{"as_of":"2026-10-23","combiner":"isotonic(sigmoid(0.6*logit(shutdown_risk_v4_prob) + 0.40*logit(honest_forecast_v1_prob)))","full_panel_auc":0.9328467969432344,"generated_at":"2026-10-23T06:30:51.697132+00:00","honest_caveats":["It mostly ranks countries by how often they have had shutdowns. It has not been shown to predict when a new shutdown will start; within-country timing is unvalidated (B1600).","The served version can change daily: an automatic gate chooses it. Read model_version before citing anything.","In the strict past->future holdout (isotonic fitted on the first 75% of dates, scored on the rest): cross-country AUC 0.917, country-ranking AUC 0.951, PR lift 9.40x over the base rate; within-country median AUC 0.440 over 13 countries, below chance (0.5). The cross-country figures compare countries; they do not time a shutdown within a country. The in-sample within-country median is cross-sectional (full panel, in-sample isotonic), not forward skill.","v9 replaces v5's multiplicative product with a logit-additive blend (alpha=0.6 on structural risk). The product collapsed to ~0 when honest_forecast was near-zero for a country, annihilating a good structural signal; the logit blend cannot zero out either factor.","In-sample within-country median AUC 0.600 (full panel, in-sample isotonic; not forward skill), lower than the 0.729 cited for v5 before: a drop, not an improvement. In the strict past->future holdout the v9 within-country median AUC is 0.440 over 13 countries, below chance (0.5); the forward cross-country AUC is 0.917. Leave-one-country-out selection only chose the blend weight (alpha=0.60); it does not show within-country timing skill.","Still country-level; sub-national shutdowns (~half of KeepItOn events) remain harder. Low-N countries are sample-size limited, not model-limited.","Same two honestly-validated base models as v5 (v4 structural + honest_forecast_v1). No new external features (unlike the gated v7 CF-Radar and v8 GDELT attempts) \u2014 purely a better combiner."],"honest_validation":{"forward_country_ranking_auc":0.9509001636661212,"forward_temporal_crosscountry_auc":0.9170363383104134,"forward_temporal_crosscountry_pr_lift":9.400199904444916,"forward_temporal_n_countries":13,"forward_temporal_within_median":0.4403034519313589,"full_panel_auc":0.9328467969432344,"full_panel_pr_auc":0.35165996720773224,"method":"full-panel scoring; within-country AUCs for countries with >=5 KeepItOn positive days + both classes; alpha chosen by leave-one-country-out","n_countries_scored_within":24,"n_countries_within_auc_ge_065":5,"n_countries_within_auc_ge_080":1,"per_country":[{"country":"IN","ensemble_auc":0.6150273202365462,"hf_only_auc":0.5316900197827076,"n":1750,"pos":972,"sr_only_auc":0.6444117818189512},{"country":"MM","ensemble_auc":0.6375712573686597,"hf_only_auc":0.3936247597762951,"n":1750,"pos":718,"sr_only_auc":0.6923004253848979},{"country":"IR","ensemble_auc":0.6167435287956461,"hf_only_auc":0.618664517687319,"n":1750,"pos":307,"sr_only_auc":0.49534989763002785},{"country":"PK","ensemble_auc":0.6190974923693817,"hf_only_auc":0.5635884852474714,"n":1750,"pos":231,"sr_only_auc":0.5894755891464253},{"country":"RU","ensemble_auc":0.6231603773584906,"hf_only_auc":0.5612558962264151,"n":1750,"pos":160,"sr_only_auc":0.6270636792452831},{"country":"IQ","ensemble_auc":0.5488604394341423,"hf_only_auc":0.5408228022046662,"n":1750,"pos":123,"sr_only_auc":0.5250098690292374},{"country":"SD","ensemble_auc":0.4979558684174901,"hf_only_auc":0.5048835792608637,"n":1749,"pos":89,"sr_only_auc":0.4345708677406255},{"country":"SY","ensemble_auc":0.36305633364456896,"hf_only_auc":0.34989484009091854,"n":1746,"pos":63,"sr_only_auc":0.4773552518650558},{"country":"TZ","ensemble_auc":0.5308475573866981,"hf_only_auc":0.5587522071806945,"n":1749,"pos":50,"sr_only_auc":0.5407239552678046},{"country":"TR","ensemble_auc":0.5857598771430971,"hf_only_auc":0.580510863957576,"n":1750,"pos":49,"sr_only_auc":0.46917779457461994},{"country":"JO","ensemble_auc":0.49332273893163825,"hf_only_auc":0.4736812757890041,"n":1750,"pos":42,"sr_only_auc":0.47970335675253706},{"country":"KE","ensemble_auc":0.5687166221628839,"hf_only_auc":0.537283044058745,"n":1747,"pos":35,"sr_only_auc":0.5785630841121496},{"country":"VE","ensemble_auc":0.8250678294573642,"hf_only_auc":0.8447674418604652,"n":1750,"pos":30,"sr_only_auc":0.3721511627906977},{"country":"GN","ensemble_auc":0.6516345723242275,"hf_only_auc":0.7799820868786387,"n":1623,"pos":28,"sr_only_auc":0.29804075235109717},{"country":"TM","ensemble_auc":0.6526234306073972,"hf_only_auc":0.6845202748557855,"n":1712,"pos":28,"sr_only_auc":0.2638064133016627},{"country":"BY","ensemble_auc":0.5820733151560219,"hf_only_auc":0.36657578010961467,"n":1750,"pos":21,"sr_only_auc":0.7220523837065189},{"country":"ET","ensemble_auc":0.613470489410339,"hf_only_auc":0.4818089178991435,"n":1750,"pos":21,"sr_only_auc":0.7534220165799113},{"country":"NG","ensemble_auc":0.6368876080691643,"hf_only_auc":0.7566488266776451,"n":1749,"pos":14,"sr_only_auc":0.47867435158501437},{"country":"ID","ensemble_auc":0.4435701991640029,"hf_only_auc":0.5420867142037539,"n":1750,"pos":7,"sr_only_auc":0.12707974756167528},{"country":"KG","ensemble_auc":0.6604067574216829,"hf_only_auc":0.6987042807938331,"n":1749,"pos":7,"sr_only_auc":0.346727898966705},{"country":"OM","ensemble_auc":0.023138254150912346,"hf_only_auc":0.022603978300180825,"n":1745,"pos":7,"sr_only_auc":0.07105868814729577},{"country":"TG","ensemble_auc":0.48975477287689273,"hf_only_auc":0.8345951283739301,"n":1743,"pos":7,"sr_only_auc":0.03427419354838712},{"country":"US","ensemble_auc":0.7153921809687731,"hf_only_auc":0.9583640685189739,"n":1750,"pos":7,"sr_only_auc":0.31755593803786575},{"country":"VN","ensemble_auc":0.08150971231866239,"hf_only_auc":0.4117695270879436,"n":1750,"pos":7,"sr_only_auc":0.026678141135972444}],"within_country_auc_mean":0.5448186889656118,"within_country_auc_median":0.599615183276718,"within_country_is_cross_sectional":true},"label_source":"Access Now #KeepItOn STOP dataset (journalist-verified)","loaded":true,"method":"trained AND validated on real, verified shutdown labels; see scripts/train-shutdown-risk-v4.py for the trainer and scripts/validate-against-keepiton.py for the prior honest cross-target validation","model_version":"shutdown_risk_v9","n_countries":60,"recommended_threshold":0.2,"schema":"voidly-shutdown-risk-info/v4","score_formula":"isotonic(sigmoid(0.6*logit(shutdown_risk_v4_prob) + 0.40*logit(honest_forecast_v1_prob)))","within_country_median":0.599615183276718}