Penrose Backtest: The Profitable Purchase Our Model Would Have Rejected
A property can rise substantially and still fail an investment framework at the original entry price. Penrose gives us a clean example — and shows why a credible backtest must publish missed opportunities, not just successful calls.
This is a retrospective test of the AskAnnaLee property investment framework. It applies the frozen methodology to historical transactions using information available, or reasonably knowable, at the historical decision date. It does not represent a recommendation AskAnnaLee actually published in 2020. Historical backtesting does not guarantee future results.
The framework would have rejected this entry. The buyer still made about 26.1%.
An exact 646 sq ft Penrose transaction entered at approximately S$991,000. The frozen model's maximum disciplined entry was only about S$975,800.
The transaction therefore failed Gate 1. Yet the same unit later sold for about S$1.25 million after roughly 3.1 years.
REJECTION
1. Why This Case Matters
The easiest historical articles to publish are winners: identify a cheap historical purchase, show that it later appreciated, and declare the framework successful.
That is not enough.
A serious backtest also has to reveal the properties or units that the model would have rejected even though they later performed well. Penrose gives us exactly that test.
A disciplined model should protect capital — but protection has a cost. Sometimes the price of discipline is missing a winner.
2. What Buyers Knew in 2020
Penrose launched in September 2020. Contemporary research in our historical ledger records 341 sales out of 566 units over the launch weekend, with broad HDB-upgrader demand supporting take-up.
The project therefore did not suffer from obviously weak market interest. The question was not whether buyers liked Penrose. The question was whether the price of a specific unit left enough valuation protection under the frozen methodology.
Public stack records also confirm that multiple 646 sq ft units transacted around the original September 2020 launch period.
3. The Frozen Price Test
For the exact 646 sq ft observation, the frozen historical methodology produced an intrinsic-value PSF of approximately S$1,661.62.
After applying the required valuation protection, the corresponding maximum entry quantum was approximately:
AskAnnaLee Historical Maximum Entry Price: ~S$975,800
Equivalent to roughly S$1,511 psf.
The actual purchase was approximately S$991,000, or around S$1,534 psf.
| Price Test | Amount | Model Decision |
|---|---|---|
| Historical Maximum Entry | ~S$975,800 | Hard ceiling |
| Actual historical purchase | ~S$991,000 | REJECT |
| Amount above ceiling | ~S$15,200 | ~1.6% above maximum |
The difference was not huge. But a frozen rule is only meaningful if it remains frozen when the answer is inconvenient.
4. What Happened Next
The exact matched transaction subsequently sold on 24 October 2023 for approximately S$1.25 million.
The holding period was approximately 3.1 years, inside the core 3–5 year validation window.
Under the historical classification rules, this is not a PASS because the transaction never qualified at entry. It is also not a Correct Rejection because the subsequent outcome was clearly positive.
5. The Result: Opportunity-Cost Rejection
OPPORTUNITY-COST REJECTION
The transaction failed the model's historical entry rule but later produced a strong positive outcome.
This means the framework sacrificed this opportunity in exchange for maintaining its required margin of safety.
WINNER
6. Does This Mean the Model Was Wrong About Penrose?
No — and this distinction is important.
The historical model did not reject every Penrose unit. Later exact-unit research found at least five larger 936–958 sq ft Penrose units that genuinely cleared the frozen entry rule.
One 646 sq ft transaction at ~S$991,000 was above the model's ceiling and is classified as an Opportunity-Cost Rejection. Other larger Penrose units entered below their respective frozen maximums and later produced gains of roughly 43.5%–47.1% in the exact audit.
This reinforces a central AskAnnaLee principle: the project name is not the investment decision. Unit size, quantum and entry price can produce different conclusions inside the same development.
7. Why We Do Not Move the Ceiling After Seeing +26%
It would be easy to look at the eventual S$1.25 million sale and argue that the original S$991,000 purchase was obviously acceptable.
That would destroy the integrity of the test.
The frozen maximum was approximately S$975,800. The transaction exceeded it. Therefore the correct historical classification remains REJECT — even though the buyer subsequently made money.
Hindsight can explain why a missed opportunity was painful. It cannot be allowed to rewrite the original rule.
8. What We Learned
If a system never rejects a property that later rises, it is probably not applying meaningful valuation discipline.
Ignoring profitable rejected transactions would create a misleading historical success record.
This entry was only around 1.6% above the frozen maximum, yet subsequently delivered a strong gain.
Other Penrose units passed the model and also performed strongly. One project can contain both qualifying buys and rejected entries.
The correct response to an opportunity-cost rejection is to record it, not retrospectively adjust the ceiling until it becomes a PASS.
9. Methodology Note
This article uses exact same-unit transaction evidence from the historical validation ledger. Exact transaction chains rank above project averages and same-size proxies. The entry rule is frozen before the later outcome is considered.
This analysis is a retrospective application of the AskAnnaLee framework to historical evidence. It does not represent a contemporaneous 2020 recommendation. The subsequent profitable result does not change the original entry classification. Historical backtesting has limitations and cannot guarantee prospective performance.
Penrose exact transaction audit — 646 sq ft historical pair: approximately S$991,000 entry to S$1.25 million exit on 24 Oct 2023, around +26.1% over ~3.1 years.
Public Penrose stack records used to verify original launch-period 646 sq ft transactions:
Penrose unit transaction records
AskAnnaLee exact-unit audit later identified qualifying 936–958 sq ft Penrose entries with matched outcomes. Those observations are maintained separately in the historical validation ledger.
A credible track record must include the ones we would have missed.
Penrose is our first published Opportunity-Cost Rejection. It makes the backtest library stronger because it shows the framework's trade-off honestly: disciplined entry criteria can reduce downside risk, but they can also reject investments that later perform well.