Bunker Demand

Bunker calls vs the same days in past years

Ships declaring a bunkering call per day, 28-day average, by day of year.

Forward demand: 131 bunkering calls expected in the next 24 h

Ships due in the next 72 h, each weighted by its calibrated probability of declaring bunkering (learned from its own past calls). 6-hour blocks.

Year-on-year change by segment

Million GT (Gross Tonnage) per day vs 12 months earlier.

The monthly read is 4.1% off with 14 days known

Bunker-sales nowcast error by days of the month observed. MPA publishes 16 d after month end.

Refuel-only stops: 36.7% of bunker calls

Calls declaring bunkering and nothing else. Discretionary stops: a read on Singapore's price competitiveness vs rival hubs.

Bunker signals against normal

Each against its own normal range: unusual = outside the 10th-90th percentile.

Ships due in 72 h, most likely to bunker first

Demand surprise: bunker sales +2.2% vs expected; direction right in 82.8% of months

Expected = same change as a year ago (what a desk without this data would assume). Nowcast and official figure, out-of-sample backtest, last 3 years. Pick another series above.

Which ships are buying fuel

GT (Gross Tonnage) of ships declaring bunkering, million GT per day, 3-month average. Only months where MPA can still size ≥ 85% of callers (scrapped ships drop out of its particulars).

The propensity model checks out: AUC 0.89

Out of sample, deep-sea calls. Raw score vs actual bunkering rate by decile; dots above the line = raw score too low (corrected before use).