voidly
BY · per-country backtest

Belarus forecast vs reality

Every Voidly Sentinel shutdown-risk forecast for Belarus, plotted against what actually happened. 111 (predicted, observed) pairs from the rolling 30-day evaluation window.

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

Forecasts evaluated
111
since Apr 17
Accuracy @ 0.5
45.9%
51/111 correct
Brier score
0.282
lower is better
Observed positive rate
19%
mean predicted 28%

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 111 predictions (newest first)

Eval dateForecastPred ≥ 0.5?Observed?Correct?
Aug 659.2%no event
Aug 571.8%no event
Aug 477.2%no event
Aug 373.5%no event
Aug 268.7%no event
Aug 166.9%no event
Jul 3170.3%no event
Jul 2944.6%no event
Jul 2849.3%no event
Jul 2756.5%no event
Jul 2667.7%no event
Jul 2565.2%no event
Jul 2464.7%no event
Jul 2362.0%no event
Jul 2267.0%no event
Jul 2163.3%no event
Jul 2059.7%no event
Jul 1962.8%no event
Jul 1860.6%no event
Jul 1758.0%no event
Jul 1665.7%no event
Jul 1569.4%no event
Jul 1458.7%no event
Jul 1356.8%no event
Jul 1254.6%no event
Jul 1144.3%no event
Jul 1045.7%no event
Jul 955.1%no event
Jul 856.2%no event
Jul 758.8%no event
Jul 657.5%no event
Jul 557.5%no event
Jul 466.5%no event
Jul 368.5%no event
Jul 258.8%no event
Jul 161.4%no event
Jun 3064.9%shutdown
Jun 2962.6%shutdown
Jun 2862.1%shutdown
Jun 2766.5%shutdown
Jun 2666.6%shutdown
Jun 2563.3%no event
Jun 2469.6%no event
Jun 2369.9%no event
Jun 2268.7%no event
Jun 212.7%no event
Jun 205.0%no event
Jun 193.3%no event
Jun 184.0%no event
Jun 173.0%no event
Jun 164.0%no event
Jun 155.5%no event
Jun 145.0%no event
Jun 134.9%no event
Jun 123.1%no event
Jun 114.5%no event
Jun 103.6%no event
Jun 94.8%no event
Jun 84.2%no event
Jun 74.0%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