MARINACHAIN × CMA CGM VIETNAM
Prediction Accuracy
Self-scored on live AIS traffic since 6 Jul 2026. Predictions are persisted when issued and graded against the recorded arrival — grades are computed from ledger data alone, stamped at arrival time, and never adjusted afterward. Scored population: transits ≥ 5 km and ≥ 45 min ending in an observed arrival at the predicted corridor's destination; only predictions issued for that corridor are graded.
SCORED TRANSITS
1,365
FIRST SCORED
6 Jul 2026
MODEL VERSIONS
v1-heuristic: 0 · v2-factors: 87,258
Horizon Error
| HORIZON | N | MAE | MEDIAN ABS | P90 ABS | % WITHIN BAND |
|---|---|---|---|---|---|
| at transit open | 1,365 | 67.2 min | 33.1 min | 106.5 min | 38.2% |
| T−3h | 226 | 84.5 min | 83.1 min | 151.9 min | 21.7% |
| T−1h | 1,221 | 27.9 min | 28.3 min | 46.2 min | 40% |
Aggregated per-transit first, then across transits.
Per-Corridor n
cor-catlai-caimep · n=1,291
cor-vungtau-dongnai · n=73
cor-vungtau-binhduong · n=1
Connections
SCORED SUBJECTS
25
MADE RATE
36%
MEAN MISS PROBABILITY
77.2%
Terminal Queues & Idle
Per-queue waits are measured from AIS stationary episodes attributed awaiting_berth; fleet idle hours and est. CO₂ count all measured stationary time (7-day window).
No terminal queue cohorts with n≥3 yet.
fleet idle hours (7d): 0
est. idle CO₂ (7d) · 0 kg
METHODOLOGY
Formula: measured fleet idle hours × 18 L/h idle burn × 2.68 kg CO₂/L diesel. Sources: DEFRA/BEIS GHG Conversion Factors 2024 for diesel; idle burn is an estimate, band 10–25 L/h. Counts measured stationary time only. No demo actions or hypothetical savings are accumulated.
closed idle episodes (7d): 0