In order to sell a house you need to be able to price it.   There are three traditional ways to do this. Each method has its degree of validity, but one is much more widely used with residential real estate than the other two.

The first method is replacement cost.  This method assumes that you know the value of the lot that the house is on, that other lots are available, and that new houses can be built on them. The idea is that you take the suare footage of the house you want to price, multiply it by the current construction costs  and add in the value of the lot.  The assumption is that someone who would have to pay $100,000 for a building lot and $100,000 to build a new house would have to pay $200,000 for a house and lot, and so could be expected to be willing to pay somewhere in the neighbourhood of $200,000 for a new, already built house.  

You can adjust this price for location, or the convenience of not having to wait while the new house is built.  You can also depreciate the existing property for age, generally at 2% per year on the construction costs.  By that method we could say that if the subject property was 10 years old, and construction costs are $100,000, the 2% per year for 10 years is 20% on $100,000, or $20,000.  Lot value is $100,000, depreciated construction cost is $80,000, so the 10 year old house is worth $180,000.

This method works well with properties that are so similar as to be interchangeable, and for which there is a large, active and fluid market.  Strata warehouse space is a good example.  Residential homes don't appraise so well using this method.

The second method is the income approach.  This approach assumes that you know how much income the property will generate, and what sort of returns investors demand.  While the returns vary from property to property, the state of the economy and from place to place, they will be relatively consistent within one locale at a specific time.  If, for example, $200,000 will get you a 3% guaranteed return in a bank, and a 10% return in a fairly risky investment, the figure of 5% per year may seem acceptable for real estate.  In simple terms we could say that a property that generates somewhere around $850 per month net rent is worth about $200,000 to someone happy with 5% ($200,000 * .05/12=$833).  If the investor pays more he receives a lower rate of return; if he pays less he recieves a higher one.  If he has the choice between two identical properties generating identical rents, he'll choose the less expensive.  

In a realistic analysis of income we'd use more complex measures, but the idea is straightforward.  This approach works well for commercial properties from shoping ceners all the way down to single apartment units used for rentals.  Its sometimes useful for residential property, but only in very stable markets.

The last method of pricing properties is the comparitive or competitive approach. This method measures what active buyers are willing to pay for property, and its the most common approach for residential property.  You simply find three properties that are currently on the market and are similar to the subject property, three recent sales that are similar to the subject proeprty, and three similar properties that were unable to sell.  You draw up a chart or use  a spreadsheet and list the properties in rows in their respective groups, with subject property at the top, the active listings in the first group of three, the sales listed right under that and the unsold listings at the bottom.  

Draw columns across the rows for lot size, taxes, age, bedrooms, bathrooms and any other relevant characteristics.  You should see a lot of similarities.  If you don't you aren't using comparable properties.  You can add or subtract from sales and list prices as appropriate (if an active listing has 3 bedrooms and the subject property has 2 bedrooms, you subtract the appropriate amount from the three bedroom price to get an adjusted 2 bedroom price).  

In a perfect world the active listings will have similar prices. The sold listings will have sales prices that are a little lower than the actives, and the expired listings (anything on the market over 90 days, usually) will have higher list prices than either group.  

The obvious conclusion is that the expired listings haven't sold because of their high prices (there can be other reasons, but if you have perfect comparables then the only difference will have to be price).  You then list the subject property a little less than the competing actives and a little higher than the recent sales.  This should give you a realistic sales price.

The problem with this approach is obvious.  Sometimes the market is dropping, in which case the sales are higher than the actives.  Sometimes the comparable properties aren't very comparable.  Sometimes there are no expired listings. When this happens you have to apply some art to the science, but its pretty clear that the comparitive/competitive approach is usually the best for run of the mill residential real estate, which is why the acromym CMA is so well known.