Key result: Best variant gain over momentum: V5 +0.031 (h63), BH q 0.56.
How it was tested: training data 2004-09 → (expanding, refit each January), test data 2014-01 → 2023-12 walk-forward; 2024-01 → 2026-06 hold-out; out-of-sample design (how it was checked on data it never saw while being built): not stated.
Horizon: 63 US sessions (126 secondary), monthly rebalance. Coverage: world cap by_exchange markets; panel 2004-09 on with >= 15 markets per date.
Stage: Validated — NOT FIT (shadow ranking for forward tracking only). Version v1; the hub record is as of 6 Nov 2026.
The verdict
NOT FIT: no pre-registered rule beats 12-1 momentum (all BH q >= 0.56) and momentum fails the display bar (WF t -0.10 / -0.46).
Development summary: Nothing beats 12-1 momentum of the composites, and momentum itself has no out-of-sample edge (WF IC -0.005 h63, -0.026 h126).
What the review allows
The walk-forward record as evidence that simple momentum / carry / macro ranks of the 22 composites carried no measurable 3-6 month relative-return information 2004-2023
The shadow ledger for forward tracking (a v2 needs a new SPEC)
What the review rules out
Display or act on the shadow ranking as a forecast
Claim a country-rotation edge from any variant
Flip the real-rate sign and re-test on the same data
What it predicts
Item
Detail
Question it answers
Rank the 22 markets by next-quarter / half-year USD return vs the cap-weighted world, to lean toward or away
12-1 momentum of the USD composite (B1), 2Y yield differential vs US (B2), real policy rate, curve slope 10Y-2Y, terms-of-trade 6m change
Output
Shadow rank 1-22 by 12-1 momentum with top-3 z components; NOT shown on any tab
Horizon
63 US sessions (126 secondary), monthly rebalance
Coverage
world cap by_exchange markets; panel 2004-09 on with >= 15 markets per date
Markets
US, India, Japan, Korea, Hong Kong, Taiwan, Canada, Singapore, UK, Germany, France, Switzerland, Netherlands, Italy, Spain, Sweden, Denmark, Norway, Poland, Finland, Australia, China
Model type
Model · Structural
Stage
Validated — NOT FIT (shadow ranking for forward tracking only)
Version
v1, first recorded 19 Oct 2026
Validation evidence
Training data: 2004-09 → (expanding, refit each January). Test data: 2014-01 → 2023-12 walk-forward; 2024-01 → 2026-06 hold-out. Out-of-sample design: not stated.
Headline results from the model's development record, as the hub publishes them.
Measure
Result
Compared with
Note
B1 12-1 momentum
IC -0.005
0
t -0.10, 120 dates
Best variant V5
IC +0.026
B1 -0.005
gain +0.031, BH q 0.56
Supplementary 2004-23, B1
IC +0.008
0
t +0.24, 232 dates incl. 2008
Limitations
price returns only; cap weights back-drifted (issuance ignored)
real-rate sign negative in both sub-periods, opposite to the pre-registered sign (not flipped)
Questions
Is the Country Relative-Return Ranker (22 markets) validated?
Its validation review dated 19 Oct 2026 rated it NOT FIT: no pre-registered rule beats 12-1 momentum (all BH q >= 0.56) and momentum fails the display bar (WF t -0.10 / -0.46).
What does the Country Relative-Return Ranker (22 markets) predict?
It answers: Rank the 22 markets by next-quarter / half-year USD return vs the cap-weighted world, to lean toward or away. Output: Shadow rank 1-22 by 12-1 momentum with top-3 z components; NOT shown on any tab. Horizon: 63 US sessions (126 secondary), monthly rebalance. Coverage: world cap by_exchange markets; panel 2004-09 on with >= 15 markets per date.
How was the Country Relative-Return Ranker (22 markets) tested?
Training data 2004-09 → (expanding, refit each January), test data 2014-01 → 2023-12 walk-forward; 2024-01 → 2026-06 hold-out, out-of-sample design: not stated. 12-1 momentum rank IC, WF 2014-2023: -0.005 ± 0.045 (h63), -0.026 (h126); Best variant gain over momentum: V5 +0.031 (h63), BH q 0.56; Hold-out 2024-26, momentum IC: +0.024 (h63, 30 dates).
Out-of-sample: results measured on data the model never saw while it was built or tuned.
Walk-forward: the model is refit on data up to a date and tested on the period after it, then the window rolls forward, so every test is out of sample.
Hold-out: a slice of data set aside and used once, for the final test.
Rank IC: information coefficient: the rank correlation between a model's scores and the returns that followed; 0 means no skill.
Shadow: scored and tracked on paper only; no list is published and nothing is acted on.
BH correction: the Benjamini–Hochberg adjustment for running many tests at once, so chance hits are not counted as findings.
12-1 momentum: the return over the past 12 months, leaving out the most recent month.