Returns predictor · v0 beta
Projected returns by community — baseline trend plus infrastructure shocks
Every Dubai community, projected over 12 / 36 / 60 months. Returns decompose into a clamped baseline drift and per-event infrastructure shocks (Blue Line, DWC Phase 2, Etihad Rail and others) with explicit confidence bands. The closest to a transparent forward-returns model that exists for Dubai property.
90 communities · 12 with infrastructure shocks
Methodology
Projection formula. For each community, the projected total PSF return over the horizon is the sum of three terms — a mean-reverting baseline drift, per-event infrastructure shocks, and an off-plan supply-pressure drag.
Baseline + mean reversion. Drift is annualised from the community's last 8 quarters of DLD price history, clamped to [-6%, +12%]/yr. The clamped value is then blended toward a 7%/yr Dubai long-run anchor — λ = 0.1 at 12m, 0.3 at 36m, 0.5 at 60m. This stops a community currently in a hot streak from projecting a flat compounded return out to 5 years; markets revert.
Infrastructure shocks. Each event (metro, rail, airport, mega-development) has a peak uplift coefficient, a ramp-up window before opening day, a saturation window after opening, and a catchment radius. The contribution to a given community is the peak uplift × distance decay × time factor × confidence weight. Distance decay is linear from 1.0 at the event centre to 0.0 at the catchment boundary. The time factor is the average value of the shock curve over the horizon window, integrated numerically.
Supply pressure. Each community has an off-plan share (offplan_pct from stats — the % of recent transactions that are off-plan filings). City baseline is 30%. Communities with a share above baseline get a small negative drag scaled by the excess pp and the horizon length, capped at -3pp. The signal: communities with a heavy unbuilt pipeline see absorption pressure when those units complete and hit the secondary market.
Confidence band. Each projection ships with a symmetric ± band that combines a calibration-uncertainty term (large while coefficients are literature defaults) and a per-year horizon-uncertainty term. Longer horizons widen the band; communities with more shocks applied widen it further.
Z-score normalisation. The Z-score column expresses each community's projection as standard deviations above or below the cohort mean for the same horizon, so picks can be ranked relative to peers.
What v0 doesn't model yet. Service-charge trajectory, foreign-buyer demand proxies, and per-building granularity (currently community-level only). Supply pressure uses the offplan_pct proxy rather than per-project unit counts because the project-level data is too sparse on completion dates. v1 calibration plan: run scripts/calibrate/predictor-coefficients.ts against the pre-2020 DLD back-fill to replace literature defaults with Dubai-specific values.
Source files for audit. Math: src/lib/predictor.ts. Coefficients: src/lib/predictor-coefficients.ts. Infrastructure events: src/data/infrastructure-pipeline.json. Every coefficient is auditable; no values are hidden behind a logo.