Marcus Vale: Long week — but I kept getting pulled back to this power-law stuff, which is... not exactly light reading.
Ben Okonkwo: Same, honestly. There's a heaviness to it once you really sit with the numbers.
Marcus Vale: Here's the deal — Cambridge Associates looked at more than 4,000 venture investments. The top 100. That's 2.5% of deals. Over 70% of all returns. I want to just let that sit for a second.
Ben Okonkwo: It sits uncomfortably.
Marcus Vale: And it's not a bell curve gone slightly skewed — this is a power-law distribution, which is a categorically different thing. Most asset classes are built around normal distributions. VC is structurally the opposite of that.
Ben Okonkwo: Right — but I want to name the specific mechanism, because I think it gets hand-waved. Early-stage companies have binary outcomes. Most fail or plateau. A very small number grow exponentially. That bimodality is the root of the distribution, it's not a consequence of bad picking — it's the underlying shape of what early startups actually do.
Marcus Vale: Which is why Uber at 5,000x and Airbnb at 2,200x aren't flukes — they're illustrations of the math working exactly as designed. Either of those returns an entire seed fund multiple times over, on its own.
Ben Okonkwo: Fred Wilson's put it plainly — the power-law curve isn't just a feature of a good fund, it's the defining structure of how returns distribute within any fund. And once you accept that, the question that actually matters is... is this designed, or is this just what happens when you're partially blind and occasionally lucky?
Marcus Vale: That's the thing — 'partially blind and occasionally lucky' actually undersells the structure. Because the ceramics analogy lands it better than any fund deck I've seen: throw forty pots, most crack, one ends up in a gallery. Except in early-stage investing the gallery pot isn't worth twice the others. It's worth a thousand times.
Ben Okonkwo: And the gallery pot doesn't make the cracked ones better. That's the part people miss.
Marcus Vale: Right — but why? What actually makes that one pot a thousand times more valuable? Because it's not just randomness.
Ben Okonkwo: So this is the mechanism I want to name precisely. Winner-take-most markets. Network effects, switching costs, scale advantages — they cause a single dominant player to capture a disproportionate share of value. Mathematically. It's not a lucky accident that Uber ended up at 5,000x for its early investors; it's that the ride-sharing market, once it tipped, had almost nowhere else for value to accumulate. The structure... kind of demanded it.
Marcus Vale: So binary outcomes at the company level, winner-take-most at the market level — those two things compound.
Ben Okonkwo: Exactly that. And the Horsley Bridge data shows it empirically — 6% of deals generating 60% of returns. That's not picking the best pots. That's the math of what early markets actually do when they tip. The concentration is baked in before any VC makes a single decision.
Marcus Vale: Which means — wait, actually this is what shifts everything — the 11.4% from the bottom half of Cambridge's portfolio isn't evidence against the power law. That's the cracked pots still being clay. Nonzero, but not the story.
Ben Okonkwo: Nonzero but not the story — that's exactly the framing. The structure isn't a failure of portfolio management. It's accurate accounting for how early markets resolve. The question that doesn't follow neatly from this is whether any firm can reliably find the gallery pot beforehand, or whether they just own enough clay.
Marcus Vale: Owning enough clay — yeah, but that's where the math starts forcing behavior. Fund-level arithmetic is brutal. To return 3x a fund, you need a handful of holdings at enormous multiples, because the majority return basically zero. That's not a choice — it's a constraint. It means you can't optimize for minimizing loss. Loss-avoidance is the wrong game entirely.
Ben Okonkwo: Which is exactly what Chris Dixon named with the Babe Ruth framing — the strikeout rate isn't a flaw in the system, it's the cost of swinging hard enough to hit home runs. Andreessen Horowitz has built an entire firm posture around that.
Marcus Vale: And Sequoia, Accel — they've all internalized it explicitly. Many bets, most fail. That's strategy, not sloppiness.
Ben Okonkwo: Right — but I want to flag the assumption that breaks if you push it. Conventional diversification, the logic built for public equities, normally distributed assets — it's designed to minimize downside across the whole portfolio. That mental model, applied to VC, actually suppresses the behavior you need most.
Marcus Vale: Because it weights loss-avoidance over outlier capture. Those are opposite objectives.
Ben Okonkwo: Completely different game. And — I mean, this is maybe the sharper version of it — the firms that haven't internalized this are probably still framing wins and losses symmetrically. Which is just... wrong accounting for the asset class.
Marcus Vale: The part that gets uncomfortable later — and we'll get into it — is whether top-quartile persistence at Sequoia or Andreessen is actually winner-picking skill, or whether survivorship bias is doing most of the work.
Ben Okonkwo: That's the one I can't resolve cleanly from the data we have.
Marcus Vale: Nobody can — yet. But the winner-picking frame has already reshaped firm structure. If the power law demands outlier capture, you're not running a portfolio management firm anymore. You're running an access business.
Ben Okonkwo: But calling it an access business is actually where the survivorship bias problem bites hardest. Because — okay, think about what we're measuring when we say Sequoia has persistent top-quartile returns. We're measuring the firms that survived. The ones that didn't build the access, didn't get the brand, they're just... gone. We're not comparing Sequoia to its 1970s peers who failed. We're comparing it to the firms that also made it.
Marcus Vale: So the persistence data is polluted.
Ben Okonkwo: Potentially, yeah. And the Y Combinator number makes this really concrete — roughly half of demo day companies fail. One in five within twelve months. These are the most curated cohorts in venture. If YC's filter can't shift the base failure rate that dramatically, what exactly is selection skill measuring?
Marcus Vale: Huh. That's — I mean, that's actually a hard number to argue around. Half of YC demo day.
Ben Okonkwo: Which means distinguishing selection skill from favorable timing is genuinely difficult. Not impossible — but the signal is buried under so much structural noise. Market timing, technology risk, competitive dynamics — VCs can't reliably control any of those. They can add capital, networks, guidance. But the outcome distribution... it might be mostly set before the term sheet.
Marcus Vale: Okay, but — portfolio sizing is where this gets actionable, right? Because the response to that uncertainty isn't indexing. Under twenty bets, you're carrying something like a 30% chance of returning less than 1x. That's not theoretical.
Ben Okonkwo: No — and you can't just flip to hundreds of bets either. Access to the deals that actually matter is constrained. You can't diligence hundreds of companies properly. The power law can't be arbitraged away through volume.
Marcus Vale: So you're stuck in this band. Too few bets and the math kills you. Too many and you lose the signal entirely. And the whole time you don't actually know if you're skilled or just... well-positioned when the market tipped.
Ben Okonkwo: That might genuinely be unanswerable from outside the fund. Maybe from inside too. And I think that's the discomfort that doesn't resolve — winner-picking, as a discipline, might be structurally indistinguishable from structured luck. Even to the people doing it.
Marcus Vale: The thing that's sitting with me now — and I don't have a clean answer — is whether that distribution is permanent or whether it's an artifact. Like, what if the power law isn't a law of nature? What if it's what happens when you have almost no information about founders at the earliest stages, and AI-assisted diligence or better founder signals actually changed that?
Ben Okonkwo: That's... hm. That's the version of the question I genuinely can't answer. Because the sources that describe this as structural — Cambridge Associates, the Horsley Bridge data, all of it — none of them address whether better information at the front end would compress the distribution over time. They just assume the scarcity of signal is a fixed condition.
Marcus Vale: And if it's not fixed — if the power law is an artifact of information scarcity rather than something baked into how markets work — then the entire model that Andreessen Horowitz, Sequoia, all of them built... it's built on a condition that was always going to be temporary. That's a different kind of uncomfortable.
Ben Okonkwo: Genuinely unsettling framing. Because the fund structure, the carry model, the winner-picking posture — it all assumes that concentration is permanent. If the distribution spreads, even partially, the whole thing needs reinventing.
Marcus Vale: And we can't answer it. Nobody can right now. I think that's actually where I want to leave this — not resolved. Just... still turning.
Ben Okonkwo: Yeah. That feels honest. Good conversation, this one.
