How close my number comes to the sale.
Anyone can put a number on a building. The fair question is whether it holds up when the building sells. So I test mine against every recorded Florida apartment sale, every month, and I put the result here, misses included.
The test
How far off, sale by sale: 18,710 apartment sales
within 10% of the price
Buildings of 5 doors and up: 2,741 sales, November 2023 to October 2026. Every apartment sale of 2 doors and up (18,710): a typical miss of 29%, 20 of 100 inside my range.
What this page holds, and what it will.
Three records, kept apart. Only the first has enough behind it to show today.
- The look backLive, every month
My valuator run on every recorded sale as it would have run 90 days before, next to what the building sold for. The figures above.
- The real recordNone sold yet
None have sold yet. I started giving owners numbers in August 2026, and Florida deeds take two to four months to post. When one of those buildings sells, it lands here with the number I gave and the day I gave it, good or bad.
- My own closed salesNot shown yet
The buildings I have listed and closed, with the price and the days on market. They go here once there are enough to mean something, and not before.
How I test it
I take every Florida apartment building that sold at arm’s length in the last three years, from the county deed records. Family transfers, foreclosure-type deeds and prices too low to be a real sale are out, and a deed that covered several parcels counts once.
For each one I go back to 90 days before the sale and run my valuator the way it runs for owners, using only what was published by that day: the 10-year Treasury rate for the month before, the sales recorded before that day, and the area rent HUD published for that year.
I don’t know what rent each building actually collected, so the engine gets the area rent, the same as an owner who taps “Not sure? Use the market rent.” That is the hardest test my number faces. An owner who gives me the real rents gets a tighter number.
Then I put my number next to what the building sold for, and I count.
I run it again every month as new sales record. Every run is kept, and each one is sealed with a fingerprint of the one before it, so nothing on this page can be changed after the fact.
By size of building
| Building | Sales | Typical miss | In my range | Ran |
|---|---|---|---|---|
| 2 to 4 doors | 15,969 | 29% | 20 of 100 | 6% low |
| 5 to 9 doors | 1,240 | 35% | 24 of 100 | 25% high |
| 10 to 19 doors | 599 | 37% | 22 of 100 | 32% high |
| 20 to 49 doors | 307 | 36% | 24 of 100 | 28% high |
| 50 doors and up | 595 | 32% | 18 of 100 | 19% high |
Every county with 30 or more sales (38)
| County | Sales | Typical miss | In my range | Ran |
|---|---|---|---|---|
| Miami-Dade County | 2,893 | 28% | 18 of 100 | 14% low |
| Broward County | 2,007 | 23% | 24 of 100 | 5% low |
| Pinellas County | 1,904 | 34% | 19 of 100 | 2% high |
| Lee County | 1,893 | 25% | 21 of 100 | 12% low |
| Palm Beach County | 1,176 | 25% | 23 of 100 | 6% low |
| Duval County | 841 | 42% | 15 of 100 | 28% high |
| Polk County | 742 | 27% | 25 of 100 | 8% high |
| Hillsborough County | 679 | 36% | 20 of 100 | 27% high |
| Sarasota County | 649 | 57% | 12 of 100 | 30% low |
| Orange County | 611 | 23% | 28 of 100 | 7% high |
| Manatee County | 519 | 24% | 24 of 100 | 7% low |
| Escambia County | 472 | 46% | 17 of 100 | 27% high |
| Volusia County | 437 | 37% | 18 of 100 | 26% high |
| Marion County | 421 | 34% | 21 of 100 | 6% low |
| Brevard County | 347 | 31% | 20 of 100 | 9% low |
| Charlotte County | 345 | 36% | 22 of 100 | 27% low |
| Leon County | 312 | 28% | 25 of 100 | 14% high |
| Bay County | 243 | 38% | 21 of 100 | right on |
| St Lucie County | 204 | 31% | 22 of 100 | 19% high |
| Pasco County | 196 | 50% | 13 of 100 | 24% high |
| Alachua County | 193 | 46% | 9 of 100 | 34% high |
| Lake County | 152 | 42% | 16 of 100 | 40% high |
| Collier County | 128 | 29% | 16 of 100 | 13% low |
| Seminole County | 127 | 21% | 30 of 100 | 14% high |
| Monroe County | 123 | 54% | 7 of 100 | 52% low |
| Osceola County | 123 | 19% | 24 of 100 | 3% high |
| Citrus County | 105 | 33% | 23 of 100 | 6% low |
| Indian River County | 99 | 23% | 26 of 100 | 2% low |
| Highlands County | 96 | 25% | 24 of 100 | 10% low |
| Martin County | 92 | 49% | 15 of 100 | 46% high |
| Hernando County | 79 | 17% | 38 of 100 | 8% high |
| Santa Rosa County | 76 | 29% | 14 of 100 | 6% low |
| St Johns County | 59 | 54% | 12 of 100 | 51% low |
| Nassau County | 49 | 48% | 8 of 100 | 44% low |
| Okaloosa County | 39 | 36% | 23 of 100 | 11% low |
| Walton County | 39 | 48% | 21 of 100 | 30% low |
| Okeechobee County | 36 | 45% | 22 of 100 | 18% high |
| Desoto County | 31 | 51% | 32 of 100 | 43% high |
| Year sold | Sales | Typical miss | In my range | Ran |
|---|---|---|---|---|
| 2023 | 976 | 31% | 17 of 100 | 16% low |
| 2024 | 6,303 | 29% | 20 of 100 | 6% low |
| 2025 | 6,610 | 28% | 21 of 100 | right on |
| 2026 | 4,821 | 32% | 20 of 100 | 5% high |
The misses
The five biggest each way, and what I know about each one.
Where my number ran high
16 doors in Duval County, sold January 2025 for $155,000.
My number 90 days before the sale: $2,326,000, 1401% high.
Why: My comp audit flags this sale: the price a door is far from nearby sales; one deed covered several parcels; it sold again within a year.
Since then: My comp audit flags it and it waits in the research queue.
The address stays off this page until I check the sale myself.
56 doors in Duval County, sold August 2024 for $565,660.
My number 90 days before the sale: $7,547,000, 1234% high.
Why: My comp audit flags this sale: the price a door is far from nearby sales; one deed covered several parcels.
Since then: My comp audit flags it and it waits in the research queue.
The address stays off this page until I check the sale myself.
6 doors in Pasco County, sold October 2026 for $100,000.
My number 90 days before the sale: $1,248,000, 1148% high.
Why: My comp audit flags this sale: the price a door is far from nearby sales; one deed covered several parcels.
Since then: My comp audit flags it and it waits in the research queue.
The address stays off this page until I check the sale myself.
10 doors in Duval County, sold May 2025 for $117,300.
My number 90 days before the sale: $1,422,000, 1112% high.
Why: My comp audit flags this sale: the price a door is far from nearby sales; one deed covered several parcels; it sold again within a year.
Since then: My comp audit researched it and found nothing that explains it. It stays in the set.
The address stays off this page until I check the sale myself.
88 doors in Escambia County, sold October 2025 for $1,265,000.
My number 90 days before the sale: $14,156,000, 1019% high.
Why: My comp audit flags this sale: the price a door is far from nearby sales; one deed covered several parcels.
Since then: My comp audit flags it and it waits in the research queue.
The address stays off this page until I check the sale myself.
Where my number ran low
2 doors in Sarasota County, sold March 2025 for $98,160,600.
My number 90 days before the sale: $430,000, 100% low.
Why: My comp audit flags this sale: the price a door is far from nearby sales; the price is far from the county's assessed value; it sold again within a year.
Since then: My comp audit flags it and it waits in the research queue.
The address stays off this page until I check the sale myself.
3 doors in Sarasota County, sold December 2025 for $98,160,600.
My number 90 days before the sale: $676,000, 99% low.
Why: My comp audit flags this sale: the price a door is far from nearby sales; the square feet a door is out of range on the county roll; the price is far from the county's assessed value; it sold again within a year.
Since then: My comp audit flags it and it waits in the research queue.
The address stays off this page until I check the sale myself.
2 doors in Okaloosa County, sold November 2025 for $38,450,000.
My number 90 days before the sale: $345,000, 99% low.
Why: My comp audit flags this sale: the price a door is far from nearby sales; the square feet a door is out of range on the county roll.
Since then: My comp audit flags it and it waits in the research queue.
The address stays off this page until I check the sale myself.
2 doors in Escambia County, sold August 2025 for $27,409,695.
My number 90 days before the sale: $356,000, 99% low.
Why: My comp audit flags this sale: the price a door is far from nearby sales; the price is far from the county's assessed value; it sold again within a year.
Since then: My comp audit flags it and it waits in the research queue.
The address stays off this page until I check the sale myself.
2 doors in Miami-Dade County, sold April 2026 for $43,131,000.
My number 90 days before the sale: $600,000, 99% low.
Why: My comp audit flags this sale: the price a door is far from nearby sales; the price is far from the county's assessed value.
Since then: My comp audit flags it and it waits in the research queue.
The address stays off this page until I check the sale myself.
5,299 of the sales in this test are flagged by my comp audit as unusual (a family price, land sold as a building, a deed over several parcels). Leaving them out, the typical miss is 24%. I keep them in the headline anyway, because I don’t get to pick which sales count.
What I changed because of a miss
Nothing yet. When a miss teaches me something about the engine, the fix goes here with its date, and the numbers before and after it, tested on sales the fix never saw.
What this test can’t tell you
- My engine’s market tiers by ZIP were last set in October 2026, from sales that include many of these. So part of this test is on sales the engine has already seen. The clean test is the months ahead, and the real record.
- Rent is the biggest input, and here it’s the area’s rent, not the building’s.
- Condition, a renovation, a motivated seller or a buyer who needed this one building are in no public record, and they move prices.
- Florida counties post deeds late, some by months. A sale shows up here once it’s recorded.
Last recomputed November 1, 2026. Run tr_20261101002031. Fingerprint ba30b563badd5dd80db89c370b60a630402a794b8434a63d69036214a157037a. The numbers on this page as data: track-record.json.
