How the ice score works

1. Inputs and weights

2. Time decay

Each report's weight fades as weight × e^(−age/τ): τ = 6 h for skater reports, 12 h for warden and official reports, 24 h for an official status. Reports older than 48 hours are ignored. An official "closed" stops overriding after 36 hours; "closed for the season" stays authoritative for two weeks.

3. Agreement: one report can't overrule strong evidence

A skater report is corroborated when another independent signal agrees with it (within 20 points): a report from a different skater or a warden within 12 hours, a fresh official status, or the weather estimate. Two reports from the same person or device never corroborate each other.

An uncorroborated skater report that contradicts strong evidence — it differs by 40+ points from a warden report or an official status less than 12 hours old, or it reports skateable ice while the weather says closed or slush — keeps only ×0.25 of its weight (×0.5 if signed in). A second, independent skater saying the same thing removes the penalty.

4. Closures

If a fresh official closure exists and no warden has reported since, the rink is closed. If more than half of the weighted signal says closed, the rink is closed. Otherwise the score is the weighted average of the non-closed signals.

5. Labels

• Glass 90–100
• Good ice 70–89
• Fair 50–69
• Rough 30–49
• Slush 10–29
• Closed 0–9

6. Confidence

Confidence = (1 − e^(−human weight / 2.5)) × (1 − 0.5 × disagreement), capped at 97%, where disagreement is the weighted spread of the human signals (0 when they all agree). If the only human signals are skater reports nobody corroborates, confidence is capped at 35%. The weather estimate alone gives 20%; one fresh anonymous report about 33%; one fresh warden report about 70%. Low-confidence scores are labelled “estimate”.

7. The weather estimate

From the last 72 hours of hourly temperature, rain and snow (Open-Meteo) for a ~2 km grid cell around the rink: sustained warmth or rain means closed or slush; fresh snow means rough; a long deep freeze means good ice. It is a prior: corroborated reports outweigh it, but a single report that claims good ice in a thaw does not.