-- COUNTRIES · CALIBRATED PROBABILITY · 24-MONTH
HORIZON
FILTER:
Risk Legend
0%25%50%75%100%
Critical — p ≥ 0.70
High — p 0.50–0.70
Watch — p 0.30–0.50
Stable — p < 0.30
Top Risks
Sovereign Risk Rankings
#CountryProb.ZF ScoreBIS GapAccel.TierConf.YoY
SEARCH A COUNTRY TO VIEW PROFILE
#CountryGradeConf. ScoreWB / BIS CoverageData LagWBBISUpdatedSource
Physical Stress Layer — ZESI
EXPERIMENTAL · TIER-1
ZESI CORE · Z_NDVI + Z_VIIRS · 1,344 OBS ·
2017–2023 · NOT YET IN PRODUCTION MODEL ·
FRED: LIVE PIPELINE (L1 REGULARISED TO ZERO)
Year:
ZESI Core
Severe StressNeutralStrong
#CountryZESI CoreZ_NDVIZ_VIIRS
Top Stress Movers (YoY)
Components —
click country
ZESI vs Probability — Blind Spots
Amber = high stress, low model probability. These may be
surfaced before macro data catches up.
Commodities
Stress Transmission Network · 9 Signals ·
Individual Country Exposure
Tracked
9
Stress Alerts
--
Live Data
6
Pending
7
FILTER BY CATEGORYFILTER BY STRESSUpdated --:--:--
Commodity
Price ↑↓
→
Household
Cost ↑
→
Fiscal
Subsidy ↑
→
F7 Social
Unrest ↑
→
Sovereign
Stress ↑
→
ZEVKORA
Risk ↑
Digital Economy Intelligence
Layer 12 · Touch globe to explore · Tap country for detail
SCORED
--
SHUTDOWNS
--
UNSCORED
--
VINTAGE
2024
DRAG TO ROTATE · TAP COUNTRY FOR DETAIL
Digital Fragility
Critical 75+
High 55-74
Moderate 35-54
Low 15-34
Minimal below 15
Digital Fragility IndexTOP 10
Mobile Money Leaders
Shutdown Monitor
Layer Status
Z_INTERNET_PENETRATION ✓ LIVE
Z_MOBILE_SUBSCRIPTIONS ✓ LIVE
Z_MOBILE_MONEY ✓ LIVE
Z_INTERNET_SHUTDOWN ✓ LIVE
Z_DIGITAL_FRAGILITY ✓ LIVE
Production model PENDING
--
--
Digital Fragility Score
--
--
--
Digital Signals
Internet Penetration--%
Mobile Subscriptions--
Mobile Money Accounts--%
Internet Shutdown--
Intelligence Reading
--
Sovereign Risk Link
Loading...View Full Profile ↗
⚠ Digital Fragility is supplementary -- not in the production sovereign probability modelSources: World Bank Findex 2024 · World Bank API · Access Now · NetBlocks
ZEVKORA assigns a
calibrated probability score
to every sovereign — the estimated likelihood of a crisis event
within a
24-month horizon.
A crisis event is any of: currency collapse, sovereign debt
default, IMF emergency programme, or acute political rupture.
The model detects
structural conditions
that historically precede crises — not when a crisis will occur,
but whether the conditions for one are assembling.
US rates, DXY, oil, stress — live pipeline, regularized out
by L1 as structural family captures same signal
ACTIVE
ABSORBED BY STRUCTURAL
Ship Connectivity
UNCTAD LSCI — trade network access, best new signal +0.0052
AUC lift
CANDIDATE
PENDING LOCK
GDELT Events
Event volume + media tone — 75% panel coverage, ablation
pending
CANDIDATE
ABLATION PENDING
ZESI / NDVI / VIIRS
Satellite physical stress — research layer, lead-time
validation in progress
EXPERIMENTAL
RESEARCH LAYER
Sanctions (GSDB v4)
Active sanctions count — integrated, tested negative lift as
sanctions occur after macro deterioration already detected
INTEGRATED
LAGGING INDICATOR
Why L1 Regularization
L1 forces redundant signals to
exactly zero
— it's a built-in feature selector. If two signals carry the
same information, one is zeroed automatically.
In practice: all 5 FRED variables were zeroed because the
structural family already captured them. The model retained only
signals adding genuinely new information.
This is why ZEVKORA avoids overfitting with 29+ input signals —
the model selects its own features.
Ablation Experiment Results — Lift vs Baseline
FEATURE FAMILY
Baseline only0.7586--
+ World Bank0.7586+0.0000
+ FRED global0.7613+0.0027
+ ZESI physical0.7593-0.0020
SHIPPING ABLATION
Baseline0.7569--
+ Ship Activity0.7581+0.0012
+ Ship Connect ★0.7621+0.0052
+ Both Shipping0.7595+0.0026
Full Physical0.7602+0.0033
SANCTIONS ABLATION
Baseline0.7659--
+ Ship Connect0.7621-0.0038
+ Sanctions0.7620-0.0039
+ Ship+Sanctions0.7603-0.0056
Full+Sanctions0.7611-0.0048
Industry Benchmark
AUC 0.65–0.70Typical IMF EWS
AUC 0.70–0.75Academic benchmark
AUC 0.75–0.80Strong performance
AUC >0.80Rare in literature
ZEVKORA: Avg 0.756 · Peak 0.896 — TOP OF STRONG
BAND
Key Design Decisions
✓
Walk-forward only
Train on past, test on future. No cross-validation. No data
leakage.
✓
fillna(0) for missing
Missing signals treated as zero deviation — conservative
assumption.
✓
Binary event definition
Currency collapse, debt default, IMF emergency, political
rupture.
✓
24-month horizon
Optimized for institutional decision cycles.
✓
L1 self-cleaning
Model selects its own features. No manual pruning, no
researcher bias.
Signal Convergence
MULTI-SIGNAL ALIGNMENT · PAST 5 YEARS · PRESENT
· 24-MONTH FORWARD PROBABILITY CONE