How List.my calculates a valuation estimate

No black box

This guide sets out exactly how the free valuation tool turns recorded transactions into an estimated market value. It covers every step, the formulas, and the thresholds we actually use.

The estimate is a starting reference, not a formal valuation. Nobody inspects the property, and the figure carries none of the condition and renovation detail a valuer records on site. For anything binding — a sale, a refinance, a legal filing — you need a licensed valuer.

Step 1: Pull comparable transactions

We query JPPH subsale records that match your selection:

• Exact match on state, district, mukim and area. A township or postcode search resolves to a list of areas instead.
• Exact match on property type. "Condo" maps to Condominium and Apartment; "terrace" covers single, one-and-a-half, double and two-and-a-half storey terrace houses, including low-cost.
• Transactions from the last 3 years only.
• Progressive size tolerance. We try within 30% of your built-up area first, widen to 50%, then drop the size filter — whichever first yields at least 3 comparables.
• Capped at the 50 most recent transactions. Three is the minimum; one or two puts the tool into rough-estimate mode.

Step 2: Strip the outliers

Developer bulk transfers and anomalous sales get removed in two phases.

Batch cap. Where the same date and area carry more than 5 transactions, we keep only the 5 highest-priced. Bulk sale and purchase agreements from developers often cluster on one date in one area, and they drag the median down.

Interquartile range filter on psf. We work out price per square foot for each transaction, sort them, then:

Q1  = 25th percentile
Q3  = 75th percentile
IQR = Q3 − Q1

keep if  Q1 − 1.5 × IQR  ≤  psf  ≤  Q3 + 1.5 × IQR

You can switch this off by adding ?outliers=0 to a valuation URL.

Step 3: Weight each sale by how recent it is

Newer sales tell you more about today's market, so each comparable gets an exponential decay weight:

weight = e^(−0.05 × age_months)

Add ?recency=0 to a valuation URL for a plain unweighted median instead.

Step 4: Take the weighted median psf

We use the median, never the average — a single luxury sale wrecks an average. The weighted median sorts the psf values by size, walks the cumulative weight, and picks the value where that cumulative weight crosses half the total.

psf(i) = price(i) ÷ effective_area_sqft(i)

weighted_median_psf = psf(k)  where  Σ weight(1..k)  ≥  ½ Σ weight(all)

Step 5: Multiply by your built-up area

base_price = weighted_median_psf × your_size_sqft

Why work from psf? Location and property type are already fixed by the comparable filter in step 1, so scaling by psf adjusts for your specific unit without counting either factor twice.

Step 6: Apply four adjustments

These are applied multiplicatively.

a. Tenure. Freehold commands roughly a 10% premium over leasehold. Some of that premium already sits inside the comparables, so we apply only the gap:

freehold_ratio    = freehold_comps ÷ total_comps
user_premium      = 0.10 if freehold, else 0
implicit_premium  = freehold_ratio × 0.10
tenure_mult       = 1 + (user_premium − implicit_premium)

b. Floor level (high-rise only). We split the comparables that carry floor data into a lower and an upper half, measure the psf gap between them, and extrapolate to your floor against the median floor. Capped at 2% per floor either way. Where fewer than 10 comparables carry floor data, we fall back to 0.5% per floor above or below the median.

premium_per_floor = (upper_psf ÷ lower_psf − 1) ÷ floor_span
floor_mult        = 1 + (your_floor − median_floor) × premium_per_floor

c. Market sentiment, capped at 15%. We compare the average price of the 10 most recent comparables against the 10 older ones behind them.

momentum      = (recent10_avg − older10_avg) ÷ older10_avg
sentiment     = clamp(0.5 + momentum, 0.1, 0.9)
sentiment_adj = (sentiment − 0.5) × 0.15 × price

d. Market trend, capped at 10%. An ordinary least squares regression on price against the ordinal index of the comparables, normalised to the average price.

slope     = (nΣxy − ΣxΣy) ÷ (nΣx² − (Σx)²)
trend     = slope ÷ avg_price
trend_adj = trend × 0.10 × price

Putting it together:

estimate = base_price × tenure_mult × floor_mult + sentiment_adj + trend_adj

Step 7: Build the range

We use 15% of the standard deviation of the comparable prices as a one-sigma proxy:

σ      = stdev(comparable_prices)
margin = 0.15 × σ
low    = max(0, estimate − margin)
high   = estimate + margin

Step 8: Score the confidence

Confidence starts at 0.5, then moves on four signals:

confidence = clamp(sum, 0.10, 0.95)

The labels you see on the result: 0.8 and above is High, 0.6 and above is Medium, 0.4 and above is Moderate, and below 0.4 is Low.

What the estimate does not know

We do not account for the condition of the unit or its renovation, the view, the specific tower or stack in a condominium, the shape or frontage of an individual lot, or any market shock after the JPPH data was published.

The effective_area_sqft value in our dataset is JPPH's best-available area figure — built-up area for most stratified properties, land area for landed. See how List.my chooses which size to use.

For anything binding — a sale, a refinance, a legal filing — use a licensed valuer.

Related reading

• Where our transaction data comes from
• How to check a property's market value in Malaysia
• Why we use the median, not the average