• Welcome to AppraisersForum.com, the premier online  community for the discussion of real estate appraisal. Register a free account to be able to post and unlock additional forums and features.

The New USPAP

Status
Not open for further replies.
The Mars dude arguing with Chat GPT :p
ChatGPT makes mistakes, - but if you argue with it, it can eventually correct itself in most cases. Sometimes I wonder whether it is worth the effort --- but it is much superior to Google Search (search.google.com). It is my official tutor and memory aid.
 
ChatGPT makes mistakes, - but if you argue with it, it can eventually correct itself in most cases. Sometimes I wonder whether it is worth the effort --- but it is much superior to Google Search (search.google.com). It is my official tutor and memory aid.
ChatGPT does make mistakes because it searches info in the internet whether it's accurate or not.
What I like about it is that it saves me time when I need answers.
After awhile, ChatGPT can only go so far. I find youtube videos to be more informative in my research on my next big vacation.
 
I've read just about every post on this thread, and it's interesting to see those who praise USPAP versus those who loathe it.

I will say something that complicates the subject. I, just like DW, know instances when Tennessee has declared something to be a violation of USPAP that really wasn't. I recall a prior chairman of the Tennessee Commission who literally lost his license because of a USPAP complaint for a minute. I knew him well before he passed a couple of years back.

It really doesn't matter how much we proclaim or decry, if the state can't get it right, even those who have complied might be found in non-compliance. That doesn't speak well of the way USPAP is used in the governing part of this equation.

Surely, we can all agree on that.
 
ChatGPT does make mistakes because it searches info in the internet whether it's accurate or not.
What I like about it is that it saves me time when I need answers.
After awhile, ChatGPT can only go so far. I find youtube videos to be more informative in my research on my next big vacation.

If you want to do R, then ChatGPT can really help. And what I say probably applies to R, JavaScript, HTML, etc., programming more than anything else. Last night I was up to 4:30am before i finally got an HTML web pages showing the results of Cluster Analysis of Pacifica to display popups the way I wanted, In particular I had to go around and around with it to get hovering the mouse over the legend of 15 clusters buttons to pop up the description of each cluster the way wanted. It's kind of like going around in circles making two steps forward and one backward, again and again. ChatGPT will forget recommendations it has made that you tell it work and step back to a previous suggestion it made that didn't work. So, it is kind of cumbersome. But at 4:30am I finally got what I wanted. (BTW its been 7 years since I did any JavaScript --- so I do get rusty, thanks to ChatGPT I can pick up these old skills faster and move forward.)
 
If you want to do R, then ChatGPT can really help. And what I say probably applies to R, JavaScript, HTML, etc., programming more than anything else. Last night I was up to 4:30am before i finally got an HTML web pages showing the results of Cluster Analysis of Pacifica to display popups the way I wanted, In particular I had to go around and around with it to get hovering the mouse over the legend of 15 clusters buttons to pop up the description of each cluster the way wanted. It's kind of like going around in circles making two steps forward and one backward, again and again. ChatGPT will forget recommendations it has made that you tell it work and step back to a previous suggestion it made that didn't work. So, it is kind of cumbersome. But at 4:30am I finally got what I wanted. (BTW its been 7 years since I did any JavaScript --- so I do get rusty, thanks to ChatGPT I can pick up these old skills faster and move forward.)
Pacifica has many different neighborhoods. I hope your clusters are in same areas and similar style homes and take into account when market conditions change.
An appraiser can see it more than computer.
 
Pacifica has many different neighborhoods. I hope your clusters are in same areas and similar style homes and take into account when market conditions change.
An appraiser can see it more than computer.
I'll probably put it up on my website and put a link on this forum. It is an interactive HTML page, so going to the legend you can click, double click or just hover over one of the differently colored cluster icons in the legend, to either:

1. Display only the properties belonging to the clicked cluster icon by removing all others
2. Add the properties removed back on to the graph
3. Remove the specific cluster from the graph
4. Put that removed cluster back on the graph
5. Show a description of the characteristics of a particular cluster.

Yes there is good correlation between the clusters and geographical areas. But, as might be expected, it is not perfect - because massive updating can alter the original characteristics of a home, such as doubling the GLA. Some neighborhoods have a very high corrleation to specific colors, i.e. clusters - and others are somewhat mixed.

One issue is that I have to round the longitude and latitude to 3 decimal places to prevent overfitting - that is about the size of one block. And MARS is instructed to make changes to the model only if it has noted a significant change over about 5 blocks. In any case the data is based on a block area and there may be several properties in that area, yet there can be only one displayed property at a time for that block, understanding there are 1884 properties to be displayed on a single page, represented all SFR sales from 9/2017-9/2023. Also overlaps are prevented, so one property may cover another. However ggplotly() is putting the most prevalent cluster on top. You can look under the upper layer by removing that particular cluster. In most cases there is nothing underneath - but often enough there maybe one or two layers underneath. If you over the mouse over a property - you will see the address and a few important details like GLA and Lot Size.

The concept is to provide a kind of AreaVibes (areavibes.com) for neighborhoods - for appraisers and real estate agents. That is just one facet of the desired end result.
 
Last edited:
some wording to help appraisers lessen violations since some violations are worse then others.
Well, sin is sin...once on your record, not many lawyers will give you a bye in court. They will accuse you of being ill-trained, or simply a poor appraiser. That's a major problem. There is no statute of limitations for your appraisal record.

There are a lot of appraisers who have been at this game 20 and 30 years. Hatch claims few appraisers are sanctioned. Well, I'd say they are getting better at avoiding sanctions certainly. But not many years ago the director stated in a newsletter they no longer update that some 49 or so complaints were lodged and 20 or so were adjudicated with the rest dismissed. At the time there were less than 900 appraisers. So 40 complaints out of 800 some appraisers means one complaint per appraiser per 20 years. Therefore, to escape having a complaint filed in a 20 or 30 year career is just luck. Are there repeat offenders? Sure. I know one girl who had 3 complaints filed, relatively mild issues apparently, because each time she was fined a bit and had to take a course or two, and report her work for six months, at which time, they pulled some of her assignments for review. She did what they asked each time. So that in the end, she retired without losing her license. And apparently the VA had no issue as she was a VA appraiser. But nevertheless, the cost of her E & O must have been high. And such complaints had to have taken a dent out of her income. And to what purpose did it serve?
ChatGPT does make mistakes because it searches info in the internet whether it's accurate or not.
Think in terms of this. Say, I ask AI to use the flood data from USGS stations that measure the height and flow of creeks in flood stage, then to predict the elevation of the largest flood on the record but adjust it according to changes in the watershed by estimating the time it takes water to concentrate in a particular spot, due to the additional concrete, roads, etc. and the reduction in pasture and trees in that watershed. (Called time of concentration - how quickly the water reaches the max height during a flood event. It's an engineering problem.) Say it predicts a height above the posted flood maps by one foot. You then build above that mark, not the flood map and think you are safe. So, you don't buy flood insurance. Bingo next year a record flood occurs, you are flooded out and you cannot sue the computer. You have no engineer to sue. You depended upon AI to do a task that a licensed civil engineer would be liable for...no recourse. AI will be doing risky things and humans won't be held liable for it. So, how soon to have AI appraisers... no license, no liability, no recourse, no ROVs, no reviews, nothing. You are stuck with the results.
That doesn't speak well of the way USPAP is used in the governing part of this equation.
One premise of the law that is fairly unique to the USA is the notion that one should give the benefit of the doubt to cases that are not clear. Prosecutors are loathe to file charges if the case is unlikely to be won. So, a trained jurist is making that judgement based upon their knowledge of the law. An investigator or board, OTOH, are not trained jurists. It adds an element of uncertainty to the case. The fact that there is no consequences to the board or investigator should the real courts overturn a decision, speaks volumes about the inadequacy of the process. Again, in most states, I bet some appraiser somewhere appealed a board decision and won. In my opinion, in those cases the appraiser is still not made whole. They are out tens of thousands of dollars in many cases, especially if their E & O wants to quickly settle cheap and the appraiser is forced to fend for themselves. Geoge calls these incidental cases, the unfortunate happenstance that is no fault of the board nor the appraiser but paid for entirely by the appraiser. Just bad luck. No. It is incompetence upon part of the boards who apply a heavy hand. And, like AI above, no consequences for failure of the board. A bad board can hurt a lot of people over time. The fact only a few per year get the hammer does not make it 'acceptable'. There ought to be consequences for failure - be it the appraiser or be it the board.

So I would propose that when an appraiser wins a case, that the BOARD AS INDIVIDUALS be required to pony up the legal fees of the appraiser. And they can do that with Board E & O or performance bond- I was under that as a member of 2 boards (the state and national board of a group) ...so that the licensed appraisers themselves are not paying out for board incompetence.
 
I'll probably put it up on my website and put a link on this forum. It is an interactive HTML page, so going to the legend you can click, double click or just hover over one of the differently colored cluster icons in the legend, to either:

1. Display only the properties belonging to the clicked cluster icon by removing all others
2. Add the properties removed back on to the graph
3. Remove the specific cluster from the graph
4. Put that removed cluster back on the graph
5. Show a description of the characteristics of a particular cluster.

Yes there is good correlation between the clusters and geographical areas. But, as might be expected, it is not perfect - because massive updating can alter the original characteristics of a home, such as doubling the GLA. Some neighborhoods have a very high corrleation to specific colors, i.e. clusters - and others are somewhat mixed.

One issue is that I have to round the longitude and latitude to 3 decimal places to prevent overfitting - that is about the size of one block. And MARS is instructed to make changes to the model only if it has noted a significant change over about 5 blocks. In any case the data is based on a block area and there may be several properties in that area, yet there can be only one displayed property at a time for that block, understanding there are 1884 properties to be displayed on a single page, represented all SFR sales from 9/2017-9/2023. Also overlaps are prevented, so one property may cover another. However ggplotly() is putting the most prevalent cluster on top. You can look under the upper layer by removing that particular cluster. In most cases there is nothing underneath - but often enough there maybe one or two layers underneath. If you over the mouse over a property - you will see the address and a few important details like GLA and Lot Size.

The concept is to provide a kind of AreaVibes (areavibes.com) for neighborhoods - for appraisers and real estate agents. That is just one facet of the desired end result.
I'm doing an appraisal in Linda Mar Area.
My 1004MC shows decrease in prices in recent quarter. Yet my intuition is that prices have not decreased.
Does your model show the decreased due to seasonality? Or correlation with mortgage rates?
 
My 1004MC shows decrease in prices in recent quarter. Yet my intuition is that prices have not decreased
How many data points do you have. You have posted more than once that if you feelz you don't have enough data points in the neighborhood. You expand the neighborhood parameters to make the 1004mc "look good".
 
How many data points do you have. You have posted more than once that if you feelz you don't have enough data points in the neighborhood. You expand the neighborhood parameters to make the 1004mc "look good".
I have sufficient data (at least 10 per period) and I if I use comps a year ago, subject's value would be higher than what had sold in past 3 months.
There still multiple offers with few homes for sale.
Remember that after Super Bowl, market picks up (if not, bad market coming). That has been my feelz for decades.
 
Status
Not open for further replies.
Find a Real Estate Appraiser - Enter Zip Code

Copyright © 2000-, AppraisersForum.com, All Rights Reserved
AppraisersForum.com is proudly hosted by the folks at
AppraiserSites.com
Back
Top