CONDITIONAL ALPHA ROBUSTNESS / SECOND CONFIRMATION LAB
CIF v0.21.3
Re-test the exact Volume-Price Thrust 7D × BTC Risk-On interaction on new assets and an outcome-purged pre-discovery block, then stress it across thirds, leave-one-asset-out and effect size.
Second confirmation
2019-07 → 2021-02
2019-07 → 2021-02
EXACT INTERACTION ONLY. No formula, EMA, regime, direction or threshold from v0.16 is changed. v0.17 adds robustness checks; it does not search for a better variant.
NEW ASSET + PRE-DISCOVERY OUTCOME HOLDOUT
Independent second confirmation
ZRXBATONTZILQTUMICXIOSTHOTLRCWAVES
Replay points730
Mature 30d730
Risk-on rows542
Risk-off rows188
Outcomes end 2021-05-29; v0.15 discovery screen starts 2021-06-01 · purged: YES.
FROZEN HYPOTHESIS
Volume-Price Thrust 7D × BTC Risk-On
Factorvolume_price_thrust_7d
Exact v0.15/v0.16 factor: trailing 7-day return multiplied by average quote volume over the latest 7 days divided by average quote volume over the preceding 30 days.
Regimerisk_on
risk_on when BTC close >= BTC EMA200; risk_off otherwise.
FIRST vs SECOND CONFIRMATION
Did the exact interaction replicate?
SampleTarget 7dTarget 30dTarget 90d+ target assetsOpp 30dΔ interaction
v0.162023-09 → 2025-080.0860.1220.084100.0%-0.2660.368
v0.172019-07 → 2021-020.1000.0930.051100.0%0.0490.083
PRESERVED v0.16 GATE
Base interaction replication
Target 30d ρ0.093
Positive target assets100.0%
Opposite 30d ρ0.049
Target−opposite Δ0.08366.7% positive Δ assets
minimum_eligible_assetsPASS
target_median_30d_rhoNO
target_positive_30d_assetsPASS
target_7d_positivePASS
target_90d_positivePASS
interaction_assetsPASS
interaction_deltaNO
interaction_positive_assetsPASS
STRONGER ROBUSTNESS FIREWALL
Thirds · leave-one-out · effect size
First / middle / last third ρ-0.268 / -0.078 / 0.053
Worst leave-one-out 30d ρ0.087
High−Low median 30d return9.49 pp
Assets with positive High−Low77.8%
first_third_positiveNO
middle_third_positiveNO
last_third_positivePASS
leave_one_asset_out_floorPASS
positive_effect_sizePASS
effect_size_asset_consistencyPASS
RISK / TAIL DIAGNOSTICS
High-thrust vs low-thrust states
Median return spread9.49 pp
25th-percentile return spread3.72 pp
Median drawdown delta2.05 pp
InterpretationDESCRIPTIVE
TURNOVER / PERSISTENCE
How fast does the signal state move?
Median lag-1 factor ρ0.058
High-state 1-week retention44.4%
Weekly state transitions65.5%
Production implicationTURNOVER LATER
SECOND CONFIRMATION VERDICT
SECOND_CONFIRMATION_FAILED
SECOND_CONFIRMATION_FAILED. The exact v0.16 interaction did not pass the preserved confirmation gate on the second independent sample.
Base confirmationNO
RobustnessNO
Live forward shadow eligibleNO
Contextual architectureNO — 2nd family required
SECOND-CONFIRMATION FIREWALL
What v0.17 is allowed to conclude
- The exact v0.16 Volume-Price Thrust 7D factor, BTC EMA200 regime rule, direction and target regime are unchanged.
- The v0.17 asset universe was not used in the v0.13-v0.16 factor/regime test universes.
- The v0.17 future-return labels end before the v0.15 discovery screen begins.
- The full v0.16 interaction gate is preserved; v0.17 adds stronger thirds, leave-one-asset-out and effect-size robustness checks.
- Risk-tail and turnover/persistence metrics are diagnostics only and cannot rescue a failed alpha confirmation.
- Passing v0.17 can only make this one interaction eligible for live forward shadow validation.
- The contextual architecture gate remains unmet until at least one second independently confirmed factor family exists.
- No factor formula, regime threshold, combined alpha score, MOS v2, production gate, AQS formula, Risk formula or MOS v0.8 formula changes in v0.17.