Seven windows tested. Two failed. All shown.
We score every district using only pre-cutoff data, then measure how well those scores predicted subsequent price growth. No cherry-picking.
Spearman rank correlation (HPSSA 5yr)
Top-decile hit rate vs random baseline
Q4/Q1 quartile lift over 5 years
Districts tested (all England & Wales)
All tested windows
n = 2,292 per window| Window | Source | ρ |
|---|---|---|
| 2018 to 2023 | ONS HPSSA | +0.236 |
| 2019 to 2024 | Land Registry | +0.188 |
| 2013 to 2018 | HPSSA | +0.548 |
| 2010 to 2015 | HPSSA | +0.614 |
| 2016 to 2018 | HPSSA | +0.197 |
| 2017 to 2022 | HPSSA | −0.321 |
| 2019 to 2024 | HPSSA (recon) | −0.100 |
What a score band has historically meant
13 cohorts, 2022 to 2023| Score band | Beat county median |
|---|---|
| 60+ | 66% |
| 50 to 60 | 53% |
| 40 to 50 | 44% |
| under 40 | 36% |
Share of districts in each band whose realized three year price growth beat their county median, measured across 13 monthly scoring cohorts from Jan 2022 to Jan 2023. Cohorts are backfilled reconstructions and windows overlap. Backtest evidence, not a prediction.
How the test works
- 01Reconstruct signal values using only data available before the cutoff date.
- 02Score all 2,292 districts. The engine sees nothing from after the cutoff.
- 03Measure Spearman ρ between score rank and actual price growth rank over the following 2 to 5 years.
The 2018+ windows are genuinely out of sample. Earlier windows informed signal selection and are labelled “reconstruction.” Temporal holdout is used instead of random split to avoid spatial leakage between neighbouring districts.
Where it breaks
The model identifies structural conditions that historically precede price growth: business formation, deprivation retreat, transport investment. It works in every window where structural factors drove prices.
It does not predict macro regime changes. COVID triggered the largest urban to rural price shift in modern history and inverted the signal (ρ = −0.321). The post-COVID rate shock produced a mild negative (ρ = −0.100), now trending toward neutral.
If you believe another disruption of that scale is imminent, this model will not warn you.
Past performance does not guarantee future results. Trajectory scores are statistical indicators derived from publicly available data, not financial advice. Property investment carries risk including loss of capital. Always seek independent financial advice before investing.