655 S Hope St

For every unit we estimate what it was worth on the lien date and set that against the County’s assessed value. Where the assessed value is the higher of the two, the annual savings are that gap × 1.18738%, ad valorem only; fixed charges that do not move with value are left out. The arithmetic for one unit is shown on that unit’s own page.

Our own leave-one-out error is 5.98% at the median and 19.36% at the 90th percentile. A discount smaller than the median error is not a signal, and we mark it “in noise”. A discount smaller than the 90th-percentile figure can still turn out to be our model rather than the property — each unit’s own page says which side of that line it falls on, in plain words.

Red does not mean you have no case. It only means that on the data we have, this unit’s market value is not below the assessed value. If you know something we cannot see — a fresh sale in the building, serious damage, the wrong square footage — send it to us and we will run it again.

What the badges mean:

• Prior sale The only evidence is this unit’s own last sale price indexed forward to today, not someone else’s sale in the building. Our own audit puts this channel’s spread at 47% for units on the same floor, against 6% for in-building comparable sales: two units on the same floor with the same footprint can come out several times apart purely because they were bought in different years.
• Recent sale The assessed value is within 5% of this unit’s own recent sale, so the roll was set from that sale. We have no case that it sits above market.
• In noise The discount is smaller than our model’s own median error.

What we publish for one unit is on that unit’s own page: the assessed value, the gap between it and our market estimate, what that gap is worth a year, how wide our error bar is on that particular unit, and which kinds of evidence produced the estimate. That page is one click from the grid above, and there is no payment step anywhere before a filing is made.