voidly
SD · per-country backtest

Sudan forecast vs reality

Every Voidly Sentinel shutdown-risk forecast for Sudan, plotted against what actually happened. 112 (predicted, observed) pairs from the rolling 30-day evaluation window.

Updated every 30 min · CC BY 4.0 · Raw JSON · Current forecast →

Forecasts evaluated
112
since Apr 17
Accuracy @ 0.5
69.6%
78/112 correct
Brier score
0.126
lower is better
Observed positive rate
6%
mean predicted 24%

Forecast time series

Blue line: forecast probability. Green ✓ markers: forecast was right. Red ✗ markers: forecast was wrong. Dashed line at 0.5 is the binary decision threshold.

0.000.250.500.751.00threshold 0.50Apr 17Aug 6

Y axis: forecast probability · Faint tick: distance to observed outcome (0 or 1)

All 112 predictions (newest first)

Eval dateForecastPred ≥ 0.5?Observed?Correct?
Aug 676.1%no event
Aug 571.5%no event
Aug 471.5%no event
Aug 371.5%no event
Aug 273.3%no event
Aug 173.3%no event
Jul 3173.3%no event
Jul 3071.6%no event
Jul 2966.5%no event
Jul 2817.9%no event
Jul 2717.9%no event
Jul 2619.3%no event
Jul 2516.2%no event
Jul 2416.2%no event
Jul 2316.2%no event
Jul 2212.7%no event
Jul 2112.7%no event
Jul 2012.7%no event
Jul 1914.2%no event
Jul 1814.2%no event
Jul 1714.2%no event
Jul 1621.2%no event
Jul 1521.2%no event
Jul 1417.0%no event
Jul 1319.2%no event
Jul 127.2%no event
Jul 117.2%no event
Jul 106.7%no event
Jul 93.0%no event
Jul 834.0%no event
Jul 726.9%no event
Jul 626.9%no event
Jul 534.6%no event
Jul 434.6%no event
Jul 334.6%no event
Jul 234.6%no event
Jul 159.3%no event
Jun 3059.3%no event
Jun 2970.0%no event
Jun 2866.4%no event
Jun 2766.4%no event
Jun 2666.4%no event
Jun 2570.1%no event
Jun 2470.1%no event
Jun 2378.4%no event
Jun 2275.2%no event
Jun 214.2%no event
Jun 205.4%no event
Jun 196.1%no event
Jun 185.5%no event
Jun 174.9%no event
Jun 165.2%no event
Jun 155.1%no event
Jun 143.3%no event
Jun 134.2%no event
Jun 124.5%no event
Jun 114.9%no event
Jun 105.6%no event
Jun 94.6%no event
Jun 84.1%no event

How to read this

  • Each row is one historical forecast. We made the prediction at eval_date, then waited the 7-day horizon, then graded against the observed outcome.
  • Forecast % is the calibrated probability we published that day. Post-recalibration (2026-05-20) these now match actual observed rates much better — see the refit finding.
  • Pred ≥ 0.5 shows the binary alert decision. Note we usually fire alerts at a lower threshold (see /v1/sentinel/global_heatmap for the live cutoff) — the 0.5 column here is for backtest scoring consistency with the global confusion matrix.
  • Correct = both (pred ≥ 0.5) and (observed) agree.

Related