Nine strategy deaths run through one schema — the killers were the channels the map omitted, and the one survivor owned the channel
Built for Weather argued that every strategy is built for specific market weather — an operating envelope of statistical states — and that what dies is almost never the machine but the weather. This companion is the register: the famous deaths, run one by one through a fixed schema. What did the strategy harvest; what states did it require; how did the position travel to the boundary; who held the financing; what actually killed it; and which observable would have flagged it in advance.
The pattern across the cases is uncomfortable for the map itself. Portfolio insurance and the short-VIX products manufactured their own failure state. The Quant Quake killed funds through their neighbours' leverage. The franc killed carry through a policy gap no price process contained. LDI and convertible arbitrage died with their asset logic substantially intact, through the financing leg. Momentum's crash came from a concave pocket inside a convex strategy, All Weather's worst year was the one state its own construction could not diversify — and Graham's net-nets simply stopped occurring. The register closes with its great exception — funding that survives — and with what the exception actually proves.
Built for Weather set out the framework; this piece applies it, the way the tokenised-equity register applied the claim ladder: every case through the same questions, so the entries are comparable and the pattern is legible. The recap in three sentences. A strategy's envelope is the configuration of six observable properties it requires — continuation or reversal, priced versus delivered fear, relationship stability, dispersion, depth, and funding. Its failure geometry is how the position travels to the boundary: rules that cut die gracefully, rules that add die catastrophically. And envelopes are not exogenous weather — crowding narrows them, and scale can manufacture the failure state outright. Three layers are kept distinct throughout: state variables (the six properties), failure channels (endogeneity, crowding, policy discontinuity, funding withdrawal, correlation breakdown, capacity), and boundary geometry (cut versus add; graceful versus catastrophic).
Each register entry states: what was harvested · the envelope required · the position path at the boundary · the funding structure · what actually killed it · what flagged it ex ante. The narrative around each entry is short by design; the schema is the point. Structures are assessed, not the people who ran them — with one deliberate exception at the end.
The purest violations of the exogenous-weather assumption bracket the modern era: 1987 and 2018, the same mechanism thirty-one years apart, in institutional and retail packaging respectively.
Portfolio insurance sold a synthetic put: sell index futures mechanically as prices fall, and the portfolio behaves as if protected. The replication mathematics assumed the one thing the strategy itself would destroy at scale — price continuity. By October 1987 the insured pool was large enough that the model's selling became a dominant source of the market's falling — a replication rule that looked stabilising at desk scale converted, at market scale, into a destabilising feedback loop, each leg down triggering the next tranche of programmed sales (Brady Report, 1988). The crash had more than one cause; the feedback loop is the mechanism the record supports. The XIV note ran the identical loop in product form: an exchange-traded note short VIX futures, whose end-of-day rebalancing required buying those futures as the volatility shock intensified. On 5 February 2018 the rebalancing demand fed the spike it was short; the note lost roughly its entire value in a session and was terminated under its acceleration provisions (Credit Suisse, 2018). Neither strategy was surprised by the weather. Each, at scale, was the weather.
The lesson is the framework’s endogeneity principle, demonstrated rather than asserted: at sufficient scale, a strategy changes the state distribution it depends on. The observable is not a market property at all — it is the strategy's own footprint.
In August 2007, market-neutral equity funds — long cheap stocks, short expensive ones, hedged against everything visible — lost heavily over a few days in which the broad market did nothing unusual, then largely rebounded within the week. The forensic reconstruction (Khandani & Lo, 2011) points to a single multi-strategy holder deleveraging: its selling moved the factor portfolios everyone else also held, forcing the next holder to cut, whose selling moved them further. No statistical property had shifted. The failure travelled through occupancy.
The upgrade this forces: an envelope is not private property. Whoever holds the same positions holds part of your risk, and the compressed premium that made the strategy look safe was the evidence of company.
Carry against the Swiss franc floor was, in statistical terms, a beautiful trade: the central bank had promised a boundary, and selling francs against it harvested the spread with the volatility visibly suppressed. On 15 January 2015 the promise was withdrawn overnight and the franc moved by tens of percent in minutes — through every stop, gapping past the prices at which the models assumed exits existed. Retail FX brokers failed; carry books lost years of income before a trade could clear.
Read as a payoff, a peg is a short straddle written by a central bank, and the premium is collected by whoever trades as if the peg were weather. Policy-manufactured envelopes do not migrate. They end.
Two cases died while their asset logic remained substantially intact, which makes them the register's most instructive pair — and the reason Built for Weather promoted funding into the taxonomy itself (the Brunnermeier–Pedersen distinction, 2009).
Convertible arbitrage in 2008 held positions that were, on the models, cheap and getting cheaper — the hedge intact, the mispricing widening in the strategy's favour. It did not matter. The trade lived on borrowed balance sheet, and 2008 attacked the balance sheet from every side at once: prime brokers raised haircuts and cut leverage, Lehman's failure trapped rehypothecated client collateral in the administrator's estate, and the exits everyone needed narrowed together (Mitchell & Pulvino, 2012). The dislocation hurt the marks too; the funding withdrawal is what turned marks into forced sales of positions that were, on the models, right. UK liability-driven investment in 2022 was the same death in a pension wrapper: levered gilt positions held as a hedge — the liability-matching logic impeccable — and then the sequence ran in order: a fiscal announcement produced the gilt-yield shock; collateral calls forced gilt sales; the sales amplified the very move that had triggered them; the Bank of England intervened to break the spiral (Bank of England, 2022). The liability-matching objective stayed valid throughout — what failed was the implementation's liquidity architecture. In both cases the strategy's statistical envelope was open. The financing envelope was not.
The pair's shared epitaph: being right is a luxury of those who can hold the position to realisation. A strategy's true envelope is the intersection of its statistical requirements and the structural conditions that let it stay in the trade — and the funding dimension deteriorates fastest precisely when the others do.
Real estate is the limiting case and therefore not in the register: property almost never fails as an asset — its holders fail through the financing leg, which is why an asset class with no strategy attached still dies exactly one way.
Momentum carries a cutting rule and a trend pedigree, which tempts the convex label. The payoff is state-dependent: robust across its ordinary state space, it contains a locally concave region that becomes dominant after a regime transition. In the spring of 2009 it produced one of the worst strategy drawdowns on record: the long–short momentum portfolio lost roughly three-quarters of its value in three months (Daniel & Moskowitz, 2016). The mechanism was the short leg. After a panic, momentum is structurally short everything that has been crushed — which means short a call option on the rebound. When the junk rally came, the "safe" leg detonated.
The lesson generalises: convexity is not uniform across the state space. A strategy can buy insurance in most states and unknowingly sell it in one — and the one is reachable.
2022 belongs in the register twice over, because the casualty was not a strategy but the two reference portfolios. 60/40's bond leg left its envelope when inflation volatility returned — working exactly as mapped, just not as marketed. All Weather is the deeper case, because it was designed by people who understood everything above. Four macro states, equal risk to each, leverage equalising the bond sleeve: a serious answer to macro uncertainty. What the construction assumed without pricing was the stability of the correlations between its own sleeves — and that stability is itself a property of the low-inflation-volatility state. When the inflation shock pushed real rates up, equities, nominal bonds and linkers failed together: three of four boxes, with leverage on the sleeve whose envelope had closed.
The verdict is constructive, and it is this framework's inference rather than settled fact: keep the quadrant layer, and add a structural convex sleeve for the states in which the assumed cross-sleeve correlations fail. The register's point is narrower: diversification is itself a strategy with an envelope, and 2022 was outside it.
Not every envelope closes in a session. Graham's net-nets — stocks priced below net current asset value, bought as a basket — harvested multi-year reversal with a forcing mechanism attached: liquidation value put a floor under the error. The strategy assumed a supply of such stocks, and mid-century America supplied them in bulk: fragmented markets, thin coverage, no screens. Then the state stopped occurring. Databases and screens arbitraged the visible cheapness; analyst coverage filled the gaps; much of what remains below NCAV in a modern developed market is cheap for a reason — the convergence mechanism replaced by a value trap. The strategy survives only where its state still occurs: crisis troughs, and market segments the screening infrastructure never conquered, Japan's small caps being the standing example of where the characteristics have persisted.
This is envelope migration at multi-decade horizon, and it is the register's answer to every strategy marketed on a long backtest: the backtest measures a climate that may no longer exist. The disciples running net-net screens on saturated markets are not wrong about Graham. They are farming his crop in a climate zone that moved.
The register needs its exception, and Berkshire Hathaway is its cleanest public example at scale. The decomposition is fifteen years old and takes a sentence: Berkshire's returns load on cheap, high-quality, low-beta stocks, run at roughly 1.6 times leverage (Frazzini, Kabiller & Pedersen, 2018). Nothing in the factor recipe explains six decades of survival — books with the same factors and comparable leverage populate this register. What the decomposition cannot replicate is the funding structure. Insurance float is funding free of daily mark-to-market margin calls, obtained historically below the cost of Treasury bills when underwriting stays disciplined, and — the decisive property — still available in the failure state, as premiums keep arriving while market-based leverage is being withdrawn. Buffett's defining edge sits in the register's sixth column: he is the only entry whose funding envelope widens in a crisis. The edge was never the seed. It was owning land the bank couldn't call.
Two of his rules re-read cleanly in the framework's terms. Margin of safety is not a valuation doctrine but a failure-geometry instrument: it converts a catastrophic path into a graceful one by making "wrong" bleed instead of ruin. And permanent holding of wholly-owned businesses is the private-markets move done honestly — removing the mark-to-market observation of the envelope, disclosed as philosophy rather than sold as low volatility.
The exception also settles the verification problem the framework poses for every discretionary claim. A few concentrated decisions a year cannot be distinguished from luck inside a career — except at sixty years, which is the one track record long enough to verify. And the corollary arrived with the verification — call it the capacity paradox: a track record long enough to establish persistence is long enough for scale to consume the opportunity that generated it. Over the past two decades, at full size, Berkshire has broadly tracked the index.
Skill becomes verifiable, and purchasable, at the moment it stops existing at size.
Which is the register's closing entry and its verdict in one. The strategies died of the channels the map omitted; the exception survived by owning the channel. Almost every strategy dies. The ones that survive for decades are often, on inspection, funding structures wearing a strategy's clothes.
Written in a personal capacity. Analytical views only — not legal or investment advice.
Julian Gretzinger — Investor and writer on monetary history, real wealth mechanics, and financial markets. substack.com/@juliangretzinger