SIGNAL DECOMPOSITION LAB
CIF v0.21.3
Separate relative asset selection from within-asset market timing, then test both on a pre-discovery temporal holdout.
Temporal holdout
2022-08 → 2024-08
2022-08 → 2024-08
NO WEIGHT SEARCH. RAS and MTS are transformations of the unchanged MOS v0.8. Their definitions and support gates are frozen before this temporal holdout is observed.
PRE-DISCOVERY TEMPORAL HOLDOUT
Independent time block
BNBSOLXRPADADOGELINKLTCTRXDOTBCH
Replay points1050
Mature 30d1050
RAS points1050
MTS points1050
Window: 2022-08-07 → 2024-08-06 · assets succeeded 10/10.
TWO DIFFERENT QUESTIONS
One MOS → two signal operators
RAS — Relative Asset Strengthcross-sectional operator
Same-date cross-sectional percentile rank of unchanged MOS v0.8.
Question: which asset is stronger right now versus its peers?MTS — Market Timing Scoretime-series operator
Current unchanged MOS v0.8 percentile versus the asset's strictly prior 26 weekly MOS observations.
Question: is this asset in a historically strong state versus itself?RAS RESULT
Relative asset selection
Date-matched median ρ0.018
Positive dates50.5%
Top2−Bottom2 spread0.28 pp
Cross-sectional dates105
MTS RESULT
Within-asset market timing
Median asset ρ → 30d raw-0.047
Positive assets40.0%
MTS high−low return-3.45 pp
Δ ρ vs raw MOS0.031
Baseline raw MOS median per-asset ρ: -0.078. MTS uses only strictly prior weekly MOS observations.
2D DECISION LAYER
RAS × MTS outcome matrix
RAS ↓ / MTS →
LOW
MID
HIGH
HIGH RAS
-3.37 ppvs BTCN=34 · beat BTC 41.2%raw 0.3% · DD -15.2%
-2.76 ppvs BTCN=69 · beat BTC 40.6%raw 1.5% · DD -9.6%
-6.14 ppvs BTCN=212 · beat BTC 38.2%raw -1.3% · DD -12.4%
MID RAS
-2.46 ppvs BTCN=134 · beat BTC 40.3%raw 2.4% · DD -9.6%
-4.87 ppvs BTCN=142 · beat BTC 33.8%raw -1.0% · DD -13.0%
-2.95 ppvs BTCN=144 · beat BTC 38.9%raw -3.1% · DD -13.2%
LOW RAS
-3.46 ppvs BTCN=175 · beat BTC 36.0%raw -1.8% · DD -11.2%
-7.42 ppvs BTCN=89 · beat BTC 29.2%raw -0.7% · DD -12.3%
-4.57 ppvs BTCN=51 · beat BTC 33.3%raw -2.4% · DD -18.3%
High/low thresholds are fixed at 67/33. The matrix is descriptive; it does not create a combined production score.
DECOMPOSITION VERDICT
NEITHER_SUPPORTED
NEITHER_SUPPORTED. This decomposition does not receive holdout support; redesign must return to hypothesis generation.
RAS gateNO2/4 checks
MTS gateNO1/4 checks
Formula promotedNO
Production readyNO
ANTI-LEAKAGE CONTRACT
What v0.12 is allowed to conclude
- Temporal holdout predates the 2024-2026 discovery/asset-holdout window.
- RAS and MTS definitions were frozen before temporal-holdout results were observed.
- MTS uses only strictly prior score observations; the current checkpoint is not part of its reference window.
- No MOS component weights or formulas are optimized in this lab.
- No production gate, AQS formula, or Risk formula changes in v0.12.
- A historical pre-discovery holdout is not a substitute for future live confirmation.