voidly
LB · per-country backtest

Lebanon forecast vs reality

Every Voidly Sentinel shutdown-risk forecast for Lebanon, plotted against what actually happened. 59 (predicted, observed) pairs from the rolling 30-day evaluation window.

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

Forecasts evaluated
59
since Apr 17
Accuracy @ 0.5
50.8%
30/59 correct
Brier score
0.272
lower is better
Observed positive rate
25%
mean predicted 16%

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 17Jun 8

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

All 59 predictions (newest first)

Eval dateForecastPred ≥ 0.5?Observed?Correct?
Jun 84.4%no event
Jun 75.3%no event
Jun 65.2%no event
Jun 53.8%no event
Jun 43.5%no event
Jun 34.9%no event
Jun 23.6%no event
Jun 117.0%no event
Jun 13.8%no event
May 314.0%no event
May 303.5%no event
May 293.9%no event
May 2860.0%no event
May 283.4%no event
May 2760.0%no event
May 274.9%no event
May 264.1%no event
May 2550.7%no event
May 254.3%no event
May 2459.7%no event
May 2360.0%no event
May 2362.0%no event
May 2234.4%no event
May 2216.1%no event
May 2182.1%no event
May 206.8%no event
May 196.7%no event
May 186.7%no event
May 174.6%no event
May 163.9%no event
May 1516.1%no event
May 1416.2%no event
May 1315.6%no event
May 1219.1%no event
May 1119.7%no event
May 1017.1%no event
May 917.7%no event
May 819.7%no event
May 714.8%no event
May 613.4%no event
May 511.8%no event
May 410.0%no event
May 38.6%no event
May 28.7%shutdown
May 19.5%shutdown
Apr 309.9%shutdown
Apr 296.9%shutdown
Apr 288.0%shutdown
Apr 278.8%shutdown
Apr 268.1%shutdown
Apr 257.3%shutdown
Apr 247.7%shutdown
Apr 236.9%shutdown
Apr 227.0%shutdown
Apr 216.8%shutdown
Apr 207.1%shutdown
Apr 196.4%shutdown
Apr 186.9%shutdown
Apr 177.5%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