Good Heuristics, Bad Theorems

We posted a version of this observation on our Engineer Investor account earlier this year, and it kept coming back to us:

“In personal finance, the voices we respect as novices often lose their resonance as we gain expertise. Figures like Ramit Sethi, JL Collins, the RWM crew, and PFC act as ‘gateway drugs’, brilliant at communicating to a broad audience, but eventually appearing too simplistic or intransigent. There is a wide gulf between effective mass-market communication and the detailed, fact-based analysis you need once you’ve gone down the rabbit hole.”

Having now checked a pile of their specific claims against the statutes, the regulations and the original papers, we would revise it. The problem is rarely that these writers are too simple. It is that a heuristic built to change behaviour gets promoted into a theorem about finance, and the promotion happens in a single sentence, usually containing a word like “mathematically” or “unequivocal” or “forever.”

This page is the index for that work. It carries the scale we grade on, the scorecard, and the claims we set out to check that turned out to be fine.

The scale

Grading personalities produces nothing useful, because the same writer can be excellent on behaviour and careless on tax. So we grade claims, one at a time, on five levels.

The scorecard

Every row was checked against a primary source, and the confidence column says how exposed we think each verdict is. High means the quotation was read on the source page and the counterevidence comes from a primary source we read in full. Medium means the quotation is equally solid, but the evidence against it is contested in the literature or rests on a single publisher. Where a claim already has a Summitward guide arguing it properly, the row links there rather than restating it.

The shaded rows are the two we would ask you to act on. Both take a real historical average and treat it as a promise about a holding period. The two Medium rows are where we are most exposed, and we would rather point at them than have a reader find them.

The two hardest to fact-check, and why

Ramit Sethi is mostly solving a different problem. Automating savings, knowing your fixed costs, spending on what you value and negotiating your salary are not claims about finance that can be checked against a dataset. Where he does make a checkable one, the 7% return assumption, he is optimistic: he describes it as conservative because “we’re already accounting for inflation,” while the Dimson, Marsh and Staunton record puts US real equity returns at 6.6% a year since 1900 and the global figure lower still.¹ Worth noting he offers a second and incompatible rationale elsewhere, that conservative estimates give “a buffer for market fluctuations,” which is a description of a margin of safety rather than of an inflation adjustment. His spending percentages, by contrast, are explicitly labelled guidelines on the page itself, so treating them as a rule to be debunked would be attacking something he did not write.

Morgan Housel writes mental models rather than prescriptions, and the models hold up unusually well. His case that investors should aim to be reasonable rather than coldly rational, because a portfolio you abandon in a drawdown is worse than a theoretically inferior one you keep, is closer to the behavioural evidence than most portfolio advice.² His characteristic risk is narrative compression: isolating one underappreciated variable so it becomes memorable. The Buffett example is the clearest case, where time does explain the shape of the wealth curve and does not explain the return that got compounded. He concedes the point himself in the same passage, writing “Yes, he’s a good investor. But a lot of people are good investors,” so the honest criticism is about what a reader takes away rather than about what he claims.

Three criticisms that did not survive checking

We went looking for these and they were not there. Publishing the misses is the only thing that makes the hits worth anything.

• The Roth five-year rule. Mr. Money Mustache is widely said to have described the clock as running from the account’s opening date. His 2011 post actually describes converting once a year to build a “5-year pipeline” and waiting for “the first batch,” which is the correct per-conversion rule. The misstatement is in the comments on that post, from two readers in 2013 and 2014, and another reader corrected them in the thread.
• Personal Finance Club never discloses inflation. It does, including a whole explainer on nominal versus real returns and retirement targets built on 7% described as “a reasonable inflation-adjusted projection.” The defensible criticism is that the 10% and 7% illustrations are not consistently labelled, not that the caveat is absent.
• Ramit Sethi derives 7% as 10% minus 3% inflation. That arithmetic appears in write-ups about him and not on his own properties, which give two different justifications instead.

Two more cut the other way. JL Collins published in February 2026 that he had moved part of his IRAs into VT, the global fund, having argued for years that US multinationals were enough; the writer most often accused of intransigence updated in public. And in the same 2011 post that carries his multinational argument, he writes that actively managed funds “can and do outperform their indexes. This one has for me. But rarely for the long term.”⁵ That is a more careful position than the flat 2023 line in the scorecard, and it is why that row is graded Medium confidence rather than High.

What we recommend

Keep reading them. The direction is usually right, and for most people acting on a slightly wrong rule beats deliberating about the correct one. None of the thirteen rows above would have been improved by the reader doing nothing.

Then apply one test before you act on any round number, ours included: find the single input the conclusion is most sensitive to, change it to something defensible, and see whether the recommendation survives. In the mortgage claim that input is the debt rate, and the answer flips between 3% and 9%. In the Social Security claim it is the discount rate, and the answer flips at about 2.3%. In the savings-rate claim there is no such input, which is why it has lasted fifteen years and belongs in the top row.

Frequently asked questions

Are these writers worth reading?

Yes, and we say so in the row where each of them is right. The scorecard has three Robust rows and three Useful heuristic rows against two Wrong ones. A reader who followed all thirteen claims literally would still end up far ahead of one who followed none of them.

Why grade claims instead of people?

Because the same writer is often excellent in one domain and careless in another, and a personality grade hides that. Mr. Money Mustache is the clearest case: the savings-rate argument is the strongest row on the board and the Social Security valuation is one of the two weakest.

Do you have a conflict of interest here?

Yes. We sell a personal finance product and compete for the same readers. That is why every row cites a primary source you can check, and why the misses are published alongside the hits.

How do you decide what counts as Wrong?

A claim is Wrong when the mechanics or the arithmetic fails on its own terms, using the data the claim itself invokes. Both shaded rows meet that test: each takes a historical average and treats it as a guaranteed outcome over a holding period.

Key takeaways

• Grade claims, not people. Of thirteen checked claims, three are robust, three are useful but literally false, five are materially incomplete, and two fail on their own arithmetic.
• The failure mode is promotion, not simplification. A heuristic becomes a theorem in one sentence, usually the one containing “mathematically,” “unequivocal” or “forever.”
• Both Wrong verdicts share a structure. Each treats a long-run average as a promise over a specific holding period.
• Three criticisms we set out to confirm were false. Publishing those is what separates a fact-check from a campaign.
• Gateway advice is worth following. A reader who acted on all thirteen claims literally would still be far ahead of one who acted on none.

Related guides

• A 9% Mortgage Does Not Lose to 10% Stocks the first Wrong row, tested on 98 years of returns.
• What Discount Rate Belongs on Social Security? the second Wrong row, worked against SSA’s own factors.
• Investing at All-Time Highs a true statistic that loses its horizon on the way to becoming a slogan.
• Fact-Checking Mr. Money Mustache seven claims from one writer, including the ones that held.
• Personal Finance Gurus vs. Economists the academic version of this exercise, and where the models lose.

Author disclosure

Summitward sells a personal finance product and competes for the same readers as everyone graded here, which is a reason to discount our verdicts and to check the primary sources we link. We have published the criticisms that failed verification alongside the ones that held, and we have tried to make the case for each claim before the case against it. Nothing here is personalized financial advice.

Sources

1. Dimson, E., Marsh, P. and Staunton, M., UBS Global Investment Returns Yearbook 2026, published March 3, 2026. US equities returned 9.8% a year nominal and 6.6% real from 1900 to 2025. ubs.com
2. Housel, M., “Rational vs. Reasonable,” Collaborative Fund, December 11, 2018, and “Last Man Standing,” December 17, 2020, which is the source of the Buffett passage quoted here. collabfund.com
3. S&P Dow Jones Indices, SPIVA U.S. Scorecard, Mid-Year 2025, Report 1a. 93.81% of all active US domestic equity funds underperformed the S&P Composite 1500 over 20 years to June 30, 2025. spglobal.com
4. Collins, JL, “JL Goes International, and to ETFs… Oh my!”, February 8, 2026. The change is described as covering the Collins IRAs rather than the whole portfolio. jlcollinsnh.com
5. Collins, JL, “What We Own and Why We Own It,” June 14, 2011. Source of both the multinational-diversification argument and the concession that active funds “can and do outperform.” The same concession is re-quoted in his 2018 update, which fixes the wording as pre-2018. jlcollinsnh.com
6. Personal Finance Club uses the “forever” framing in at least nine posts between 2020 and 2026, including an identical paragraph in July 2026 and in February 2021. Representative: “Retirement Is a Number, Not an Age,” January 20, 2026. personalfinanceclub.com
7. Sethi, R., “Episode 58: If we want to retire, we’d have to live on $36k,” August 30, 2022: “I use 7% because I like to be conservative.” He repeats the figure in a March 2026 episode. Also the Conscious Spending Plan page, where the percentages are twice described as guidelines. Read September 14, 2026; the plan page was last modified September 9, 2026. iwillteachyoutoberich.com

Editor’s note

Educational content, not financial advice. Every quotation was read from the live source on September 14, 2026, and attributed to the specific post, page or comment it came from rather than to a writer in general; several of these sites republish and re-date material, and two of the pages cited carry no version date at all. Verdicts describe the claim as written on the date we read it. Where a writer has since updated a position, as JL Collins did in February 2026, we have said so.