voidly

ChinaCN

climbing (+24.7pp)

30-day day-by-day shutdown-risk trajectory. Each point is a per-horizon XGBoost prediction (h=1..30) with a 90% conformal band derived from absolute LOCO residuals. Peak risk highlighted.

Peak day
70.2%
d21 · 2026-06-12
Floor day
2.3%
d1 · 2026-05-23
30-day delta
+24.7pp
d30 vs d1 probability

30-day trajectory with 90% conformal band

model trajectory_v1
0%25%50%75%100%d1d7d14d21d30peak d21 70%
mean P(shutdown by day)90% conformal bandpeak day

Drivers (SHAP top 3 from the 7-day model — same features feed all horizons)

day 12.3%
2026-05-23
  • block_rate_roll7_std-0.013
  • week_of_year+0.013
  • incident_count_7d-0.010
day 74.5%
2026-05-29
  • block_rate_roll7_std-0.013
  • week_of_year+0.013
  • incident_count_7d-0.010
day 1429.7%
2026-06-05
  • block_rate_roll7_std-0.013
  • week_of_year+0.013
  • incident_count_7d-0.010
day 3027.0%
2026-06-21
  • block_rate_roll7_std-0.013
  • week_of_year+0.013
  • incident_count_7d-0.010

Honest caveats

  • Per-horizon XGBoost stacked into a trajectory — not a true joint distribution.
  • Conformal bands are absolute-residual q90 from LOCO predictions on 18 eval countries.
  • Long horizons (h>=14d) have wider effective uncertainty than the conformal q90 suggests; the same lag features feed all horizons.
  • Bands floor at [0,1]; probabilities below 0.05 are at the model's noise floor.
  • SHAP attribution above is from the 7-day point model. The trajectory uses 30 separate per-horizon XGBoost models that share the same feature set; we have not yet computed per-horizon SHAP.

Sample horizons

DayDateP(shutdown)90% interval
d12026-05-232.3%[0%, 46%]
d72026-05-294.5%[0%, 69%]
d142026-06-0529.7%[0%, 99%]
d302026-06-2127.0%[0%, 100%]
Spotlight countries