Investor Profiles, Part Two: The Systematic Turn and Its Counterexample

Scope, and the rules this article is written under. Positions are documented and sourced to published work; no quotation is reproduced anywhere; no company, fund or holding is named, including the vehicles through which these approaches were applied; and nothing here suggests any reader adopt anyone's approach. Where a subject is living, their current position has been verified as at 14 August 2026 and re-confirmed at publication QA on 18 August 2026, rather than assumed — a requirement of this pillar, because a stale claim about a person is a worse failure than a stale claim about a rule.

Setting the three together is more useful than profiling any of them alone, because the disagreement between them is the substance, and because the way a reader resolves it says more about their standards of evidence than about the people involved.

Harry Markowitz: diversification stops being a proverb

Before 1952, diversification was folk wisdom — do not put everything in one place. Markowitz's Portfolio Selection turned it into a result.

The insight was that a portfolio's risk is not the average of the risks of its holdings. Because holdings do not move together, combining them produces a whole whose variability is lower than the weighted average of its parts — and the size of that reduction depends on the correlations between them, which is a quantity that can be measured rather than felt.

Two consequences followed, and both reshaped the field. Risk became a property of the portfolio rather than of individual holdings, so a volatile holding could reduce total risk if it moved differently from everything else — which is counterintuitive and correct. And the problem became one of optimisation, which is what opened the door to everything in Pillar 25's article on quantitative analysis, including its hazards.

The honest limitation is that the inputs are estimates. Expected returns, variances and correlations all have to be forecast, and the optimisation is sensitive to them — so a result that is mathematically exact rests on quantities that are not. That is not a flaw in the mathematics; it is where the mathematics hands the problem back.

John Bogle: the one variable that is not a forecast

Bogle's published argument is narrower than it is usually reported and stronger for it. He did not principally claim that markets are efficient or that managers lack skill. He claimed that costs are certain while returns are not, and that an investor controls the first and not the second.

The arithmetic behind that claim is set out in the article on index investing: on the portal's canonical parameters, the difference between a 0.10% and a 0.90% annual charge consumes 13.7% of terminal wealth over twenty years, because a fee is levied on the balance rather than on the return and compounds against the investor exactly as returns compound for them.

The rhetorical move that made the argument land was refusing to argue about skill at all. A debate about whether managers can outperform is unresolvable in the sample sizes available, as the article on factor investing demonstrates. A debate about whether a smaller number subtracted from an uncertain one leaves more than a larger number does is not a debate. Bogle relocated the question from a contested empirical claim to an uncontested arithmetic one, and that relocation is the intellectual contribution rather than any assertion about efficiency.

The documented counterexample

Warren Buffett has published annual letters to shareholders across six decades, which constitutes the largest body of first-person practitioner reasoning available anywhere. He stepped down as chief executive of the company he led with effect from 1 January 2026, remaining as chairman, with the operating role passing to a long-serving successor.

The documented positions are consistent over a very long period and are, in outline: that a share is a fractional ownership of a business rather than a tradeable object; that the durability of a company's competitive position matters more than its current statistics, which is the subject of the article on economic moats; that concentration in a few well-understood businesses is preferable to breadth across poorly understood ones; that the market's daily prices are an opportunity rather than an assessment, which is Graham's position carried forward; and that most activity is self-defeating after costs.

Two of those positions contradict the other subjects of this article. The concentration argument runs directly against the diversification result, and the case for selection runs against the case for not attempting it.

And the portal's own standards apply here as they applied in part one, without exception for prominence. A single long record cannot distinguish skill from fortune at the bar used elsewhere in this portal. The sample-size problem: as Pillar 25 shows, the standard error of a Sharpe ratio falls only with the square root of time, so even multi-decade records leave wide intervals. The selection problem, which is larger: this record is examined precisely because it is exceptional, and in any sufficiently large population of investors some records will be exceptional whatever the underlying distribution of skill. Choosing the most remarkable outcome and then asking whether it could have arisen by chance is the wrong question, because it was selected on the answer. The reply usually offered is that a coherent method stated in advance and applied consistently is different from a lucky sequence — and that reply is reasonable, unfalsifiable, and exactly what a lucky investor with a coherent method would also say. This portal does not resolve it. What it will not do is treat an outstanding record as evidence for the reasoning that accompanied it, having refused to do so for factors, for momentum, for contrarian reversal and for managers generally.

What the disagreement is actually about

The three positions are not really in conflict about markets. They are in conflict about who the advice is for.

Markowitz and Bogle both produce conclusions that hold for a general investor — the diversification result is a theorem, and the cost arithmetic applies to everyone. Neither depends on the reader being unusual.

The concentrated selection argument, by contrast, is conditional on the reader being able to do something difficult — understand a business well enough to hold a large position through periods when the price says otherwise. It is a claim about what is possible for someone with a particular capability, not a claim about what is advisable for a reader in general, and the practitioners who make it have generally said so.

Conflating those two kinds of claim is the most common error in the popular literature. A theorem and a description of what an exceptional practitioner did are different sorts of statement, and only one of them transfers to a reader automatically.