Ranking Comparables

[Edd Gillespie]

My forum assistant is not searching for some reason. It just gives me a return that says the topic I am interested in is not being dicussed today. So here's the discussion.

I have been using Wintotal for several years and I rank the comps the way Wintotal does when the sort comps radio button in the comps power window is pressed. I have been using the cavalier thinking that if I have chosen the correct comps and inserted the correct adjustments that the machine should tell me which comp is most like the subject and what the value is.

Now, I have followed the discussion on here about those instruments of the devil, the rapidly becoming ubiquitous AVM, but it became necessary the other day for me to consult USPAP on the issue of AVM. I looked at AO 18. I'll be glad to look elsewhere, but that one seems to address the USPAP pitfalls of AVM-ing.

Seems in nutshell we should:
1. Understand how the AVM works.
2. Use it properly.
3. Determine if the AVM and the data it uses are appropriate for the intended use.
4. Be able to determine whether the output is credible.
5. Determine if the output is reliable.

If sorting cmps and value of the subject in Wintotal is an AVM, I am failing USPAP-again.
DISCLAIMER: This post in no way is intended as criticism of Wintotal . A while back (like several years) I asked the tech guy on the Wintotal 800# how does this thing work. The answer given very hesitently and criptically only after checking with a senior tech is "Its a wieghted average". "Oh, I see, thankyou", I said. I got the impression its a secret, but I accepted the answer without further question. Now I have gone along about two years with the assurance that I am using a wieghted average and after all if it wasn't OK why would Wintotal use it? Now I have also been whipped into appraiser submission to never ever use the word "average" for some reason (everybody else in the world does), so I thank my lucky stars nobody has asked me why my comps were ranked the way they were in the URAR. I just wouldn't be able to honestly say anything other than the computer did it and its a weighted average. But I have no real idea or information how it did it. Now with my extremely limited newbie background in regression analysis, I'm discovering that "average" sort of sneaks in there too. This average, but don't admit it is driving me into crime.

Two questions:

1. How does Wintotal adjusted comp sorting and subject value conclusion work? (I know its a weighted average, but how does it work- what the heck is weighted?)
2. Do you think using Wintotal to rank the comps is using an AVM?

Thanks in advance for answering the questions, and if you answered to the best of your ability go ahead and tell me what you think.

 

[Austin]

Edd wrote:

Seems in nutshell we should:
1. Understand how the AVM works.
2. Use it properly.
3. Determine if the AVM and the data it uses are appropriate for the intended use.
4. Be able to determine whether the output is credible.
5. Determine if the output is reliable.

Once again, Edd sees through the fog or again like the little boy that said the emperor has no clothes.

Edd, ask yourself this question: Based on your requirements in the above quote you placed on AVM, can you honestly say the existing appraisal method used in GSE residential appraisals can satisfy the same requirements? The answer is an emphatic no!

Two questions:

1. How does Wintotal adjusted comp sorting and subject value conclusion work? (I know its a weighted average, but how does it work- what the heck is weighted?)
2. Do you think using Wintotal to rank the comps is using an AVM?

The weighted average is an old technique that gives more weight to the sale with the least amount of adjustments. They do something like say calculating that sale 1 had 10% adjustments, sale 2 had 40%, and sale 3 50% so the results are weighted using a weighted average method. The books are full of examples of doing this. No it does not work because you are weighting the average based on one sale and so you are actually giving the most weight to one random variable that could take you anywhere.
The answer to # 2 is yes ranking is an avm. What is the difference between a manual valuation model and an automate valuation model if you use the same algorithm you get the same result? The fact that it is done automatically by the computer does not change the outcome. It is the algorithm used and not the fact that it is manual or automatic. If you want to see another example in Wintotal, check out their size adjustment system.
With the just stated in mind, go back and read your first quote again and think about it.

 

[Randolph Kinney]

I don't know anything about WinTotal. I use ACI. However, after saying that, I usually pick comp 1 to present as the most similar to the subject after adjustments. I don't "weight" the numbers to arrive at market value. However, the market value will be between the highest and lowest adjusted numbers.

AVM's are mechanical and are simply programmed by whatever algorithm known to the programmer to adjust for whatever it is programmed to adjust. It does not necessarily mean the method of choosing comparables are validated nor the final number after adjustments has more meaning than some other approach be it mechanical or human.

 

[Pamela Crowley (Florida)]

I believe Alamode weights the comps according to the total gross adjustments only.

I look at that, but do NOT use it as my final reconciliation.

 

[Couch Potato]

I don't know why companies are so reluctant to give a straight answer on such "weighted average" calculations. It should be plainly published in their manuals and help files. With a little mathematical back ground, reverse engineering the formula is not hard, but should not be necessary. The calculation given is one more thing at which to look, but should never be viewed as "the answer". Only if the comps chosen were the perfect ones and all adjustments were accurate to the penny and no needed adjustments missing would the number be "the answer"; an impossible situation. There is a reason there is no set formula for appraising; it is all about good judgment based on relevant dat, training and experience.