Ratings and reputation
See how seasons, coverage, and long-term accuracy build trust in a forecaster.
Plura separates a season rating from long-term reputation. They answer different questions: who performed best across a defined event series, and how stable a person’s accuracy has been over a longer history.
Seasons
Each season has a predefined event set. Once the season starts, its membership and sequence cannot change. This prevents selecting only convenient completed events after the fact.
A skipped event contributes zero points and zero coverage, but it remains in the season’s shared denominator.
When a result becomes ranked
A profile remains provisional until both requirements are met:
- forecasts exist for at least 20 resolved events;
- average time coverage is at least 50%.
These thresholds prevent one lucky outcome from looking like established skill.
Long-term reputation
Reputation incorporates history gradually. A small sample is pulled strongly toward a neutral level; trust grows only after sustained accuracy.
Established reputation may produce a separate secondary signal, but its influence is bounded: the weight starts at 1 and never exceeds 2. The ordinary community median remains separate and equal-weighted.
What does not improve accuracy
Account age, login frequency, and any earlier activity outside probability forecasting do not improve a season result.
Always read a result together with event count and coverage. A number without that context can create false confidence.