Data and methodology. All prices shown are US customs transaction prices, which represent the declared value of snow crab imports divided by volume. The model uses the March customs price as its pre-season anchor, but testing confirmed this choice is not critical: substituting January, January-February, or the full January-March average produces nearly identical results (LOO MAPE ranges from 11.0% to 13.0% across all anchor variants). March volume is thin in most years (median 339,000 lbs), but January-February combined is 2.7 times larger and yields the same forecast. The model is robust to the anchor period.

Validation. The LOO MAPE (Leave-One-Out Mean Absolute Percentage Error) of 7.0% is the model's honest accuracy measure. It works by removing each pack year in turn, fitting the model on the remaining years, and predicting the held-out year blind. The average of those blind prediction errors is 7.0%, meaning the model is typically off by about $0.40 on a $6 price or $0.60 on a $9 price when forecasting a year it has never seen. This is the real-world accuracy to expect for PY2026.

2026 forecast. March customs at $10.69 with a rising off-season (+5.8% January to March) and no inventory carryover produces a point forecast of $8.77/lb, with an 80% probability range of $7.17 to $10.72.