voidly
PK · per-country backtest

Pakistan forecast vs reality

Every Voidly Sentinel shutdown-risk forecast for Pakistan, 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
25.0%
28/112 correct
Brier score
0.580
lower is better
Observed positive rate
34%
mean predicted 59%

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 675.5%no event
Aug 568.6%no event
Aug 469.7%no event
Aug 380.7%no event
Aug 278.9%no event
Aug 178.0%no event
Jul 3176.6%no event
Jul 3075.8%no event
Jul 2973.0%no event
Jul 2869.8%no event
Jul 2771.1%no event
Jul 2680.7%no event
Jul 2575.8%no event
Jul 2473.9%no event
Jul 2386.3%no event
Jul 2271.0%shutdown
Jul 2184.0%shutdown
Jul 2074.6%shutdown
Jul 1969.5%shutdown
Jul 1875.1%no event
Jul 1769.8%no event
Jul 1671.0%no event
Jul 1574.0%no event
Jul 1476.0%no event
Jul 1368.9%no event
Jul 1276.4%no event
Jul 1176.7%no event
Jul 1080.4%no event
Jul 981.6%no event
Jul 878.3%no event
Jul 781.2%no event
Jul 678.9%no event
Jul 580.6%no event
Jul 486.8%no event
Jul 378.6%no event
Jul 278.6%no event
Jul 173.7%no event
Jun 3073.7%shutdown
Jun 2966.5%shutdown
Jun 2868.5%shutdown
Jun 2764.7%shutdown
Jun 2669.3%shutdown
Jun 2576.4%no event
Jun 2476.4%no event
Jun 2376.7%no event
Jun 2279.5%no event
Jun 2195.0%no event
Jun 2095.0%no event
Jun 1995.0%no event
Jun 1895.0%no event
Jun 1795.0%no event
Jun 1695.0%no event
Jun 1595.0%no event
Jun 1495.0%no event
Jun 1395.0%no event
Jun 1295.0%no event
Jun 1195.0%no event
Jun 1095.0%no event
Jun 995.0%no event
Jun 895.0%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