Track record
How accurate were our predictions?
Every week we forecast where each city's hotel and flight prices will be in 4 weeks, as a range we expect 4 in 5 forecasts to land inside. We log each one before we know the answer, never edit it, and grade it against what our own tracking later measures. This page is the scorecard, misses included.
- Typical miss
- —
- “No change” would have hit
- —
- Waiting to be graded
- 0
- Voided
- 0
- Typical miss
- —
- “No change” would have hit
- —
- Waiting to be graded
- 79
- Voided
- 0
How we grade
What we forecast. For hotels, the typical (median) nightly rate across the 30 hotels we sample in each city for one fixed weekend, over the last 30 days. For flights, the typical round trip from New York across the next two months of departure dates: the same figure our trip receipt uses.
How. The forecast leans a little from today's price back toward the city's own recent average, and the range comes from how far our past 4-week moves actually went. While that history is short, the range is a deliberately wide default.
Grading. In range means the price we measured on the target date fell inside the range. A forecast is voided, not counted, only when there was nothing to measure, for example when we moved the hotel weekend we track. “No change would have hit” is the bar to beat: the same range around today's price, as if we'd predicted nothing would move.