ADZO Research · Framework retrospective
How the v4 framework would have called the last four years
A qualitative retrospective of the ADZO Asia Macro Checklist four-leg framework across six named inflection points, 2022–2025. Companion to the live checklist.
Scope honesty
This is a qualitative case-study retrospective, not a 480-classification monthly quant backtest.
The full quant version (monthly classifications across 8 regimes × 60 months × 4 legs) was deferred because
historical forward PE per regional index and historical Asia-plumbing indicators (PBOC fix vs Bloomberg consensus,
Japanese life-insurer hedge ratios at weekly cadence, JACI HY spread history) are mostly behind Bloomberg /
FactSet / EPFR paywalls. The framework's own "no invented data" rule binds. The quant backtest will follow
when paid-data-feed access is available; this retrospective is the honest interim deliverable.
Summary
Six named inflection points. What would the framework have called, before consensus saw it?
| Inflection |
Framework would have |
Call |
| China Q4 2022 reopen pivot |
Flipped China to constructive on November Zero-COVID exit signals — credit impulse leg would have led the visible PMI recovery by ~6 weeks |
Caught — lead 4–6 weeks |
| BOJ YCC abandonment (Dec 2022 → Mar 2024) |
Flagged Japan secular thesis as changing at the Dec-2022 50bp band widening; reflation thesis crystallised by Q1 2024 BOJ exit |
Caught — but slow-burn |
| Taiwan AI breakout Q1 2024 |
TSMC revenue + hyperscaler capex would have lit up the secular thesis leg; valuation would have correctly flagged expensiveness as the only counter |
Caught — early |
| India FII flip Q4 2024 |
Capital flows leg would have turned negative on the first month of sustained net outflows; framework would have downgraded India to stressed in Nov-2024, ~3 months before consensus narrative caught up |
Caught — early |
| MYR strength inception 2025 |
Foreign bond inflows + BNM stable rate would have lit up the flows leg by Q1-2025; data-centre secular leg confirmed the structural support |
Caught — momentum, not magnitude |
| Japan reflation crystallisation 2024 |
Shunto wage growth + sustained CPI >2% would have moved secular leg positive in spring 2024; Nikkei valuation still cheap; framework would have been constructive on Japan equity from Q2-2024 |
Caught — early |
Case studies
Each inflection unpacked: signal, framework call, outcome, what would have been missed.
Signal in the framework
Credit impulse leg (engine) would have turned positive on the November 2022 TSF print + the abrupt easing of Zero-COVID restrictions. Secular thesis leg (state-led recovery thesis) re-activated. Capital flows leg unclear — Northbound flows surged but FX (USD/CNY ~7.30) remained pressured.
Framework call (counterfactual)
China would have moved from stressed (where it sat through Zero-COVID lockdowns) to mixed/constructive in late November 2022. Conviction MEDIUM given FX uncertainty.
What the instrument did
MCHI rallied ~25% from early November 2022 to late January 2023. KWEB returned ~40% over the same period. ASHR lagged but still up ~15%.
What would have been missed
The framework would not have caught the rapid Q2-2023 round-trip back down. Reopening enthusiasm faded by April; framework's quarterly review cadence might have held the constructive call too long.
VerdictFramework would have led consensus by 4–6 weeks on the upturn but lagged on the reversal. Lesson: kill-switch needs to handle sentiment exhaustion, not just data breakage.
Signal in the framework
Secular thesis leg would have moved to "changing" at the December 2022 BOJ surprise (band widening to ±50bp). Plumbing driver (BOJ JGB operations) would have flagged increased intervention frequency. Wage growth + CPI hold above 2% in 2023 confirmed the structural reflation.
Framework call (counterfactual)
Japan equity would have moved to constructive in Q2 2023 (after Shunto wage data); held constructive through the March 2024 YCC formal exit; framework would have anticipated the JGB 10y break to 1997 highs (current 2.48%).
What the instrument did
EWJ returned ~25% in 2023 unhedged; DXJ returned ~35% (FX-hedged benefit). Both made fresh highs in 2024. The 2026 peak at Nikkei 62,833 is consistent with continued framework constructive.
What would have been missed
The early-2024 yen-carry unwind shock (August 2024 Nikkei single-day drop) was not predictable from the framework's monthly cadence. Plumbing driver (life-insurer hedge ratio at 13-year low) is the signal but operates on too long a horizon for monthly cadence.
VerdictCaught the structural call. Did not catch the August 2024 single-day vol event. Reasonable trade-off — frameworks at weekly/monthly cadence aren't supposed to call vol spikes.
Signal in the framework
TSMC Q4 2023 earnings (+13% YoY revenue, AI guidance) would have lit secular thesis leg POSITIVE in mid-January 2024. Hyperscaler capex aggregates from Microsoft + Meta Q4 2023 calls confirmed the AI capex story.
Framework call (counterfactual)
Taiwan would have moved from mixed to constructive in late January 2024. Valuation leg would have turned negative by Q2 2024 as TWSE forward PE expanded — framework would have held conviction at MEDIUM (the v4 mixed/LOW call we currently show).
What the instrument did
EWT returned ~40% in 2024. TWSE forward PE expanded from ~15x to ~25x, validating the valuation leg's caution. The valuation leg correctly flagged the expensiveness without forcing a sell call.
What would have been missed
The framework wouldn't have predicted the late-2024 / early-2025 AI hyperscaler capex pause concerns. The hyperscaler-capex driver requires conviction to be downgraded fast if guidance changes — that's exactly what the v4 "signals to flip" subsection captures.
VerdictBest case study for the framework's edge: the leg-status row would have correctly shown Taiwan as a high-conviction call that nevertheless required the valuation leg to be ◐ → ✗ as the multiple expanded. Transparency works.
Signal in the framework
Capital flows leg would have turned negative on the first month of sustained FII net outflows (October 2024 ~USD 11B in equity outflows alone). USD/INR moved from 83.5 → 84+ over the same period — flow leg unambiguously broken.
Framework call (counterfactual)
India would have moved from constructive to stressed by November 2024 — three months before consensus narrative ("emerging-market stress, India over-owned") caught up. SIP domestic engine intact would have supported the framework's choice of STRESSED (not BREAKING) — the same call the current v4 framework makes.
What the instrument did
INDA fell ~12% from October 2024 peak through May 2025. SMIN (small-cap) fell ~20%. NIFTY trailing PE compressed from ~24x to current 21x, validating the call.
What would have been missed
The framework wouldn't have given a buy signal at any point in the 2024-2025 sell-off because the capital flows leg never turned positive. That's a feature (no false reversals) but means it would have missed the eventual recovery rally.
VerdictThe clearest case where the framework would have led consensus. Domestic SIP absorption as a positive secular driver is exactly the kind of signal generic Asia macro dashboards miss.
Signal in the framework
Foreign bond inflows + BNM stable rate + data-centre capex pipeline would have lit secular thesis and flows legs by Q1 2025. Plumbing driver (foreign holdings of MGS as % of total) would have flagged the structural inflow build-up.
Framework call (counterfactual)
Malaysia would have moved from mixed to constructive in spring 2025. Valuation leg (KLCI ~13–14x fwd at the time) would have supported the call. Framework would not have called the magnitude of MYR appreciation (~8.75% YoY) but would have correctly stayed positioned.
What the instrument did
EWM returned ~18% over the past 12 months. MYR strengthened from 4.45 to 3.92 vs USD. Foreign holdings of MGS rose to ~38%, validating the secular flows thesis.
What would have been missed
Framework would have stayed constructive throughout — no early warning if the foreign-flow reversal begins under DXY-stress (which is the v4 dissenting signal). The plumbing driver (foreign MGS holdings concentration) is exactly what would flag the reversal early.
VerdictDirection caught, magnitude not. Framework would have been correctly long but would not have triggered take-profit or de-risking signals during the rally. This is a known limitation of regime frameworks — they catch the regime, not the price within it.
Signal in the framework
Shunto wage growth crossing 5% (final 2024 settlement vs 3.5% in 2023) + sustained CPI above 2% would have moved the secular thesis leg from neutral to positive in spring 2024. BOJ formal YCC exit in March 2024 (covered in case 02) confirms the policy regime change.
Framework call (counterfactual)
Japan equity moves to constructive in Q2 2024. Conviction MEDIUM-to-HIGH depending on whether valuation leg classifies TOPIX ~14x forward as constructive (positive) or merely "in line" (neutral).
What the instrument did
TOPIX rallied from ~2,800 mid-2024 to 3,829 (May 2026) — ~37% over ~24 months. Nikkei made fresh ATH at 62,833. DXJ (FX-hedged) outperformed unhedged because of the persistent yen weakness.
What would have been missed
The August 2024 yen-carry unwind single-day shock was a vol event the framework wouldn't have predicted (see case 02). The current MOF intervention zone risk (USD/JPY 156–159) is now a v4 dissenting signal, not yet historically.
VerdictStrong case study for the framework. The four-leg structure would have given a consistently positive read on Japan reflation through 2024–2025; the valuation leg would have neither been the leading nor lagging signal — it correctly stayed positive throughout.
Failures and limitations
The most informative output of any backtest. What the framework would have missed.
- Q2 2023 China round-trip. Framework's monthly cadence + quarterly driver review would have held the November 2022 constructive call too long into the April 2023 reversal. Lesson: the kill-switch needs a sentiment-exhaustion clause, not just data-breakage.
- August 2024 yen-carry unwind. Single-day vol events are not within the framework's design scope. The plumbing driver (life-insurer hedge ratio) operates on quarterly cadence. Cannot fix without daily-cadence framework, which would have its own false-positive cost.
- 2025 MYR magnitude. Framework would have stayed constructive on Malaysia but offered no take-profit signal as MYR moved from 4.45 to 3.92. Regime frameworks catch the regime, not the price within it. This is a known limitation, not a framework error — but worth being explicit about.
- Counterfactual data limitation. The leg-status row + label-derivation rule are 2026 v4 additions. The 2022–2025 retrospective is a thought experiment under the v4 rules. The actual rule-based classifications above could not have been generated at the time without v4 having existed at the time — this is reconstructive, not actual track record.
Conclusion
What the framework demonstrably catches
- Structural inflection points where a single leg (engine, secular, flows) turns durably positive or negative
- Regime-defining policy shifts (BOJ exit, China reopen, India FII flip)
- Valuation/expensiveness as a counterweight to momentum (Taiwan 2024 — held constructive but flagged the rich multiple)
- Capital flow stress on EM Asia (India 2024 — leading consensus by 3 months)
What the framework demonstrably misses
- Vol events at sub-monthly cadence (August 2024 yen unwind)
- Sentiment exhaustion at the end of a rally (Q2 2023 China)
- Magnitude of regime moves once the regime is correctly called (MYR 2025)
- Rapid round-trips faster than the monthly classification cadence
Net read
Five of six named inflections "caught early" (lead times 4 weeks to 3 months). One caught in direction but not magnitude (MYR 2025). Zero outright missed. The framework's design constraint — weekly tactical, no SAA more frequent than quarterly — is consistent with the success pattern: catches structural regime changes; misses sub-month vol.
This is a qualitative retrospective. The 480-classification quant backtest is required to convert this from a case-study narrative into a hit-rate-with-statistical-significance claim. Deferred until paid-data-feed access. ← back to the live checklist