We Backtested PrimeKey Across 718 Launches. Here’s What Happened.

When Jackie and I first built PrimeKey Analysis, the idea was never to create another property “prediction” model.

We were trying to solve a much more practical problem.

How do you reduce the probability of buying the wrong property?

There are always projects that look exciting during launch weekend. Beautiful showflats, persuasive narratives, limited-edition discounts, “first-mover advantage”, freehold versus leasehold arguments, transformation stories — you know the drill.

But when you strip all that away, a property still has to survive the next ten years.

• Will people still want to live there?

• Will tenants want to rent there?

• Will there be enough buyers when you eventually want to sell?

• Will banks have enough recent transactions to support valuations?

• Will the neighbourhood have new demand coming in?

• Will the lease still be healthy?

That was the thinking behind PrimeKey Analysis.

Our white paper describes PrimeKey as a research and shortlisting tool, designed to identify “safe, high-probability assets rather than high-risk speculation”.

The philosophy is deliberately defensive: reduce the ways a property can fail first, then allow the upside to take care of itself.

But there was one uncomfortable question we eventually had to answer.

Does the scorecard actually work?

So we decided to backtest ourselves.

718 launches. Ten years each.

For the backtest shown in these graphics, we reconstructed the PrimeKey fundamentals of 718 Singapore new-launch projects, based on what would have been knowable around their respective entry periods.

We then tracked each project to its ten-year mark and measured the change in its project median transacted PSF.

This is important.

We weren't asking whether somebody bought the best-facing unit, negotiated an unusually low price, renovated beautifully or got lucky with timing.

We wanted to test something much simpler:

If a condominium had stronger PrimeKey fundamentals at the beginning, did the development itself tend to experience stronger price growth over the following decade?

The answer surprised even us.

The median ten-year growth formed a remarkably clean ladder:

Excellent: +40.7%

Good: +30.8%

Fair: +22.5%

Poor: +6.8%

Grade by grade.

No reversal.

And when we looked at performance relative to projects entering during the same market periods, the broad ordering remained intact.

That second point matters because someone could reasonably say:

“Of course a condo launched in 2009 would perform differently from one launched in 2013. You’re just measuring the property cycle.”

Correct.

Which is why looking only at nominal returns isn't enough.

The reconstructed backtest also examines entry-year-relative performance, and the refreshed analysis continues to show a positive relationship between PrimeKey scores and ten-year new-launch outcomes.

At the same time, we should be precise: the very top Excellent-versus-Good comparison still has less statistical power than the broader Good–Fair–Poor ladder, so we should not pretend the model has proven that every Excellent project will beat every Good one.

That distinction is important.

The interesting result isn't that PrimeKey can predict winners.

It is that progressively stronger fundamentals were associated with progressively stronger long-term outcomes.

And I think there is a logical reason why.

PrimeKey isn't really trying to predict price appreciation

This may sound strange coming from the person who created it.

PrimeKey doesn't contain a “future price” variable.

There is no machine-learning algorithm forecasting that Condo A will appreciate 37.2% while Condo B appreciates 19.8%.

Instead, PrimeKey asks 8 much more boring questions.

• How close is the MRT?

• Is the property inside or near a meaningful growth area?

• What is the future land pipeline?

• How large and liquid is the development?

• How much tenure remains?

• What school demand exists nearby?

• How deep is the future HDB upgrader pool?

• And what does rental demand look like?

That's basically it.

The framework intentionally excludes historical profit-and-loss ratios from its scoring because previous gains are heavily dependent on entry price and on conditions that may no longer exist.

In other words:

We aren't buying yesterday's performance. We're trying to understand tomorrow's demand.

And when you look at the eight pillars that way, the backtest becomes less mysterious.

Strong properties simply have fewer ways to fail

Take MRT connectivity. A condominium near a train station doesn't appreciate because the MRT magically makes prices rise every year.

It helps because the property remains relevant to more people. More owner-occupiers.

More tenants. More future buyers.

Knight Frank research referenced in our white paper found a meaningful premium for homes close to MRT stations, while also observing price effects associated with new station announcements.

PrimeKey is essentially measuring the size and durability of that audience.

The same logic applies to Project Size.

Large developments tend to transact more frequently. More transactions mean more comparable sales. More comparables help banks establish valuations. More listings create market depth, while extensive facilities and economies of scale can strengthen buyer appeal.

The white paper references ERA Research and EdgeProp findings highlighting the performance and liquidity gap between larger developments and boutique projects.

This doesn't mean small developments are bad.

It means liquidity is an investment characteristic.

And liquidity becomes very important when you are the person trying to exit.

Demand doesn't come from one place

Another reason I believe the ladder held is that PrimeKey deliberately looks for multiple independent sources of demand.

Consider schools.

NUS Business School research cited in our white paper found sizeable housing premiums around popular primary schools, driven by families competing for proximity.

That is one demand pool.

Then you have the MOP cluster — HDB owners becoming financially and legally able to upgrade.

Another demand pool.

Then rental demand.

Another.

Then connectivity.

Another.

Then infrastructure transformation.

Another. A project that performs strongly across several of these does not need every narrative to come true.

That's the defensive aspect of PrimeKey.

If one driver disappoints, another may still support the asset.

Compare that with a property where the entire investment thesis is:

“It is freehold.”

Or:

“There is an MRT coming in 12 years.”

Or:

“The developer is famous.”

Or:

“The showflat is crowded.”

Those may all be valid considerations.

But a single strong argument is still a single point of failure.

Even GLS is not simply an “oversupply” story

Government Land Sales is probably one of the most misunderstood PrimeKey pillars.

Many buyers instinctively think:

More land nearby = more supply = bad.

But that isn't always how developing precincts work.

New GLS sites also create future price anchors.

Developers purchase land at increasingly current prices, incur current construction costs and subsequently launch new products at prices needed to maintain their margins.

Earlier developments can then be repriced against these newer benchmarks.

Huttons research referenced in the PrimeKey white paper documents examples where successive site releases and launches helped lift surrounding resale benchmarks.

Again, PrimeKey isn't saying supply is automatically good.

The pillar has a shelf life.

Once the land has been awarded, built and occupied, it eventually changes from future pipeline into competing stock. The white paper explicitly recognises this.

That's why PrimeKey scores are dated assessments, not permanent labels.

Tenure protects the exit, not just the ownership period

Remaining tenure is another pillar where people often reduce the discussion to:

“Freehold good. Leasehold bad.”

The reality is more nuanced.

A healthy lease does two jobs.

First, it protects against accelerating lease decay.

Second, it preserves financing access for the person who buys from you next.

The NUS Institute of Real Estate Studies research referenced in our framework discusses the premium associated with freehold properties and how the leasehold discount changes as tenure declines.

This matters because property investing is ultimately an exit business.

You don't realise your paper gain by admiring your valuation report.

You realise it when another buyer can — and wants to — purchase the property from you.

PrimeKey repeatedly tries to answer that question:

Who is standing behind me when I eventually want to sell?

Why I think the ten-year test was especially revealing

Property fundamentals often need time.

A transformation area may take years to develop.

HDB owners need time to reach MOP.

Land parcels need time to be sold and repriced.

Schools create recurring demand across repeated Primary One cohorts.

Transaction liquidity compounds gradually as developments establish resale histories.

This may explain why the ten-year result looks cleaner than shorter-horizon tests.

Our separate five-year study was notably noisier.

Once market-entry effects were adjusted for, there was still evidence of a useful ranking signal, particularly across the broader grade ladder, but the raw five-year data was much less pronounced with insufficient transaction volumes and a relatively short passage of time in the real estate investment context.

The internal conclusion was that PrimeKey's evidence is stronger for long-horizon new-launch selection than for short-term nominal performance.

Which actually makes sense.

PrimeKey was never designed as a flipping model.

It is a fundamentals model.

But there is one thing PrimeKey cannot tell you

This is probably the most important paragraph in this article.

PrimeKey does not know what you paid.

The white paper states this very clearly:

“Price is not an input.”

PrimeKey can tell you that a development has strong fundamentals.

It cannot tell you whether paying $2,800 psf for it is sensible.

It also doesn't know whether you're buying the best stack, a terrible layout, a west-facing unit beside the bin centre, or an unusually overpriced penthouse with a huge roof terrace.

Those are separate decisions that our agents have to study and advise you on.

Our Sky Habitat versus Sky Vue case study illustrates this perfectly.

The two developments were extremely similar on the eight fundamentals PrimeKey measures, yet their outcomes diverged substantially because one entered at a meaningfully higher price.

As we wrote in that case study:

PrimeKey tells you which building. It does not tell you what to pay.

That is not a weakness we should hide.

It is the boundary of the tool.

So what did the backtest really prove?

Not that an Excellent grade guarantees 40.7%.

It doesn't.

Not that a Poor property cannot make money. Even a poor fundamentals property at a huge discount can still have the potential to bring profits.

In a strong Singapore property market, even weaker assets can appreciate.

And certainly not that PrimeKey replaces valuation, negotiation or professional judgement.

What the results suggest is something more useful:

When a new launch scored progressively better across the fundamental drivers PrimeKey measures, its median ten-year outcome also tended to be progressively better.

Excellent above Good.

Good above Fair.

Fair above Poor.

For a scorecard originally designed around a simple defensive principle — buy assets that are difficult to fail — that is an encouraging result.

Perhaps the best way to describe PrimeKey isn't as a system for finding the property that will make you the most money.

I would describe it this way:

PrimeKey helps remove properties that require too many things to go right.

And once you think about investing like that, the backtest results become surprisingly intuitive.

Great investing is not always about predicting the biggest winner.

Sometimes it is simply about systematically avoiding the obvious losers.

That was what PrimeKey was built to do.

And after putting our own scorecard through ten years of market history, the evidence so far suggests that the fundamentals mattered far more than we expected.

The ladder held.

Now the more interesting question is what it tells us about the properties being launched today.

Stuart Chng is the Managing Partner of Navis and Chief Agency District Director at Huttons and the co-creator of Navis Atlas and PrimeKey Analysis.

He adores music and can play a few instruments decently without upsetting his neighbours. When not doing so, he enjoys pillow fighting with his son and coming up with silly puns which barely amuses his wife. 

Professionally, he is a licensed real estate agent, avid investor in options, stocks and real estate, team leader, speaker and columnist for several property newsletters and blogs and is often quoted in media interviews on 938FM, Channel 8, PropertyReport, PropertyGuru and other publications.

Throughout his career, he has helped many clients grow their wealth through selecting great real estate investments and managing their portfolios actively. Read his clients' reviews here.

Stuart has also coached many top million dollar producing agents from top Singapore real estate agencies. Read his agents' reviews here.