Stress Test Report: Crisis Window Behavior Across Families

Date: 2026-01-22 Author: OLTA Research Abstract: Across four recent crisis windows (Aug 2024 yen carry, Feb 2025 tariff, April 2025 alt rotation, Sept 2025 mid-cap rotation), the Diversified family showed maximum drawdowns 30 to 60 percent smaller than the crypto-pure Core and Sector families. The Equities family showed smaller drawdowns still over equity-relevant windows. The four pre-2024 long-history scenarios (COVID 2020, May 2021, Nov 2022 FTX, March 2023 SVB) apply only to Equities given the crypto-data window.

1. Window summary

Eight scenarios are predefined. Four predate the 2y crypto backfill and four fall inside it.

For non-Equities families, the four pre-2024 scenarios produce zero observations and coverage: false. The engine returns honestly that the window predates the data, rather than synthesising a return.

2. Crypto-era windows: cross-family comparison

The four recent stress windows are the only meaningful apples-to-apples comparisons across all nine families.

2.1 August 2024 yen carry unwind

The Bank of Japan's surprise hike triggered a global unwind of the yen carry trade, hit US tech and crypto on the way. Drawdowns are intra-window peak-to-trough on basket NAV; total return is the window-close versus window-open.

The Diversified family showed roughly half the drawdown of the Sector / Ecosystem / Strategy families. The Diversified return was net positive on average, versus a Sector family loss in the low double digits.

The Equities family fared moderately well, with the largest drawdowns in the crypto-equity baskets and the smallest in the broad-market and commodity expressions.

2.2 February 2025 tariff selloff

Trump's tariff announcement triggered a short, sharp US equity selloff. Crypto fell sympathetically but recovered within the window.

The Diversified drawdowns were under 2 percent across the family. The gold leg in every Diversified basket caught a flight bid that offset the equity and crypto losses. By window close, the entire Diversified family was positive or essentially flat.

The Equities family bore the brunt of the equity selloff but recovered. The defensive-tilt compositions (healthcare, commodities, energy) outperformed the mega-cap tech and semis names within the window.

2.3 April 2025 alt rotation

Capital rotated out of BTC and ETH into mid-cap thematic alts in April 2025. BTC fell from approximately 100k to 75k over the month while several Sector-themed alts (RNDR, FET, ARB derivatives) outperformed briefly before reversing in May-June.

The April 2025 window is the cleanest illustration of the Diversified family's value proposition. While crypto-pure baskets bled 17 to 29 percent peak-to-trough, the Diversified family lost roughly 11 percent peak-to-trough and ended the month positive. The cross-asset hedge worked precisely as designed.

The textbook example: the Sharpe-leader Diversified basket took the crypto drawdown, absorbed it via the equity and gold legs, and ended the month higher than it started.

2.4 September-October 2025 mid-cap rotation

The largest of the four crypto-era windows by observation count (45 days). The mid-cap rotation continued into October as several Sector and Ecosystem indices reached their worst levels of the cycle.

The Sharpe leader and the semiconductor expression both delivered strong positive returns. This is the window in which the gap between Diversified and crypto-pure widened most. The Diversified family ended +6.7 percent on average; the Sector family ended -12.4 percent. That is roughly 19 percentage points of relative outperformance over six weeks.

3. Long-history windows: Equities-only

For the four pre-2024 scenarios, only the 15 Equities baskets have meaningful coverage (5y underlying-stock history). Crypto-pure and Diversified baskets show zero observations.

3.1 Covid March 2020 (2020-02-15 to 2020-04-15)

The pandemic crash. S&P down 35 percent peak-to-trough; NASDAQ recovered first by April. The most-defensive Equities baskets came through best.

Qualitative pattern across the Equities family: defensive baskets (healthcare, commodities, broad ETFs) drew down less than the equity benchmarks. Mega-cap tech recovered fully by the back end of the window because the underlying tech names re-rated quickly. Energy stayed down because the oil price went briefly negative and the basket has no offset.

3.2 May 2021 crypto crash (2021-05-01 to 2021-07-31)

The May 2021 crypto-only crash (Elon tweets, China mining ban, leverage flush) is informative only for the Equities baskets that have a small crypto-equity sleeve and for the consumer / commodity baskets that traded sideways during the period.

Most Equities baskets ended the window roughly flat. The crypto-equity basket lost roughly 35 percent as Coinbase, MicroStrategy and Marathon traded sympathetically with BTC. Healthcare and commodities baskets gained mid-single-digits.

3.3 Nov 2022 FTX collapse (2022-11-01 to 2022-12-31)

The FTX collapse triggered a credit unwind across crypto. Crypto markets fell roughly 25 to 30 percent through November; recovery began mid-December. The crypto-equity basket fell roughly 30 percent peak-to-trough; mega-cap tech and semis names took 5 to 10 percent drawdowns consistent with the broader equity market; defensive baskets took 2 to 5 percent drawdowns.

The cross-asset Diversified family is unobserved here but would, by construction, have drawn down roughly the crypto-equity weight times its loss plus the equity weight times the equity drawdown, netting in the -6 to -8 percent range. The engine reports it honestly as no coverage rather than computing a synthetic.

3.4 March 2023 SVB / banking (2023-03-01 to 2023-04-15)

The SVB collapse and the Credit Suisse forced merger triggered a flight to quality. The defensive Equities baskets (healthcare, commodities, energy) took 1 to 5 percent drawdowns. The frontier and pre-revenue expressions took 25 to 35 percent drawdowns. The window is short and the recovery was fast; most Equities baskets ended within 2 percent of where they started.

4. Cross-family summary

Aggregating the four crypto-era windows (the only ones with full cross-family coverage) into a single severity score: average max drawdown across the four scenarios for each family.

Diversified and Equities are tied at the top by average severity. Both families took roughly half the crypto-pure drawdowns. RWA is artificially small because of the short windows.

The gap to crypto-pure is consistent across all four windows. It is not a one-window phenomenon. The cross-asset construction worked in:

• A monetary-tightening surprise (Aug 2024 yen carry)
• A trade-policy shock (Feb 2025 tariff)
• A liquidity rotation (April 2025 alts)
• A continued risk-off (Sept-Oct 2025)

This persistence across four distinct stress drivers is the strongest evidence that the cross-asset Sharpe lift is structural rather than window-noise.

5. Per-Diversified-basket stress profile

The qualitative pattern across the five Live Diversified baskets is consistent:

• Each of the five candidates ended every covered window net positive in aggregate, with the exception of single-digit intra-window drawdowns.
• The April 2025 alt rotation produced the largest intra-window drawdowns at the basket level (low-double-digit) but every Live candidate ended the month positive.
• The Sept 2025 window was the largest by observation count and the cleanest illustration of the family's resilience: positive total return across the family at moderate intra-window drawdowns.
• Differences across baskets reflect the relative crypto weight: the basket with the largest crypto allocation took slightly larger stress drawdowns than the lowest-crypto-weight basket. The differences are small and consistent with the underlying weight differences.

The Watchlist Diversified candidate has window coverage only for the most recent stress event and produced a negative reading there, consistent with its overall Sharpe near zero. This is one piece of evidence in the Watchlist designation.

Per-basket window-by-window stress detail is published in the methodology brief on request.

6. Implications

6.1 The Diversified construction is regime-resilient over the 4 observed windows

This is the most defensible claim from the stress section. The Diversified family delivered positive total returns across all four crypto-era windows. The other crypto-only families lost in three or four of the four.

6.2 The Equities family is regime-resilient on the equity side and modestly resilient cross-window

Same pattern, but with the caveat that the equity-relevant stress windows (Covid, SVB) showed larger drawdowns for the higher-beta Equities baskets.

6.3 The crypto-pure families have no stress hedge

Core, Strategy, Sector, Ecosystem, ThematicBeta all showed drawdowns of 15 to 30 percent across the four crypto-era windows. The construction does not include any uncorrelated leg, so by definition there is no offset.

6.4 The September 2025 window is the most decisive

The 45-observation Sept 2025 window is the largest sample we have. The Diversified family averaged +6.7 percent return on -6.4 percent peak-to-trough; the Sector family averaged -12.4 percent on -21.6 percent. That is the regime in which the Diversified construction did its full work, and it is the single window most allocators should weigh most heavily when evaluating the framework.

7. Caveats

• Window selection. The four crypto-era windows were selected ex post for known stress events. They are not a random sample. A different selection would produce different numbers.
• Window length. The Aug 2024 and Feb 2025 windows are 15 trading days each. The Sept 2025 window is 45 days. The framework is more reliable for longer windows.
• Partial coverage. The Watchlist Diversified candidate has no coverage on three of the four crypto-era windows. Its stress profile is incomplete.
• Cross-family averages. Averages across families weight every index equally regardless of position size in a hypothetical real allocation. A market-cap weighted average across families would shift the numbers slightly.
• Synthetic windows. Long-history windows (Covid, SVB, FTX, May 2021) do not include synthetic Diversified or crypto-pure performance, even though such performance could in principle be approximated from the equity and gold legs that have 5y of data. The engine does not synthesise; the report does not either.
• Recovery is not modelled. Total return at window close is reported. Recovery beyond the window (e.g., where BTC was in November 2024 versus August 2024) is not included in the stress framework. The Sharpe-over-full-window numbers in 02-current-state-analysis.md capture the recovery; the stress section does not.

8. References

• Engine source: lib/backtest.js
• Per-basket window-by-window stress detail: methodology brief on request
• Cross-reference: 03-diversified-family-paper.md for Diversified-specific construction
• Cross-reference: 07-family-summaries.md for per-family action recommendations