What is an H-Value?

How it is calculated

Every H-Value is a weighted median of comparable sales. Four steps, in order:

1. Find the comparable sales. Sales the board recorded in the last 12 months, searched outward in three rings: other units in the same building first, then the same neighbourhood, then a widening radius. A sale in the same condominium is comparable in a way nothing else is — same building, same era, same amenities, same land — so it outranks every neighbourhood comp.
2. Score each one for similarity. Bedrooms, bathrooms, parking, floor area, distance and the age of the building, each weighted by how much it moves a price. Floor area and bedroom count carry the most; parking the least. The shape of the house counts on a freehold home, where it describes the land, and is ignored on a condominium, where the board uses the same field for the inside of a unit.
3. Adjust each sale forward to today. A sale eight months ago in a moving market is not today’s price, so each comp is carried forward on the same quarterly index the market charts are drawn from. Without this an estimate silently lags the market by the average age of its evidence.
4. Take the weighted median. The value with half the weight either side — a median rather than an average, because one renovated outlier moves an average and should not move an estimate.

The range beneath the figure is the spread of those adjusted comps, so it widens exactly where the evidence disagrees with itself.

When we refuse to give you a number

A portal that always has an estimate is a portal that sometimes invents one. There are three cases where this one does not:

• Fewer than 5 comparable sales. Below that the comps are anecdotes. A unique house on a large lot in a thin market gets no H-Value, and that is the truthful answer rather than a failure.
• A property type we have never measured on at least 30 sales. If we cannot say how wrong we usually are, we do not put a number on the page.
• A property type where the measured error is above 8%. Those get a band instead of a figure — see the next section.

How accurate it is

We test the estimate by running it against sales that have already closed and comparing what it said to what the home actually sold for. Measured against 803 recent sales on September 4, 2026:

“Typically within” is the median error: half our estimates were closer than that, half further away. It is the honest summary figure — an average would be dragged around by a handful of homes we got badly wrong.

Why detached and condo apartment show a range

These are the property types where a single figure would be false precision. On detached homes we are typically out by 10.5%, and only 49% of our estimates land within 10% of the real price. So the page shows the band that actually holds four homes in five — ±22% — rather than a number that looks authoritative and is wrong half the time.

The limit is the data, not the arithmetic. A condominium’s price turns on floor, view and exact square footage, and the board publishes floor area as a band spanning several hundred square feet. A detached house turns on frontage and condition, and the board records lot size on fewer than half of all sales. No amount of cleverness recovers a measurement nobody took.

On att/row/townhouse, semi-detached, condo townhouse the error is small enough that a figure is worth naming, and those listings show one.

What it cannot know

The estimate has never been inside the house. It is built from the fields the board records, and it is blind to everything else:

• Condition and finish. A gut renovation and a kitchen from 1978 look identical in the data.
• View, exposure and floor. Two condominiums with the same layout in the same tower can differ by six figures.
• The lot itself. A ravine backing, a corner, a laneway, an easement.
• Why it sold for that. A sale between family, a power of sale, a closing date somebody badly needed — all of them recorded at face value.

This is why a large gap between an H-Value and an asking price usually means the comparable sales do not describe this home, rather than that the price is wrong. Where the two diverge by more than 10%, the listing page says so plainly instead of presenting the estimate as a verdict.

How it differs from other numbers on a home

Assessed value (MPAC)

What the province values the property at for tax purposes, on a valuation date that can be years old. It is not an attempt to estimate today’s market price and routinely differs from one by a wide margin.

Appraised value

A licensed appraiser’s opinion, formed after visiting the property, for a lender or a court. An H-Value is not this and is not a substitute for it — most importantly, nobody will lend against it.

Asking price

What a seller and their brokerage decided to list at, which is a strategy as much as a valuation. In this market it is sometimes set deliberately below expected value to attract competing offers.

H-Value

What comparable homes nearby actually sold for, adjusted to today, with the error published. A starting point for a conversation with somebody who has seen the house.

No model, on purpose

There is no machine-learning model behind an H-Value, and there will not be. Two reasons, and the second is the one that decided it.

The first is that every number here can be shown to you. The listing page lists the sales the estimate was built from, with their dates and their adjusted prices, so you can disagree with the arithmetic rather than take it on faith.

The second is the board’s agreement. It permits derivative analytical tools built on the feed, and separately prohibits putting feed data through an AI system. A weighted median of comparable sales is plainly the first and plainly not the second. A trained model would be arguable, and this is not a thing worth arguing about.