The Forecast Accuracy Ledger
Every month, when actual home prices publish, we grade the previous month's forecasts against them: ours, and Zillow's published forecast, metro by metro. Wins, losses, and the size of every miss go here. A forecaster who won't show you the scoreboard is asking for faith.
Like-for-like: the one-month move
Zillow restates its whole price history every month. By the time new prices publish, the starting point of last month's forecast has moved by a median of about 1.0%. Grading a forecast level against restated data credits or blames the forecaster for that rewrite. This table applies each forecast's predicted change to the restated starting point, so it grades the month alone. The no-change column is the score of a forecast that predicts zero movement.
On this basis the two engines are close, and both beat the no-change forecast in every graded month.
| Forecast month | Sialim median miss | Zillow median miss | No-change median miss |
|---|---|---|---|
| June 2026 | 0.095% | 0.093% | 0.196% |
| July 2026 | 0.099% | 0.093% | 0.208% |
| August 2026 | 0.087% | 0.091% | 0.234% |
Both engines are graded against the same realized values from the same Zillow release. For months graded before we archived each release, the starting point and the outcome come from the latest release instead of the one that first published them.
Against the published level
The original scoreboard. It includes the restatement of the starting month, which neither engine controls.
| Forecast month | Sialim median miss | Zillow median miss | No-change median miss |
|---|---|---|---|
| June 2026 | 0.89% | 0.93% | 0.91% |
| July 2026 | 0.67% | 0.64% | 0.59% |
| August 2026 | 0.89% | 0.86% | 0.77% |
A "win" is the engine whose forecast landed closer to the realized value for that metro. Both engines are graded against identical realized values. Zillow forecast © Zillow, graded here for comparison.