Newer Condos Score Higher. Almost None of It Is Returns.

Sort every condo in Singapore by the decade it was completed and the scorecard reads like a straight line. Projects finished before 1990 average 5.55. The 1990s average 5.67, the 2000s 5.90, the 2010s 6.21, and the 2020s 6.27. Newer is better, by 0.72 points end to end.

Take the gap apart and the line stops meaning what it looks like it means. The single heaviest factor in the scorecard is historical capital appreciation, at 25% of the total. Across forty years of building it contributes minus 0.06 to the advantage newer projects hold.

Everything below is computed from the published RPS dataset — 2,357 projects, of which 2,350 carry a completion year. Multiplying the seven sub-scores by their published weights reproduces each project's published total to within 0.03 points, which is a free check that the weights and the column mapping are both right.

The ladder

The grade distribution moves further than the mean does. One project in 250 completed before 1990 earns an S; in the 2020s it is one in five. That is a real difference in what the scorecard is measuring, and it is worth knowing which of the seven factors produces it.

Where the 0.72 actually comes from

Each sub-score runs 0 to 10 and carries a fixed weight. Multiply the change in each sub-score by its weight and the gap decomposes exactly.

Three factors do 106% of the work — rental yield, project size and school catchment — and the other four either barely move or move against.

The yield gap is the largest single contributor despite carrying one of the smallest weights, because the sub-score itself moves so far: 2.65 to 5.66, more than doubling. Old freehold stock in prime districts is expensive relative to what it rents for. Newer leasehold stock in the suburbs is not.

Project size is the second contributor and the most mechanical. The median project completed before 1990 has 40 units. The median 2020s project has 140. The scorecard rewards scale because scale produces resale liquidity and maintainable facilities, so an era that built bigger scores better on it by construction.

Why performance is flat, and what that does and does not mean

The historical-performance sub-score is not a measure of raw price growth. It is a percentile rank within the project's own completion-year cohort. The scorecard says so in its own reasoning text: a 2015 project scoring 10 is described as being "in the top 3% of its 2015-era cohort", and a 1985 project scoring 8 as "top-quartile performance within its 1985-era cohort".

Ranking within cohort means the factor is designed to have no era trend. Its mean across decades is flat because it cannot be anything else. So the honest reading of the table above is not that new condos have appreciated no better than old ones — this dataset cannot answer that question — but something narrower and more useful: the scorecard's single heaviest factor deliberately contributes nothing to the vintage premium. Everything the ladder shows is coming from the other 75%.

What the cohort ranking does reveal is dispersion. The share of projects scoring 3 or below on performance runs 6.1% before 1990 and 23.8% in the 2010s. The share scoring 8 or above runs 15.6% and 27.9% over the same span. Both tails fattened; the median fell from 6.0 to 5.0. Relative to its own peers, a newer project is far more likely to be a clear winner or a clear loser than an older one is.

Two things that did not improve

MRT access is the flat line nobody expects. The sub-score averages 7.62 for pre-1990 stock and 7.49 for the 2020s — slightly worse, after four decades of network expansion. It is polarising rather than improving: the share of projects with direct walk-in access rose from 7.6% to 11.1%, while the share scoring 3 or below nearly doubled, from 2.3% to 4.2%. Twin Vew, completed 2021, sits 1.55km from Clementi. The Garden Residences, 2021, sits 1.75km from Serangoon. The older stock was built on land the network was later routed through; a meaningful share of the newer stock was not.

Future transformation also slipped, 3.50 to 3.18. Sites with announced transformation ahead of them are, on average, not where the last decade of building happened.

Control for size and tenure and most of the ladder disappears

Newer projects are bigger and more often leasehold, and both of those independently move the score. Compare each project instead against the average of its own size band and tenure type:

The 0.72-point spread falls to 0.38, and the ordering breaks at the top: on a like-for-like basis the 2010s outscore the 2020s. Roughly half the vintage premium is not vintage at all. It is the fact that recent building has been large-format and leasehold, in places where those two things score well.

What the extremes look like

The best pre-1990 project in the dataset is Ocean Park in D15 — 298 units, freehold, completed 1984, scoring 7.54 and the only S grade of its era. Lakepoint Condo (D22, 1983) and The Dairy Farm (D23, 1989) both reach 7.34.

The weakest 2020s projects are a tight cluster: One Draycott (D10, 2023, 64 units) at 3.50, Les Maisons Nassim (D10, 2023, 14 units) at 3.75, Lotus @ Jervois (D10, 2023, 36 units) at 3.84, 19 Nassim (D10, 2023, 101 units) at 4.01. All four are prime-district boutiques, which is the same profile that dragged Keppel Land to the bottom of our developer table — small, expensive, thin on yield and scale.

That is the practical read. Completion year is a decent proxy for how a project scores, but it is a proxy for size, tenure and yield, not for how the building has done. A 1985 project of 300 units near a station and a school is not competing with the era average, and a 2023 boutique of 30 units in D10 is not inheriting one. Look at the seven sub-scores; the decade tells you very little the sub-scores do not tell you better.

Method

Computed from the published TRIBE Resale Project Scorecard dataset at tribesg.com/rps-data-v2.json, 2,357 projects. Seven projects carry no completion year and are excluded, leaving 2,350. Sub-score weights are historical performance 25, schools 20, project size 16, MRT 13, rental yield 10, tenure 10, future transformation 6; applying them reproduces each published total to a mean absolute difference of 0.013 and a maximum of 0.030. "Adjusted" compares each project to the mean score of its size band (under 100, 100–299, 300–699, 700+ units) crossed with its tenure type. Freehold includes 999-year tenure. Grade and sub-score figures are as published in the dataset.