voidly
TH · per-country backtest

Thailand forecast vs reality

Every Voidly Sentinel shutdown-risk forecast for Thailand, plotted against what actually happened. 118 (predicted, observed) pairs from the rolling 30-day evaluation window.

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

Forecasts evaluated
118
since Apr 17
Accuracy @ 0.5
65.3%
77/118 correct
Brier score
0.220
lower is better
Observed positive rate
22%
mean predicted 5%

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 12

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

All 118 predictions (newest first)

Eval dateForecastPred ≥ 0.5?Observed?Correct?
Aug 120.6%no event
Aug 110.3%no event
Aug 100.3%no event
Aug 91.5%no event
Aug 80.9%no event
Aug 70.6%no event
Aug 60.8%no event
Aug 52.3%no event
Aug 43.8%no event
Aug 33.7%no event
Aug 23.5%no event
Aug 11.1%no event
Jul 311.5%no event
Jul 301.5%no event
Jul 291.8%no event
Jul 282.4%no event
Jul 272.2%no event
Jul 261.6%no event
Jul 251.8%no event
Jul 241.3%no event
Jul 231.5%no event
Jul 221.7%no event
Jul 211.4%shutdown
Jul 205.5%shutdown
Jul 191.4%shutdown
Jul 180.5%shutdown
Jul 171.1%shutdown
Jul 161.2%no event
Jul 152.1%no event
Jul 142.1%no event
Jul 132.4%no event
Jul 121.6%no event
Jul 112.0%no event
Jul 102.9%no event
Jul 92.8%no event
Jul 85.1%no event
Jul 72.6%no event
Jul 61.8%no event
Jul 52.2%no event
Jul 41.9%no event
Jul 32.2%no event
Jul 21.9%no event
Jul 14.2%no event
Jun 302.1%no event
Jun 292.4%no event
Jun 281.8%no event
Jun 271.6%no event
Jun 261.9%no event
Jun 251.7%no event
Jun 241.8%no event
Jun 231.8%no event
Jun 221.8%no event
Jun 213.6%no event
Jun 205.1%no event
Jun 193.3%no event
Jun 183.0%no event
Jun 175.0%no event
Jun 164.6%no event
Jun 155.7%no event
Jun 145.1%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