3-month forecasts
Landed inside the range
60%
of 13,712 forecasts checked
Typical miss
1.7 pts
median gap between forecast and outcome, as a share of the ZIP's price level
Forecasts from December 2025 on · checked through July 2026 data
Track record · updates as new data lands
The forecast, checked against what happened.
Every month the HavenScore Forecast Model (v1) publishes a price range for every scored ZIP at 3, 6, and 12 months out. When each horizon arrives, we compare the range we published with the median value that actually landed, and post the result here — hits and misses, both.
Forecasts published since April 2026 · checked against data through July 2026
01 · How to read the scorecard
Two numbers, no adjectives.
Landed inside the range — the share of forecasts where the ZIP's realized median value fell inside the range we published. The range is built to catch about four in five outcomes, so 80% is the mark to read this number against — above it the ranges ran wide, below it they ran tight.
Typical miss— the median gap between the forecast's central estimate and the realized value, in percentage points of the ZIP's price level. Half of the forecasts landed closer than this, half farther.
Realized values are the same Zillow-based median home values published on every ZIP page. Each ZIP's live forecast is scored once per horizon when its check date arrives; results stay in the tally whether they land or miss.
02 · The scorecard
Results by horizon.
6-month forecasts
Landed inside the range
60%
of 2,654 forecasts checked
Typical miss
3.5 pts
median gap between forecast and outcome, as a share of the ZIP's price level
Forecasts from December 2025 on · checked through July 2026 data
12-month forecasts
No 12-month forecast has reached its check date yet. The first batch gets scored once December 2026 data lands — results appear here then, whatever they show.
311,989 forecasts published and waiting to be scored
03 · Read it honestly
What this page can and can't tell you.
This track record is young. Early results are a start, not a verdict — the table fills in as more forecasts reach their check dates, and we post each month's results either way.
Source data gets revised upstream, so past results here can shift slightly as revisions land. We recompute from current data every time rather than freezing a flattering snapshot.
A hit rate describes the past; it doesn't guarantee the next forecast. What the published range means is on the forecast methodology page, and the full not-a-guarantee statement is in the disclaimers.