Each predictor has a mechanism written down before it was tested. The percentile compares today with all earlier months. "Past evidence" is the out-of-sample lift (1.0 means no better than the base rate) from a causal walk-forward, with the number of scored entries.
| Predictor | Percentile | Read | Past evidence |
|---|---|---|---|
| Quota change (6 months) | 41 | normal range | lift 1.76 on 577 entries; since 2017: 1.81 |
| Quota change (12 months) | 43 | normal range | lift 1.76 on 577 entries; since 2017: 1.81 |
| Bids per quota (6 months) | 72 | normal range | lift 1.70 on 42 entries; since 2017: 1.70 (too few to rely on) |
| B/A spread | 31 | normal range | lift 1.41 on 276 entries; since 2017: 1.46 |
| Premium vs its 12-month mean | 72 | normal range | not scored on its own |
| Premium momentum (6 months) | 77 | normal range | lift 1.29 on 389 entries; since 2017: 1.44 |
| Supply cycle (de-registrations 6 to 18 months ago) (12 months) | 73 | normal range | lift 1.25 on 66 entries |
| Predictor | Percentile | Read | Past evidence |
|---|---|---|---|
| Quota change (6 months) | 74 | normal range | lift 1.76 on 577 entries; since 2017: 1.81 |
| Quota change (12 months) | 75 | normal range | lift 1.76 on 577 entries; since 2017: 1.81 |
| Bids per quota (6 months) | 68 | normal range | lift 1.70 on 42 entries; since 2017: 1.70 (too few to rely on) |
| B/A spread | 31 | normal range | lift 1.41 on 276 entries; since 2017: 1.46 |
| Premium vs its 12-month mean | 64 | normal range | not scored on its own |
| Premium momentum (6 months) | 71 | normal range | lift 1.29 on 389 entries; since 2017: 1.44 |
| Supply cycle (de-registrations 6 to 18 months ago) (12 months) | 92 | tail, no supported effect | lift 1.25 on 66 entries |
How often the monthly premium was lower, rose more than 5%, and the median change, after 3, 6 and 12 months. The last ten years carry a rising trend.
| Category A | Category B | |||||
|---|---|---|---|---|---|---|
| Window | Lower | Up >5% | Median | Lower | Up >5% | Median |
| 3 months ahead, last 10 years | 37% | 45% | +3.1% | 40% | 43% | +2.6% |
| 6 months ahead, last 10 years | 33% | 55% | +6.5% | 32% | 50% | +4.7% |
| 12 months ahead, last 10 years | 32% | 66% | +15.3% | 31% | 62% | +12.0% |
| 3 months ahead, since 2002 | 44% | 40% | +2.0% | 47% | 40% | +1.2% |
| 6 months ahead, since 2002 | 43% | 48% | +3.7% | 43% | 45% | +3.3% |
| 12 months ahead, since 2002 | 47% | 49% | +4.8% | 48% | 48% | +3.0% |
A COE lasts ten years, so cars registered ten years ago are a schedule of expiries. De-registrations track that schedule closely and set the quota, and the premium has followed the supply cycle with a lag of about 18 to 24 months (a correlation of about −0.7). The schedule now stands at about 6,770 vehicles a month, 7,573 a year ahead, against 5,181 actual de-registrations over the last 12 months, so quota is set to keep rising through 2027 to 2028.
That would normally ease premium pressure, but the premium reached record highs through 2024 to 2026 while de-registrations were already rising, so demand or a longer lag is overriding it for now. This is a description of where the market sits in the cycle, not a forecast of a turning point: there are only two full cycles in the data.
URA price and rental indices (2009 Q1 = 100), quarterly to 2026Q2. Condo means private non-landed. Executive condominium prices come from URA transaction records for the last five years; the long price history is for the private index.
Same discipline as for cars: mechanisms written down first, percentile against all earlier quarters, and a causal walk-forward. A family is "supported" only if its lift is at least 1.4 over all test years and also since 2017, on at least 100 entries. No property family passes.
| Predictor | Percentile | Read | Past evidence (lift: all years, since 2017) |
|---|---|---|---|
| Price momentum (4 quarters) | 40 | normal range | 2.13 (532 entries), 0.69 |
| Rent growth (4 quarters) | 55 | normal range | 1.20 (457 entries), 0.57 |
| 2-year government yield change (4 quarters) | 45 | normal range | 0.52 (856 entries), 0.28 |
| Supply pipeline change (4 quarters) | 74 | normal range | 0.35 (229 entries), 0.28 |
| Units transacted change (4 quarters) | 23 | normal range | 0.81 (622 entries), 0.83 |
| Unsold approved units change (4 quarters) | 42 | normal range | 1.38 (101 entries), 1.38 |
| Price-to-rent vs its 5-year mean | 32 | normal range | 0.31 (38 entries), 0.31 |
| Vacancy rate vs its 5-year mean | 43 | normal range | 0.76 (158 entries), 0.17 |
| New sales / launches | 81 | tail, not supported | 1.13 (79 entries), 1.13 |
| Predictor | Percentile | Read | Past evidence (lift: all years, since 2017) |
|---|---|---|---|
| Price momentum (4 quarters) | 60 | normal range | 2.13 (532 entries), 0.69 |
| Rent growth (4 quarters) | 55 | normal range | 1.20 (457 entries), 0.57 |
| 2-year government yield change (4 quarters) | 45 | normal range | 0.52 (856 entries), 0.28 |
| Supply pipeline change (4 quarters) | 72 | normal range | 0.35 (229 entries), 0.28 |
| Units transacted change (4 quarters) | 23 | normal range | 0.81 (622 entries), 0.83 |
| Unsold approved units change (4 quarters) | 42 | normal range | 1.38 (101 entries), 1.38 |
| Price-to-rent vs its 5-year mean | 51 | normal range | 0.31 (38 entries), 0.31 |
| Vacancy rate vs its 5-year mean | 43 | normal range | 0.76 (158 entries), 0.17 |
| New sales / launches | 81 | tail, not supported | 1.13 (79 entries), 1.13 |
How often the index was lower four and eight quarters later. The last ten years were almost one-way up, so the longer window is the fairer guide to risk.
| Segment | Window | Lower after 4q | Median 4q | Lower after 8q | Median 8q | Worst 4q |
|---|---|---|---|---|---|---|
| Condo (non-landed, all regions) | since 2004 | 23% | +4.4% | 21% | +7.6% | -30% |
| last 10 years | 6% | +4.8% | 0% | +9.7% | -1% | |
| Condo, core central | since 2004 | 34% | +2.0% | 34% | +4.8% | -32% |
| last 10 years | 25% | +2.8% | 16% | +5.9% | -6% | |
| Condo, rest of central | since 2004 | 27% | +2.8% | 24% | +7.2% | -32% |
| last 10 years | 11% | +4.9% | 0% | +10.3% | -2% | |
| Condo, outside central | since 2004 | 19% | +4.6% | 13% | +10.7% | -17% |
| last 10 years | 0% | +4.9% | 0% | +11.1% | +0% | |
| Landed | since 2004 | 27% | +4.2% | 20% | +8.7% | -21% |
| last 10 years | 14% | +5.8% | 0% | +9.3% | -1% |
| Series | Peak | Trough | Fall | Quarters down | Quarters to recover |
|---|---|---|---|---|---|
| Condo (non-landed, all regions) | 1981Q2 | 1986Q2 | -47% | 20 | 20 |
| Condo (non-landed, all regions) | 1996Q2 | 1998Q4 | -42% | 10 | 36 |
| Condo (non-landed, all regions) | 2008Q2 | 2009Q2 | -26% | 4 | 4 |
| Condo (non-landed, all regions) | 2013Q3 | 2017Q2 | -10% | 15 | 9 |
| Landed | 1984Q1 | 1986Q2 | -34% | 9 | 13 |
| Landed | 1996Q2 | 1998Q4 | -48% | 10 | 47 |
| Landed | 2013Q3 | 2017Q2 | -16% | 15 | 15 |
| Vacant private homes (share of available) | 6.4% (percentile 37 of history) |
| Unsold approved units | 15,057 (percentile 4) |
| Non-landed supply pipeline | 51,677 units |
| Units sold / launched, uncompleted | 1.19 |
| Private transactions this quarter | 6,148 (+20% year on year; 2,141 new sale, 3,813 resale) |
| EC stock available / vacant | 40,302 / 1,686 (4.2% vacant) |
| EC supply pipeline | 6,541 units (3,916 under construction) |
| EC launched and sold (uncompleted, cumulative position) | 3,035 sold, 149 launched but unsold |
| EC new-sale units sold this quarter | 169 |
Median price per square foot of single-unit private sales, by month. These are medians of what happened to transact, so they move with the mix of projects; new-sale medians are lumpy because they depend on which projects launch. Not a quality-adjusted index.
| Median, single-unit sales | Month | Units | Price | Per sq ft | Per sq ft, year on year |
|---|---|---|---|---|---|
| EC new sale | Sep 2026 | 25 | S$1,731,000 | S$1,855 | +5.4% |
| EC resale | Sep 2026 | 130 | S$1,595,944 | S$1,497 | +6.7% |
| Condo new sale | Sep 2026 | 271 | S$2,738,897 | S$2,519 | -1.3% |
| Condo resale | Sep 2026 | 988 | S$1,744,379 | S$1,791 | -0.1% |
Every registered HDB resale since March 2012 (HDB, via data.gov.sg). Type a block number and street, or a town. Prices depend on the unit's storey, size, renovation and remaining lease, so a block's recent sales are a guide to its range, not a valuation.
Worked example of the method on one block: timing the sale of a five-room flat at BLK 184 Jelebu Rd.
Two checks. A forward ledger, scored as outcomes arrive, and a causal replay of the base rates over the past (each forecast uses only windows already finished at the time).
| Ledger started | 2026-10-16 |
| Base-rate forecasts logged | 18 (scored 0) |
| Condition entries logged | 0 (horizons scored 0) |
The first live outcomes arrive once the data for December 2026 are published, a few weeks after that month. Each month (COE) or quarter (property) the dashboard logs what its base rates imply before the outcome is known; every time a supported condition becomes active it is logged as an entry. Scored items appear below.
No scored outcomes yet. They will appear here as the horizons pass.
| Series | Anchor | Horizon | Higher: ten-year base rate | Higher: all history |
|---|---|---|---|---|
| COE A | 2026-09 | 3 months | 62% | 56% |
| COE A | 2026-09 | 6 months | 66% | 57% |
| COE A | 2026-09 | 12 months | 63% | 53% |
| COE B | 2026-09 | 3 months | 59% | 53% |
| COE B | 2026-09 | 6 months | 66% | 57% |
| COE B | 2026-09 | 12 months | 64% | 52% |
| Condo (non-landed) | 2026Q2 | 4 quarters | 83% | 66% |
| Condo (non-landed) | 2026Q2 | 8 quarters | 83% | 69% |
| Condo, core central | 2026Q2 | 4 quarters | 66% | 66% |
| Condo, core central | 2026Q2 | 8 quarters | 71% | 66% |
| Condo, outside central | 2026Q2 | 4 quarters | 85% | 80% |
| Condo, outside central | 2026Q2 | 8 quarters | 83% | 87% |
| Condo, rest of central | 2026Q2 | 4 quarters | 78% | 73% |
| Condo, rest of central | 2026Q2 | 8 quarters | 83% | 76% |
| HDB resale index | 2026Q2 | 4 quarters | 66% | 68% |
| HDB resale index | 2026Q2 | 8 quarters | 61% | 67% |
| Landed | 2026Q2 | 4 quarters | 76% | 68% |
| Landed | 2026Q2 | 8 quarters | 80% | 71% |
The forecast for each anchor is the share of the last ten years’ windows (and, separately, of all earlier windows) in which the series was higher after the horizon. A Brier score of 0.250 is what a coin flip scores; lower is better. The ten-year base rate beat the coin in 2 of 18 series and horizons, the all-history base rate in 10. For COE the mean score is 0.261 (ten-year) and 0.254 (all history): about a coin flip. For the property series it is 0.285 and 0.234: the long history is the better guide, because a trailing decade over-reacts to the regime just before it.
| Series | Horizon | Windows scored | Actually higher | Brier, ten-year base rate | Brier, all-history base rate | Brier, always “higher” | Direction right |
|---|---|---|---|---|---|---|---|
| COE A | 3 months | 170 | 61% | 0.251 | 0.248 | 0.394 | 57% |
| COE A | 6 months | 167 | 58% | 0.261 | 0.251 | 0.419 | 46% |
| COE A | 12 months | 161 | 55% | 0.278 | 0.265 | 0.447 | 24% |
| COE B | 3 months | 170 | 57% | 0.256 | 0.251 | 0.429 | 52% |
| COE B | 6 months | 167 | 57% | 0.260 | 0.250 | 0.425 | 44% |
| COE B | 12 months | 161 | 52% | 0.261 | 0.260 | 0.478 | 34% |
| Condo (non-landed) | 4 quarters | 162 | 70% | 0.290 | 0.229 | 0.302 | 59% |
| Condo (non-landed) | 8 quarters | 158 | 73% | 0.281 | 0.224 | 0.266 | 65% |
| Landed | 4 quarters | 162 | 67% | 0.281 | 0.232 | 0.327 | 59% |
| Landed | 8 quarters | 158 | 72% | 0.285 | 0.220 | 0.278 | 59% |
| Condo, core central | 4 quarters | 46 | 59% | 0.325 | 0.280 | 0.413 | 28% |
| Condo, core central | 8 quarters | 42 | 69% | 0.294 | 0.251 | 0.310 | 36% |
| Condo, rest of central | 4 quarters | 46 | 70% | 0.281 | 0.239 | 0.304 | 70% |
| Condo, rest of central | 8 quarters | 42 | 81% | 0.252 | 0.208 | 0.190 | 81% |
| Condo, outside central | 4 quarters | 46 | 76% | 0.243 | 0.213 | 0.239 | 76% |
| Condo, outside central | 8 quarters | 42 | 81% | 0.235 | 0.210 | 0.190 | 81% |
| HDB resale index | 4 quarters | 102 | 66% | 0.319 | 0.244 | 0.343 | 37% |
| HDB resale index | 8 quarters | 98 | 64% | 0.333 | 0.257 | 0.357 | 34% |