Methodology

Everything below is open. If you spot an error, email me — I correct within 48 hours.

1. Data sources

• MOE Phase 2C ballot results — published annually each August. Vacancies, applicants, admitted, and the "indicator" (which distance/citizenship bucket had to ballot for the last seats).
• SC Schooling — aggregator of the above; I cite it where I rely on its derived figures, but every number traces back to MOE.
• OneMap (data.gov.sg) — official government API for postal-code → coordinates → distance to schools.
• SingStat — births by planning area; informs the cohort projection layer.
• URA + HDB — planning-area boundaries and BTO completion schedules; informs medium-term demand projection.

2. The MOE phase priority order

P1 admission walks through phases in order:

1. Phase 1 — sibling currently in the school (Singapore Citizens).
2. Phase 2A1 — sibling currently in school or parent alumnus (members of school's alumni association).
3. Phase 2A2 — parent is school staff, or MOE staff.
4. Phase 2B — parent volunteer (PVP 40+ hours), clan/religious affiliation, or active community-leader status.
5. Phase 2C — everyone else, with this internal priority order when balloting is needed:
   1. SC <1km
   2. SC 1–2km
   3. SC >2km
   4. PR <1km
   5. PR 1–2km
   6. PR >2km
6. Phase 2C Supplementary — leftover Phase 2C; foreigners participate here.
7. Phase 3 — non-citizen / non-PR foreigners.

3. How chances are computed

For each school × your profile:

1. Filter for gender compatibility (single-gender schools).
2. If you qualify for Phase 1 / 2A1 / 2A2 / 2B → near-certain admission, high chance.
3. Otherwise, fall through to Phase 2C. Compute your bucket (SC or PR × distance band).
4. Compare your bucket to the most recent year's "indicator". If you rank above the indicator → walk in. Below → shut out. Equal → ballot (50% baseline, refined by MOE-published bucket-specific ballot chance where available).

4. Projection model (v1)

For each school, I fit a simple OLS regression on (year, subscription %) for the historical years available, then project forward. I report:

• Projected subscription % for your target year, with a ±band based on YoY volatility.
• Projected indicator category (open / PR-balloted / SC-balloted), using the mode of the last 3 historical categories with a recency tiebreak.
• Confidence (high / medium / low) based on data points available and volatility.

This is intentionally simple. Future versions will layer in birth-cohort and BTO data.

5. What I don't model (yet)

• School quality / sentiment / reviews — too subjective and too prone to complaints.
• Property prices — important but not in v1.
• Specific MOE policy shocks — flagged in commentary, not in the model.

6. Refresh cadence

MOE data refreshed annually in August. Projection model re-run on every data refresh. Last update: 2025.