You run a $50M venture fund. You split it into equal checks — more checks means each one is smaller — then fire a vintage: every startup's return is drawn live from a heavy-tailed distribution. Most return nothing (they die dark); a rare one blooms gold at 20×, 50×, 100× or more.
Because returns follow a power law, the whole fund's fate is set by its single best deal, not its average one. The dashed line is the distribution's true mean — yet most vintages land below it, because that mean is propped up by the giant winners you usually miss.
A fund returner is one company whose payout equals the entire fund — it needs a multiple of at least your check count. So concentration (few checks) makes a single winner matter; spreading into 40 tiny checks needs a monster just to move the needle.
That is why index-fund diversification fails here: averaging tames a bell curve, but under a power law it just dilutes the one outcome that mattered. Fire many vintages and feel the variance — that variance is the asset class.
This lab hit a snag — the lesson continues without it.