The idea. Every event has a typical sales curve — the shape of how tickets sell over time. We match an event to similar past events (same artist → venue → genre, same city) and use their curve to project where it finishes.

Two modes:

• No sales yet → we estimate from similar artists / venues / capacity (peer prior).
• Already selling → we compare its pace to the typical curve (pace ↑ ahead / ↓ behind) and project the final.

Three scenarios. We show pessimistic / realistic / optimistic, not one number. The width is set by confidence and calibrated so the range contains the true final ~80% of the time.

Confidence blends: how similar the past events are (artist>venue>genre), how far through the sale window we are (biggest factor), whether it's same-day-sale, whether it's exclusive to us, and whether it has sales yet.

Capacity ceiling 🎟️. A forecast can never exceed the venue's seats. If our raw estimate is above capacity, we clamp it and flag sells out. If a low-confidence estimate is far above capacity, we don't trust it — it shows — unreliable instead of a fake sell-out.

⚠ Partial visibility. If an event also sells on other platforms (non-exclusive), we only see Platinumlist sales — so the number is a lower bound.

Tier ≠ capacity. The tier (XS–XL) is the expected revenue bucket, not the venue size. A huge venue can show XS if we expect it to sell little — that's a signal the event needs attention, not a contradiction.

Accuracy. Validated on 6,965 past events: HIGH confidence ~79% within ±30%, MEDIUM ~69%. Accuracy improves as the event nears (median error 49% at 60 days out → 10% at 1 day out). Trust HIGH/MEDIUM; treat LOW and partial-visibility as direction, not a quote.