Does the undervalued screen actually work? The track record.
A screen is only worth anything if its calls hold up. So we backtested it across every cooling-measure regime since 2009 — the GFC recovery, both ABSD waves, the 2013 TDSR, COVID and the 2023 round. At each point we scored projects using only the data available then (no look-ahead), flagged the cheap ones, and measured whether they went on to out-grow their fairly-priced neighbours.
62%
flagged out-grew peers (all cycles)
+2.0pp
median extra growth
4/5
cycles beating 50%
786
flags tested
By market regime
| Market regime (scored with no look-ahead) | Window | Precision | Extra growth | Flags |
|---|---|---|---|---|
| GFC recovery → first ABSD Post-GFC: IAS & interest-only loans abolished · SSD introduced (1yr); LTV 80% · SSD to 3yr; LTV 70% (2nd loan) | 2009-2011→2014 | 68% | +3.7pp | 172 |
| 2013 TDSR & ABSD downturn ABSD raised + MSR 30% (12 Jan 2013) · TDSR framework (29 Jun 2013) | 2012-2014→2017 | 65% | +1.6pp | 99 |
| 2017 easing → 2018 ABSD SSD 4→3yr; TDSR eased (11 Mar 2017) · ABSD raised; LTV −5pp (6 Jul 2018) | 2015-2017→2020 | 47% | -0.4pp | 125 |
| COVID → 2021/22 rounds ABSD raised; LTV −5pp (6 Jul 2018) · ABSD raised; TDSR 60→55% (16 Dec 2021) · TDSR rate floor +0.5pp; 15-mo wait (30 Sep 2022) | 2018-2020→2023 | 62% | +1.7pp | 185 |
| 2023 ABSD ABSD raised; TDSR 60→55% (16 Dec 2021) · TDSR rate floor +0.5pp; 15-mo wait (30 Sep 2022) · ABSD raised sharply (27 Apr 2023) | 2021-2023→2026 | 64% | +2.0pp | 205 |
The honest caveat: the one cycle it didn’t beat a coin-flip was 2015–2020 — a broadly flat-to-declining market where “cheap” projects didn’t differentially recover. The screen works best in normal or recovering markets; in a broad downturn, cheap can stay cheap. This is evidence the method has worked, not a promise about any single project.
The cooling-measure timeline
The regimes above straddle Singapore’s major property measures (dates from MAS / MND / IRAS media releases):
2009-09
Post-GFC: IAS & interest-only loans abolished
2010-02
SSD introduced (1yr); LTV 80%
2010-08
SSD to 3yr; LTV 70% (2nd loan)
2011-01
SSD to 4yr, up to 16%; LTV 60%
2011-12
ABSD introduced (8 Dec 2011)
2013-01
ABSD raised + MSR 30% (12 Jan 2013)
2013-06
TDSR framework (29 Jun 2013)
2017-03
SSD 4→3yr; TDSR eased (11 Mar 2017)
2018-07
ABSD raised; LTV −5pp (6 Jul 2018)
2021-12
ABSD raised; TDSR 60→55% (16 Dec 2021)
2022-09
TDSR rate floor +0.5pp; 15-mo wait (30 Sep 2022)
2023-04
ABSD raised sharply (27 Apr 2023)
Live tracking
Beyond the historical backtest, every nightly run logs the current flag list, so the screen’s calls can be followed forward in real time from here on:
| Run date | Flags | Scored | Model R² |
|---|---|---|---|
| 2026-07-11 | 95 | 528 | 0.8743 |
| 2026-07-11 | 95 | 528 | 0.8743 |
Method: at each as-of point we re-run the whole neighbourhood model on data up to that date only, flag projects trading below their nearest same-era peers, then compare their subsequent $psf growth to fairly-priced neighbours within ~2 km. Built on the undervalued screen. Not financial advice; past performance of the method is not a guarantee of future results.