Which inflection signals held up historically?
Each signal class (A–F) is scored on every month it newly fired in a county since …, by what that county's typical home value did over the next 3, 6 and 12 months. This is a backward-looking scorecard of a price-only rule set. It does not forecast prices and is not investment advice. Ranked by 12-month excess hit rate.
| # | Signal | Dir. | Fired | Hit 3M / base | Hit 6M / base | Hit 12M / base | Median fwd 12M | Mean fwd 12M | Excess median | Excess mean | Excess hit 3M | Excess hit 6M | Excess hit 12M |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
Hit = county value moved in the signal's direction over the window; base = share of all county-months moving that way over the same window. Excess = the county's forward change minus the median forward change of all counties over the same months. Excess hit = share of signals that beat (up classes) or lagged (down classes) that all-county median; 50% is a coin flip. Green/red show whether the result is on the signal's side.
| Signal | ||||
|---|---|---|---|---|
Periods are by the month the signal fired (2020–present only includes signals with a complete forward window). Value tiers are terciles of county ZHVI across all counties in the month the signal fired. County size uses Census Vintage 2020 population.
- Input: Zillow ZHVI county series (all homes, mid-tier, smoothed, seasonally adjusted). No interpolation across missing months.
- Event: the first month a county enters a signal class (a new signal), not every month it stays there. Rules:
- A · Potential Recovery r12 < 0, r3 > 0, acceleration > 0
- B · Early Warning r12 > 0, r3 < 0, acceleration < 0
- C · Accelerating Uptrend r12 > 0, r3 > 0, p3 > 0, acceleration ≥ +0.25pp
- D · Accelerating Downturn r12 < 0, r3 < 0, p3 < 0, acceleration ≤ −0.25pp
- E · Potential Peak r12 > 0, 0 ≤ r3 < 0.5%, acceleration ≤ −0.25pp
- F · Potential Bottom r12 < 0, −0.5% < r3 ≤ 0, acceleration ≥ +0.25pp
- Forward change: (Vt+h / Vt − 1) for h = 3, 6, 12 months. Signals without a complete forward window are excluded for that horizon.
- Benchmark: median forward change of all counties with data over the same months. Excess = county forward change − benchmark.
- Ranking: 12-month excess hit rate, so a class is not rewarded simply for firing in a rising national market.
- Sample floor: cells with fewer than 30 scored signals show “insufficient” instead of a number.
- Caveats: computed on today's revised ZHVI history (not point-in-time data); signals overlap in time and space so observations are not independent; results describe the past only.
See Methodology and the Inflection Screener. Not investment advice; see the disclaimer.