Entirely correct, and the framing is good. The crudest binary process, summed, converges on the same smooth normal curve you get from processes that could land anywhere. That is the central limit theorem, and the observation that one shape keeps reappearing in finance regardless of what the underlying pieces look like is the right takeaway.

He calls it "a coin flip formalized over a century ago." Abraham de Moivre published the normal approximation to the binomial in 1733. Laplace generalized it in 1810. The Lindeberg and Lévy statement most people learn arrived in the 1920s. "Over a century" is true the way "over a dollar" is true of a hundred dollar bill. Say three centuries and the point lands harder, because the whole thesis is that the edge comes from very old free ideas.

The biography checks out. Diaconis, born 1945, left George Washington High School at fourteen to tour with Dai Vernon, took a BS in mathematics from City College of New York in 1971, a Harvard PhD in statistics in 1974, and went straight to Stanford. The underlying claim holds too. Asked to produce a random binary string, people alternate too often and truncate runs, a finding that has survived fifty years of replication.

The single most cited demonstration of "humans invent patterns in noise" is the 1985 Gilovich, Vallone and Tversky basketball paper that gave us the hot hand fallacy. In 2018 Joshua Miller and Adam Sanjurjo showed that paper contains a subtle selection bias. Counting substreaks inside a finite sequence is not the same as sampling from it, and the naive estimator is biased downward. Correct for it and the hot hand comes back at roughly 13 percentage points.

So the flagship study proving that smart people see patterns in randomness was itself a case of smart people misreading a random sequence. That is the best joke available in this entire subject and the thread walks past it.

Dave Bayer and Persi Diaconis, Trailing the Dovetail Shuffle to its Lair, Annals of Applied Probability, 1992. They model the riffle shuffle, derive the probability of any arrangement from the count of rising sequences, and measure distance from a truly random deck in total variation distance. The numbers for a 52 card deck are the reason "seven shuffles" entered the culture.

Nothing happens for four shuffles. Then the deck falls off a cliff between five and seven. That abruptness is the cutoff phenomenon, and it generalizes to three halves of log base two of n.

He writes that "the same formula quant desks use to check whether a price sequence has actually lost its structure" traces back to shuffling. Total variation distance and mixing time are the standard currency of Markov chain Monte Carlo convergence diagnostics, which is genuinely everywhere in quantitative finance, but "the formula quant desks use to test a price series" is not a thing that exists. Desks test serial structure with autocorrelation, variance ratios, Ljung and Box, Hurst exponents, entropy estimators. The lineage he is gesturing at is real and the specific claim is invented.

Correct and cleanly explained. The example is stars and bars, the standard bijection between distributing n identical balls into k labeled boxes and arranging n stars with k minus 1 dividers in a row. The count falls out as n plus k minus 1, choose k minus 1. The graphical device goes back at least to Ehrenfest and Kamerlingh Onnes in 1914, and the name was popularized by William Feller in An Introduction to Probability Theory and Its Applications, first edition 1950.

"A counting trick older than modern finance itself" undersells it again. Bijective counting is as old as combinatorics. This is the one layer where nothing needs fixing except the framing.

The characterization of the Medallion method is accurate and the returns are real. Roughly 66 percent annualized before fees from 1988 through 2018, with no negative year in that window. One hundred dollars in 1988 becomes about 398.7 million dollars in 2018 on the gross figures. Gregory Zuckerman's The Man Who Solved the Market is the standard account, and Bradford Cornell has a paper arguing the record "stretches explanation to the limit," which is a polite way of saying nobody has a satisfying story for it.

The thread's lesson is that a pattern repeating thousands of times independent of any story beats a pattern you went looking for. Fine. But Renaissance runs funds that outsiders can buy, built by the same people on the same philosophy. In 2020 Medallion returned 76 percent. Through December 25 of that same year the Renaissance Institutional Equities Fund was down 22.62 percent and Renaissance Institutional Diversified Alpha was down 33.58 percent.

Same firm. Same data first religion. Roughly a hundred point spread. Whatever Medallion has, it is not the philosophy, because the philosophy is available to the public funds and the public funds lost money. The thread cites Renaissance as proof that method beats narrative, and Renaissance is the cleanest available demonstration that method alone is not the thing.