Time-weighted Scoring
Time-weighted scoring gives different weights to forecasts based on when they were made, for example weighting earlier forecasts more to reward early insight.
Definition
Time-weighted scoring applies weights to forecasts based on time, rather than treating every forecast equally. A common goal is to reward earlier, harder forecasts more than last minute updates.
Why it matters
Forecasts close to settlement often benefit from public information and can be easier. Time-weighting can make evaluation fairer by accounting for forecast horizon.
Common approaches
• Weight by time to resolution (longer horizon = higher weight).
• Score at fixed checkpoints (T-30d, T-7d, T-24h) and average.
• Decay weights for late updates, but keep an audit trail of all changes.
Tradeoffs
Time weighting adds complexity and can be gamed if rules are unclear. If used, document it in methodology and keep the implementation consistent.