Prediction track record
Every Defaxon forecast is recorded before its outcome window opens, then scored against what actually happened. This is that record, in the open — confirmed predictions with the date we called them, the probability we gave and when they came true, alongside the misses. We show the misses on purpose: a track record that hides them can’t be trusted, and a 40%-likely call is meant to be wrong much of the time. In-sample backtests never count here — only genuine, pre-registered, out-of-sample forecasts.
Confirmed predictions
Open predictions
Forecasts we’re making right now, before the fact — public and timestamped, so you can come back and check. Each is genuinely predictable (recurring or significantly correlated), never the daily firehose.
No open predictions in the next 72 hours.
Full ledger
Every scored forecast — hits and misses — newest first.
| Outcome | Prediction | Predicted on | Probability | Occurred |
|---|---|---|---|---|
| ✓ Hit | Flight route KSMF→KLAS | Jun 14, 2026 | 85% | Jun 16, 2026 |
| ✓ Hit | Coastal Flood Advisory in US | Jun 14, 2026 | 85% | Jun 16, 2026 |
| ✓ Hit | Coastal Flood Statement in US | Jun 14, 2026 | 85% | Jun 17, 2026 |
| ✓ Hit | Baseball | Jun 13, 2026 | 85% | Jun 16, 2026 |
| ✓ Hit | Wildfire | Jun 14, 2026 | 47% | Jun 16, 2026 |
| ✗ Miss | Flight route KSMF→KLAS | Jun 14, 2026 | 85% | — |
| ✓ Hit | Coastal Flood Advisory in US | Jun 14, 2026 | 85% | Jun 15, 2026 |
| ✓ Hit | Coastal Flood Statement in US | Jun 14, 2026 | 85% | Jun 15, 2026 |
| ✓ Hit | Baseball | Jun 13, 2026 | 85% | Jun 15, 2026 |
| ✓ Hit | Wildfire | Jun 14, 2026 | 47% | Jun 15, 2026 |
| ✓ Hit | Baseball | Jun 13, 2026 | 85% | Jun 14, 2026 |
| ✓ Hit | Wildfire | Jun 13, 2026 | 47% | Jun 14, 2026 |
How scoring works
How do you decide a prediction 'came true'?
Every forecast is recorded BEFORE its outcome window opens, with a target time and a probability. When the window closes we check whether a matching real event actually occurred — a hit if it did, a miss if it didn't. Only these out-of-sample, pre-registered forecasts appear here; retrospective backtests (postdiction) are never counted as predictions.
Why do you show the misses too?
Because a track record that hides its misses can't be trusted. A 47%-likely forecast is supposed to be wrong about half the time — showing the full ledger is what makes the confirmed predictions credible, and lets you judge whether our probabilities are honest.
Why so few entries so far?
The prediction engine is young and deliberately conservative — it forecasts only genuinely recurring or significantly-correlated patterns, never the daily firehose. The ledger grows as forecasts come due and as the historical archive deepens. We'd rather show a short honest record than a padded one.