Technical Specification

Engine v2.10, data as of 2026 Q2. Every formula, constant and rule behind the published readings, as implemented.

1. Inputs

Every input is quarterly, from 2000 Q1. A market needs 16 quarters of history before its readings begin, and 2020 Q2 – 2021 Q4 is left out of what counts as normal. The last two inputs are national; the rest are measured market by market.

Input	What it measures	How the model uses it
Occupancy rate, o	Share of all multifamily units in the market that are occupied at quarter end; units still leasing up count as unoccupied.	Standing: the rate standardized against its own history, so occupancy above the market’s own norm adds to Standing. Momentum: its change over the past four quarters. Leading: the change in that four-quarter change versus a year earlier, entered negatively.
Effective rent growth, g	Percentage change in average effective rent per unit (asking rent net of concessions) over the trailing 12 months, all units.	Standing: the 12-month growth rate itself. Momentum: that rate versus four quarters earlier.
Supply, s	Units under construction at quarter end, as a share of the market’s existing inventory.	Leading: the level, against the market’s own history.
Price growth	12-month percentage change in a market-wide sale price index covering all multifamily properties.	Repricing: that 12-month growth versus four quarters earlier, so whether price growth is speeding up or slowing down.
Transaction volume	Number of multifamily sales closed in the quarter, less the same quarter a year earlier, divided by the prior year’s average quarterly count; capped between −100% and +200%.	Transactions: the year-over-year change, against the market’s own history.
Price ÷ replacement	Price per unit of 4- and 5-star properties divided by a national construction-cost index.	Replacement: this market’s percentile of the ratio against its own history, plus the share of all markets past their own 90th percentile.
HY spread	Option-adjusted spread of U.S. high-yield corporate bonds over Treasuries (national).	Credit: its percentile against its own history, and whether it is higher than a year earlier.
SLOOS	Net share of banks reporting tighter standards on multifamily real estate loans in the Federal Reserve’s quarterly senior loan officer survey (national).	Credit: the net share tightening in the latest quarter.

2. Operators

Standardization

Z_t(x) = (x_t − μ_t) ÷ σ_t

(2.1)

μ_t and σ_t are the mean and (population) standard deviation of the market’s own observations strictly before t, excluding 2020 Q2 – 2021 Q4. Z is undefined until at least 16 such observations exist. No observation at or after t enters its own standardization.

Z^±2(x) = min(2, max(−2, Z(x)))

(2.2)

Percentile

p_t(x) = share of prior observations (same exclusions, n ≥ 16) strictly below x_t

(2.3)

Change and blending

Δx_t = x_t − x_t−4 · Δ²x_t = Δx_t − Δx_t−4

(2.4)

F̃_t = 0.6·F_t + 0.4·F_t−1

(2.5)

3. Fundamentals (F)

Standing and Momentum

F_S = ½·Z(g) + ½·Z(o)

(3.1)

Occupancy is 1 − vacancy, the form the data arrives in; standardizing either gives the same number with the sign flipped. Occupancy above the market’s own norm adds to Standing.

F_Δ = ½·Z(Δg) + ½·Z(Δo)

(3.2)

Standing and Momentum are not capped. Both are blended by (2.5) to give F̃_S and F̃_Δ.

Strength and Position

R_F = √(F̃_S² + F̃_Δ²)

(3.3)

Position = 25·(F̃_S − F̃_Δ) ÷ max(R_F, 1.5)

(3.4)

Signal strength is used to normalize the Position term while distinguishing stronger directional signals from markets that remain close to normal. It combines a market’s current standing and momentum into a single measure of how far fundamentals have moved from normal conditions.

The Position term divides the directional spread, (F̃_S − F̃_Δ), by overall Signal Strength, √(F̃_S² + F̃_Δ²). This was designed to measure the directional signal relative to its overall magnitude, rather than allowing unusually large raw readings to dominate the score. Because of the way the two components are combined, the formula naturally limits the Position contribution to approximately ±35.

Signal Strength also dampens readings at the other end of the spectrum. When Signal Strength falls below 1.5, the Position term is scaled linearly toward zero rather than receiving its full weight. Because Signal Strength combines the squared standing and momentum terms, a reading below 1.5 generally means both components remain relatively close to zero — or normal conditions. For perspective, if standing and momentum contributed equally, each would be only about 1.06 at a Signal Strength of 1.5. The threshold is therefore set “ex ante” as a reasonable neutral zone around normal conditions, rather than selected afterward to improve historical results.

Leading

F_L = 8·Z^±2(s) − 4·Z^±2(Δ²o)

(3.5)

F = Position + F_L

(3.6)

4. Market Mood (M)

Repricing and Transactions

S_P = −8·Z^±2(Δ Price Growth)

(4.1)

S_T = −4·Z^±2(Transaction Volume)

(4.2)

While S_R > 0, lean-in readings are muted: S_P = max(S_P, 0) and S_T = max(S_T, 0).

Credit

S_C = 10 · 1[national stress] · 1[local transmission]

(4.3)

National stress holds when the HY spread is at or above its 90th expanding percentile (2.3) and higher than four quarters earlier, or when SLOOS net tightening is +40 or more. Local transmission holds when

8·Z^±2(Δ Price Growth) + 4·Z^±2(Transaction Volume) < 0

(4.4)

evaluated before the S_R mute.

Replacement

S_R = 45 · min(1, max(0, (p − 0.90) ÷ 0.10)) · 1[B ≥ 0.60]

(4.5)

p is the market’s percentile (2.3) of Price ÷ Replacement. B is the share of markets with p ≥ 0.90 in the same quarter, computed only when at least 30 markets report.

M = S_P + S_T + S_C + S_R

(4.6)

5. Cycle Score (C)

C = min(100, max(0, 50 + F + M))

(5.1)

Every term is in points toward defense: a positive term pushes C toward 100 (defend), a negative term toward 0 (lean in).

Term	Range (points)	Capped by
Position	−35.4 to +35.4	Formula (3.4)
Leading, supply	−16 to +16	Z^±2
Leading, occupancy turn	−8 to +8	Z^±2
Repricing S_P	−16 to +16	Z^±2
Transactions S_T	−8 to +8	Z^±2
Credit S_C	0 or +10	Indicator
Replacement S_R	0 to +45	Ramp

A missing Leading or Market Mood input contributes zero. Position requires both Standing and Momentum for the current and prior quarter.

6. The Hour (H)

θ = atan2(F̃_Δ, F̃_S)

(6.1)

H = 6 + 6·C ÷ 100 if θ ∈ (0°, 180°] · H = (6 − 6·C ÷ 100) mod 12 otherwise

(6.2)

The rising side is the one where blended Momentum is positive. C = 0 maps to 6 o’clock and C = 100 to 12 o’clock on either side.

Readings

All three readings are functions of H alone.

Reading	Hour bands
Fundamental Phase	Recovery [6, 9) · Expansion [9, 12) · Hypersupply [12, 3) · Recession [3, 6)
Market Mood	Disbelief [7, 8) · Optimism [8, 11) · Euphoria [11, 12:30) · Anxiety [12:30, 2) · Fear [2, 5) · Despair [5, 7)
Investment Posture	Aggressive [5, 8) · Active [8, 10) · Cautious [10, 11) · Defensive [11, 3:15) · Cautious [3:15, 5)

7. Constants

Constant	Value	Role
History minimum	16 quarters	Before any Z or percentile is defined
Excluded window	2020 Q2 – 2021 Q4	Omitted from every mean, deviation and percentile reference set
Cap	±2	All Leading and Market Mood Z-terms
Blend	0.6 / 0.4	Current / prior quarter, Standing and Momentum
Position scale	25	Eq. (3.4)
Strength threshold	1.5	Eq. (3.4)
Leading weights	8 · 4	Supply · occupancy turn
Mood weights	8 · 4	Repricing · Transactions
Credit step	10	Eq. (4.3)
HY percentile	0.90	National stress
SLOOS threshold	+40	National stress
Replacement maximum	45	Eq. (4.5)
Replacement ramp	0.90 → 1.00	Own-history percentile
Breadth trigger	0.60 of ≥ 30 markets	Eq. (4.5)

8. Limitations

Every historical reading uses only the data that existed at the time, so none of them was built with hindsight.

The weights and thresholds are round numbers, not values tuned to fit past outcomes. They were still chosen while looking at this history, so the results here are not a live track record.

A reading describes a market as a whole. Submarkets and individual properties can behave differently.