What makes a comp similar enough for an ARV

Similarity is a match score from 0 to 1. It says how close a sold house is to the subject on five facts: bedrooms, bathrooms, living area, year built, and lot size. It is a weight input. It is separate from the dollar adjustment for bedrooms and bathrooms, and it does not read the distance or the sale date. The houses on this page are invented so the arithmetic can be checked by hand. They are not closed sales, and this page does not look up a listing.

A good comparable sale for an after-repair value is one that scores high on those five slices and also sold recently and nearby. Recency and distance have their own multipliers. Those two, and the weighted median they feed, are on how to pull comps. This page owns the match score only.

Five slices, 100 points

Each slice has a fixed point budget. Bedrooms can contribute 20. Bathrooms can contribute 15. Living area can contribute 25. Year built can contribute 20. Lot size can contribute 20. The maximum is 100, and that maximum stays 100 even when a slice has nothing to score. The similarity score is the points divided by 100, rounded to two decimals. A full match is 100 / 100 = 1.00. A total of 45 points is 0.45. A total of 80 points is 0.80.

The bands are checked from the closest match outward, and a house lands in one band per slice. Points from two bands on the same slice are not added.

Bedrooms and bathrooms

Bedrooms use the absolute difference in the count. A difference of 0 scores 20. A difference of 1 scores 15. A difference of 2 scores 8. Any other difference scores 0, including a difference of 3 or more and a half bedroom. The bedroom test matches those three counts exactly. A count of 0 is still scored. Two houses that both have 0 bedrooms differ by 0 and score 20. A subject with 0 bedrooms and a comp with 3 bedrooms scores 0 on this slice. The slice is not skipped because the count is empty.

Bathrooms use the absolute difference as well. A difference of 0.5 or less scores 15, so a half-bath gap is a full bath match. A difference above 0.5 and up to 1 scores 10. A difference above 1 and up to 2 scores 5. A difference above 2 scores 0. A subject with 2 baths and a comp with 2.5 baths scores 15. The same subject against a comp with 3 baths scores 10. Baths are scored when the count is 0, on the same rule as beds.

Those points are not dollars. A one-bed gap scores 15 similarity points and can still receive a bedroom dollar adjustment in a later step of the same pipeline. The sign of that adjustment, the pair gates, and the clamps live on bedroom and bathroom comp adjustments.

Living area, year built, and lot size

Living area is scored only when both sizes are above 0. The ratio is the smaller living area divided by the larger one. A ratio of 0.90 or better scores 25. A ratio of 0.80 or better, and under 0.90, scores 20. A ratio of 0.70 or better, and under 0.80, scores 15. A ratio of 0.60 or better, and under 0.70, scores 10. A wider gap, with both sizes still above 0, scores 5. There is no zero for a large size gap when both living areas are on file. If either living area is missing or stored as 0 or below, the slice stays at 0 and the 25 remains inside the 100-point maximum.

Year built is scored only when both years are above 0. The gap is the absolute difference in years. A gap of 5 years or less scores 20, and a gap of exactly 5 still scores 20. A gap above 5 and up to 10 scores 15. A gap above 10 and up to 20 scores 10. A gap above 20 and up to 30 scores 5. A gap above 30 scores 0. A year of 0 or below on either house leaves the slice at 0, and the 20 stays in the maximum.

Lot size is scored only when both lots are above 0. The ratio is the smaller lot divided by the larger one. A ratio of 0.80 or better scores 20. A ratio of 0.60 or better, and under 0.80, scores 15. A ratio of 0.40 or better, and under 0.60, scores 10. A wider gap, with both lots still above 0, scores 5. A missing lot, or a lot stored as 0 or below, leaves the slice at 0 of 20. The maximum does not shrink to 80 to forgive the gap. The missing slice pulls the score down.

One invented subject, three scores

The subject is invented: 3 beds, 2 baths, 1,500 square feet, built in 1980, on an 8,000 square foot lot. Nothing here was fetched from a listing site.

The close comp is 3 beds, 2 baths, 1,400 square feet, built in 1976, on a 7,000 square foot lot. Beds differ by 0, so that slice is 20. Baths differ by 0, so that slice is 15. Living area is 1,400 / 1,500 = 0.933, which is 0.90 or better, so that slice is 25. The year gap is 4, which is 5 or less, so that slice is 20. The lot ratio is 7,000 / 8,000 = 0.875, which is 0.80 or better, so that slice is 20. The points are 20 + 15 + 25 + 20 + 20 = 100. Similarity is 1.00. When this score reaches the after-repair value, the weight is multiplied by 1 + 1.00, which is 2.

The farther comp is 4 beds, 3 baths, 1,000 square feet, built in 1955, on a 3,000 square foot lot. Beds differ by 1, so that slice is 15. Baths differ by 1, so that slice is 10. Living area is 1,000 / 1,500 = 0.667. That ratio is in the 0.60 band, so the slice is 10. Both sizes are above 0, so the slice stays at 10 rather than falling to 0. Year built differs by 25, which is inside the 30-year band, so that slice is 5. The lot ratio is 3,000 / 8,000 = 0.375, which is under 0.40, so the lot slice is 5, the widest lot band. The points are 15 + 10 + 10 + 5 + 5 = 45. Similarity is 0.45. The weight is multiplied by 1.45. The one-bed gap is already inside those 15 bed points, and the dollar adjustment for that same gap is still a later step.

Take the close comp again and leave the lot size off. Beds, baths, living area, and year still total 20 + 15 + 25 + 20. The lot slice is 0 of 20. The points are 80. Similarity is 0.80. The weight is multiplied by 1.80. The house did not become a different match on the four facts that are present. The empty slice lowered the score because the maximum stayed at 100.

What the score does next

The after-repair value multiplies a comp’s weight by 1 plus this score only when the score is above 0. A score of 1.00 multiplies by 2. A score of 0.45 multiplies by 1.45. A score of 0.80 multiplies by 1.80. A score of 0 leaves recency and distance as the only multipliers. Distance and the sale date are not inputs to the five slices. A sale can be a perfect physical match and still be old, or a weak physical match and still be next door. Those facts stack as separate multipliers on the comps page. This page does not rebuild that median.

A low score stays in the set. The pipeline keeps the fetched comps, sorts them by this score from highest to lowest, and still passes the full set to the after-repair value. There is no similarity cutoff that drops a comp before the median. A 0.45 comp is lighter than a 1.00 comp. It is still in the list the median reads. Once you have an after-repair value you trust enough to type into an offer, the purchase-price worksheet is the MAO calculator. The calculator does not recompute this score. It takes the value you give it.

Questions