Day-Ahead & Extended Evaluation

Work in progress — playground deployment

This page documents the evaluation framework currently deployed on the playground. It is a draft and may change before the production rollout — treat figures and examples here as preliminary.

Day-Ahead (D+1) and Extended (D+2 through D+N) challenges request a forecast for a specific calendar day. Each forecast covers exactly that day, so there's no lead-time slot model and no fallback chain. Your submission contributes equally at every timestamp.

For the universal mechanics. Per-submission contribution, per-timestamp aggregation, and daily score, see Evaluation. This page covers what's specific to D+N.

One session per target day

Today, Predico opens day ahead forecasting challenges at one specific session (and gate closure), per resource. So for any target day D and horizon h, there is exactly one challenge to submit to.

One gate, one submission per (target day, horizon) slice. The submission's contribution is what's evaluated at every timestamp of D.

Qualification

Submit your full Q10, Q50, Q90

Your day qualifies if you submit your full Q10, Q50, Q90 to the opened session for that target day and horizon. Skipping it triggers a peer-Q75 penalty for that (day, horizon) slice.

If your day fails, your score is replaced by the 75th-percentile of qualifying peers' scores. That counts as a "missed day" for monthly league eligibility (see Rewards).

Peer-Q75 penalties are only emitted when at least three peers also submitted real scores for the same slice; otherwise the slice is left unpenalised. The longer the horizon, the lower the typical participation, so D+N pools may be smaller.

Looking ahead — multiple sessions per day

The framework generalises to multiple sessions per (target day, horizon): every submission stacks vertically over D and is averaged at each timestamp, and qualification requires a full Q10/Q50/Q90 to every opened session. We'll document those mechanics here when additional gates go live.