Y Combinator likes to act as if it has cracked the code of startup investing. Spray money at hundreds of fledgling companies, let most of them wither, and then point at the one that survives to justify the whole exercise. The thesis depends on the power law: a fat-tailed distribution where one or two outliers carry the entire portfolio. It sounds clever until you remember that fat tails are not a permanent feature of the universe. They thin over time. That is not opinion, it is statistics.

Reversion to the mean is not just some academic quirk. It is the gravitational pull of reality. Burton Malkiel documented this exhaustively in “A Random Walk Down Wall Street” even the most successful investment strategies see their alpha decay to zero as markets become efficient. What worked for Benjamin Graham in the 1930s stopped working by the 1960s. What worked for momentum traders in the 1990s evaporated by 2010. YC’s spray-and-pray approach is no different. It is just another arbitrage opportunity waiting to be competed away.

The irony is that Nassim Taleb, who popularized the black swan events YC depends upon, would likely be the first to spot the flaw in their thinking. In “Antifragile,” Taleb distinguishes between environments that generate true tail events and those that merely appear to. The early internet was what he would call an “Extremistan” environment with winner-take-all dynamics, network effects, and virgin territory. But even Extremistan environments eventually migrate toward “Mediocristan” as they mature. The variance compresses, the outliers become less extreme, and the distribution normalizes.

Consider the biological parallel through Stephen Jay Gould’s lens of punctuated equilibrium. The Cambrian explosion produced bizarre and wonderful forms like the five-eyed Opabinia, and the alien-like Hallucigenia, but most of these experimental designs failed. What survived were a few basic body plans that still dominate today. The internet had its Cambrian moment from roughly 1995 to 2015. Facebook, Google, and Amazon are the arthropods, chordates, and mollusks that survived. Everything since has been variations on established themes.

The trouble for YC is that the digital frontier that once spawned these freakish giants has been colonized. The internet was once a primordial swamp where anything could crawl out and grow legs. Today it looks more like a crowded savanna where the big predators have staked their territory and scraps are all that remain. Yes, a new species might appear now and then, but the odds of another hundred-billion-dollar behemoth emerging are shrinking by the day.

The statistical mechanics here are unforgiving. Benoit Mandelbrot showed in “The (Mis)behavior of Markets” how power laws in financial markets are temporal phenomena. They appear during phase transitions. YC is betting on continued phase transitions in an ecosystem that has largely found its equilibrium. It’s like expecting earthquakes in Kansas; the tectonic conditions simply aren’t there anymore.

The venture capital industry itself provides a perfect case study in mean reversion. David Swensen’s analysis in “Pioneering Portfolio Management” shows how venture returns have steadily declined since the 1980s. The top quartile venture funds from 1980-1990 averaged 40% annual returns. By 2010-2020, that figure had collapsed to under 15%. The entire asset class is reverting to mean, and YC, despite its brand, cannot escape this gravitational pull.

YC’s response has been to shovel more seeds into the soil, as if sheer volume can override ecological constraints. They’ve expanded batch sizes, loosened admission standards, and increasingly fund companies that would have been rejected five years ago. This isn’t growth; it’s desperation. It’s the behavior of a predator in a depleted hunting ground, forced to lower its standards or starve.

But in mature environments, more seeds do not mean more towering redwoods. Most simply die faster, crowded out, starved of light. Look at YC’s recent cohorts, another AI wrapper, another B2B SaaS tool, another marketplace with a slight twist. These aren’t the zero-to-one innovations Peter Thiel describes as truly valuable. They’re n to n+1 iterations in crowded spaces. The fat tail requires genuine discontinuity, not marginal improvement.

From an information-theoretic perspective, the situation is even clearer. Claude Shannon showed that channel capacity has fundamental limits. The internet as an information channel for business model innovation is approaching its Shannon limit. You can only extract so much economic value from connecting buyers and sellers, from advertising, from subscriptions, from data brokering. The low-entropy, high-value opportunities have been mined out. What remains is high-entropy noise where we have thousands of startups competing for slivers of remaining inefficiency.

Eugene Fama’s efficient market hypothesis, despite its critics, contains a kernel of truth that applies here: abnormal profits attract competition until they disappear. YC’s early success has spawned hundreds of accelerators, thousands of angel investors, and entire ecosystems dedicated to startup creation. The very success of the YC model has ensured its eventual failure. The observation changes the observed until the original thesis no longer holds.

The statistical truth is that as the tail of the distribution thins, the entire strategy collapses under its own weight. You can keep buying lottery tickets, but it does not matter if the lottery quietly replaced the jackpot with a consolation prize. The expected value of a lottery ticket remains negative regardless of how many you buy. YC is essentially arguing that if they buy enough negative expected value bets, the variance will save them. But as any student of the Kelly Criterion knows, this is a path to ruin. John Kelly’s formula, refined by Edward Thorp in “Beat the Dealer,” shows that even with positive expected value, over-betting leads to bankruptcy. YC is over-betting in a game where the expected value itself is declining.

So YC can keep telling itself it is playing on a power law curve. The reality is that nature, markets, and math all conspire to grind extreme outcomes back toward the mean. Their thesis was brilliant in a world of wide-open possibilities. In the long run, it is nothing more than a victim of statistical inevitability. The mean always wins, not because it’s stronger, but because it’s patient. Statistical gravity is weak but relentless. And YC, for all its posturing, cannot escape the pull.