The ledger: every prediction, scored against what happened
Who called it?
GLOVIZ does not try to out-forecast the institutions. It keeps their receipts. Every forecast that enters the database, from a weather model, a grid operator or the IMF, is stored with a timestamp, and when the outcome arrives the error is computed. The naive baseline, tomorrow equals today, is the honesty benchmark: a forecaster that cannot beat it is not adding information.
Forecast against outcome, the British gridNational Grid · 30 minutes · gCO2/kWh · the cleanest test bench in the catalogue
Source: National Grid ESO Carbon Intensity API
The scoreboard
Mean absolute error per predictor, lowest first. Rows appear as forecasts mature: a prediction can only be scored after its target time has passed.
Grid CO2 intensity GBgCO2/kWh · lower is better
| Predictor | Scored points | MAE |
|---|---|---|
| Naive (tomorrow = today) | 3 | 0 |
| Drift (recent trend continues) | 3 | 0.66 |
Methodology: forecasts are captured at ingest, before the outcome can overwrite them. Scoring matches the nearest observation within 30 minutes of the target. Ingestion status.