Actionable insights — choosing a sleeve, and expressing it

30:18 1. Choose the sleeve by its cost of carry, not by its expected return

1. Start from the exposure you want, not the instrument — here, "I wanted exposure to commodities," because stocks and bonds are both short inflation.
2. Test the obvious vehicle first and look at the realised numbers rather than the theory. He ran the commodity indices and "the returns were not that great."
3. Decompose the disappointment before rejecting the exposure. The diagnostic question is whether the asset is bad or whether the holding cost is eating it: "it's not because commodities are terrible, it's because commodities have negative carry because you have to pay for storage."
4. Name the mechanism in the market's own terms so you can check it: in futures, that storage cost shows up as contango — each expiring contract rolled into a more expensive one, a bleed that repeats every roll regardless of the spot price.
5. Look for a proxy with the same exposure and a negligible carry. Gold qualifies on both legs: "with gold it's very minimal," and, empirically, "gold mimics the commodity indices over time."
6. Apply the test to every sleeve. Cash yields little but costs nothing to hold; a house has real carry (taxes, maintenance, a mortgage); an 80-vol asset carries an attention cost that never appears on a statement.

• Any commodity fund whose chart badly lags spot over years — that gap is the roll cost, not bad luck; a futures curve in steep contango before you buy the index; a "diversifier" whose annual carry exceeds the diversification it buys.

32:00 2. Express the real-estate sleeve with what you already own — then admit its flaw

1. Count the equity in your home, not its price, as the real-estate allocation: "if you own a home, the equity in that house could be considered your real estate allocation."
2. Write down the defect rather than quietly accepting it: it is one property in one local market — "it's not ideal because it's one house in this idiosyncratic geographic area." An asset can fill a sleeve and still be undiversified within it.
3. If you do not own a home, use the listed proxy: "you can simply buy a REIT ETF" — REITs hold "apartments, offices, malls… data centers and cell phone towers," which "by and large is a pretty good proxy for all real estate."
4. Check that the sleeve has a long enough data history to be worth backtesting at all: "the real estate indices have been around since 1972."
5. Confirm the addition on both axes before committing to it — return and risk-adjusted return: "when you add real estate… the returns go up and the Sharpe goes up by quite a bit," because "it's not super correlated with anything else."
6. Do not expect it to carry the growth: "stocks over the last 100 years have returned about 10%, real estate has returned about 5%." It is there for the correlation and the inflation gearing.

• Home equity larger than every financial sleeve combined (very common, and it means the portfolio is already concentrated); a REIT fund whose composition has drifted into data centres and towers, i.e. into the same theme as your equity sleeve.

19:46 3. Size to your own risk of ruin, not to a return target

1. Ask the question in the ruin direction rather than the growth direction: not "what could this become," but "what would a 50% fall in this do to my life."
2. Use the wealth-level test. Dillian's generalisation is blunt: "wealthy people think about the risk of ruin and middle class people don't." Decide, honestly, which of the two behaviours your current allocation reflects.
3. Run the lottery thought-experiment on your own balance sheet: if the amount were already enough, the highest-expected-return allocation stops being the right one — "would you take $300 million and put it all in SPY?… nobody does that."
4. Separate need from want: the 401(k) holder "wants to double and triple"; the winner wants to still be solvent in thirty years. Only the second one has a ruin constraint, and only the second one should dominate sizing.
5. Re-ask it periodically, because memory decays: "people were thinking about it 2010, 2011, 2012 when the financial crisis was still pretty fresh… coming up on 20 years later people have totally forgotten."
6. Note the limit of the outsourced fix. Vanguard's advisor alpha finding — a third-party referee saying "stop trading" was worth about 3% a year (11:56) — improves conduct but not experience: "even if you have an advisor, if you take a 50% drawdown, you're still going to be stressed."

• A plan whose only stated number is a return; a portfolio you would have to keep working ten extra years to recover from; an advisor relationship being used as a substitute for an allocation you can actually hold.

43:34 4. Measure drawdown peak-to-trough, not by calendar year

1. Treat annual return tables as a presentation artefact. A crash that straddles a December boundary is split across two rows and each row looks survivable.
2. Re-measure the same episode from its high to its low, whatever the dates: "this gets kind of lost in the annual numbers, you don't really see it, but the total drawdown from the summer of 2007 to March of 2009 was 57%."
3. Compare the two numbers deliberately. The calendar year 2008 shows −38%; the actual experience was 57%, and 57% is the number that decides whether someone sells.
4. Extend the window as far back as the data allows before deciding what is possible: "the S&P 500, it's 89% if you go back to 1929."
5. Apply the same measure to the portfolio you are proposing, so the comparison is like-for-like: max drawdown 12%, second worst 9%, "the third, fourth and fifth worst, 1%."
6. Do the same check on a log-scaled chart, which compresses exactly this — his standing complaint on the 2026-SEP-08 page.

• Any performance table presented only in calendar years; a "worst year" figure quoted without a peak-to-trough figure beside it; a fund's marketing drawdown measured on month-end closes rather than intraday extremes.

22:15 5. Choose the wrapper for its pricing frequency — once-a-day NAV as a defence

1. Recognise that two products holding the identical portfolio can produce different outcomes, because one shows you a price continuously and the other does not.
2. Prefer the lower-frequency price for long-horizon money: an open-end mutual fund prices once, after the close, at NAV; "you only get one price per day… and that's it."
3. State the reason as a causal chain rather than a preference: "the more information you're getting on price, the more it affects your decision making and it causes you to do stupid things."
4. Weigh it against the real advantages of the ETF — intraday liquidity and, particularly, better behaviour in illiquid underlyings ("especially when you're talking about illiquid stuff like high yield bonds"). The choice is horizon-dependent, not absolute.
5. Where a low-frequency wrapper does not exist for a sleeve, build the friction yourself: check on a schedule rather than on impulse. The daily-vs-annual checking statistic on the 2026-SEP-08 page is the same idea from the other end.
6. Apply the same test to any asset before adding it: how often would owning this make me look? An asset that fails this test is excluded even if it improves the risk-adjusted return — see insight 6.

• Choosing an ETF over its mutual-fund twin purely out of habit; a brokerage app whose home screen is a live P&L; adding an asset whose price you would check more often than the rest of the portfolio combined.

41:41 6. Let the attention test veto the optimizer

1. Run the optimisation honestly first, and report the result even when it goes against you: adding Bitcoin "increased the Sharpe of the portfolio" in 2019, and he says so.
2. Then apply a second screen the optimiser cannot see: how much of your attention will this position consume? "Guess which one you're going to be staring at every day?"
3. Note that the screen is size-independent. "Even if it was only 2% of the portfolio, you're just going to be staring at it all the time because it's so volatile" — so trimming the position does not fix it.
4. Use a volatility label as the trigger. An "80-vol" asset — one whose annualised swing is around 80% — fails the screen regardless of its return history.
5. Decide in advance which objective wins when the two screens disagree. For him the objective is stated plainly: "the whole purpose of this is to minimize stress."
6. Where exclusion is impractical (you already own the asset), reclassify rather than churn: count it half toward gold and half toward stocks.

• A backtest improved by exactly the asset you already can't stop watching; a small "satellite" position taking more of your attention than the core; any sizing argument that assumes attention scales with position size.

24:39 7. Audit "diversification" for reflexivity before you rely on it

1. Check the top-weight concentration of any index you hold before calling it diversified: "the top seven stocks make up 35% of the index."
2. Ask what the flows into it mechanically do: "if you want to buy the index, you have to proportionally buy the top seven stocks and they get bigger." Buying the measure moves the measure — the property Dillian names as reflexive.
3. Use the active-manager population as a read-through: "most large cap managers underperform. The only ones that are outperforming are the ones who disproportionately own the top seven stocks" — a market where outperformance requires more concentration, not better selection.
4. Track the share of assets that is indexed as a regime variable: 2% of AUM in the late 1990s, 56–60% now, "150 million people who are doing the same thing."
5. Stress-test the exit rather than the holding. The failure he points at is a liquidity event, not a valuation one: "what you saw during the pandemic when the market was down 35% was a mass liquidation of index funds, and it happened very quickly."
6. Keep the epistemic humility he keeps: "Michael Green obviously has done a lot more work on the flaws of indexing than I have." Read the primary source before acting on the conclusion.

• Rising top-10 index weight; a portfolio of several "different" funds whose top holdings are the same seven names; falling cross-stock correlation alongside large single-name moves, which is when the index masks what is happening underneath it.