution is completely inappropriate here.

"You've got to pull the weeds if you want a beautiful garden" classic.<p>-P