voidly
Atlas · 1d / 7d / 30d forecasts

Multi-horizon forecast index

Three honest forecast horizons per country — each with its own XGBoost model, its own SHAP attribution, and its own 90% conformal interval. Click a country for per-horizon detail.

Trained 2026-05-21 · Recommendation: ship_all_three · model info

1d LOCO AUC
0.907
7d LOCO AUC
0.883
30d LOCO AUC
0.845

20 spotlight countries · ranked by 30-day risk

#Country1d7d30d30d range (90%)Cons.
1Iran IR22%36%95%
[45, 100]
ok
2China CN22%36%95%
[45, 100]
ok
3Azerbaijan AZ27%59%91%
[41, 100]
ok
4Uzbekistan UZ27%59%91%
[48, 100]
ok
5India IN22%59%91%
[41, 100]
ok
6Egypt EG36%59%83%
[33, 100]
ok
7Venezuela VE36%36%83%
[33, 100]
ok
8Myanmar MM36%36%83%
[42, 100]
ok
9Pakistan PK36%59%83%
[33, 100]
ok
10Turkey TR36%36%83%
[33, 100]
ok
11Russia RU22%36%70%
[20, 100]
ok
12Belarus BY15%59%70%
[29, 100]
ok
13Cuba CU15%36%70%
[20, 100]
ok
14Ethiopia ET22%49%70%
[20, 100]
ok
15Turkmenistan TM22%36%70%
[20, 100]
ok
16TJ TJ27%59%70%
[20, 100]
ok
17Kazakhstan KZ27%59%70%
[20, 100]
ok
18AF AF27%59%70%
[20, 100]
ok
19Bangladesh BD27%59%70%
[20, 100]
ok
20Saudi Arabia SA15%59%70%
[20, 100]
ok

Cons. = monotonicity (P(1d) ≤ P(7d) ≤ P(30d)). 30d range = 90% conformal interval.

Why three horizons?

A single 7-day number is what most journalists ask for, but it hides two important things: what is going to happen tomorrow, and what the operational ceiling is over the next month. We train three independent XGBoost models and publish all three with honest LOCO AUC numbers and per-horizon SHAP attribution.

The monotonicity check is a free honesty signal: longer windows must contain shorter windows, so if P(1d) > P(30d) the three models are disagreeing and you should treat the headline numbers with caution. See /atlas/models for the full registry, or /methodology for the full pipeline.

Related