Using a predictive model to find undervalued properties.

Detroit, Michigan
Starting Out
2yFollow

Using a predictive model to find undervalued properties.

David D.
New to Real Estate · Member since 2024 · 19 posts · 3 votes

I am curious if anyone has employed the simple strategy I was thinking of for investing in any kind of property:

1. Use something like ATTOM API to get historical sales snapshots, or just all the historical sales of the property class for a metropolitan area. 
2. Train a regression model using square footage, bathroom count, or more advanced features to predict the price of the sales. 
3. Get the error of your predicted housing price (if you're using price per square foot, convert to housing price) down to $10K-$40K or so. 
4. Find properties with characteristics that predict they should be selling for 2 standard deviations above what their actual price/value is (or basically just properties that are selling for way below what the model predicts they should sell for).

The challenge here is of course step 3. I just started experimenting with Downtown Detroit with ATTOM's "sales snapshots" and I haven't gotten that kind of performance yet. However, I'm sure many people have, especially if they have images or floor plans of the houses/other features and lots of data. 

I apologize if this is all just fairly standard financial modeling. Very key thing here, I don't have a finance/economics background, and am just getting started in this area, but this struck me as a good strategy to use. Also looking at neighborhood trends. My worry is that this wouldn't work because any undervalued property has its price for a "reason", e.g. it is already at its equilibrium price by the time you've determined that its "undervalued", and in fact you won't be able to flip it for whatever reason for a huge profit.

I don't understand how that could be the case though. What seems more intuitive for me is that these are the homes that just happen to have so far been ignored by other investors, or else passed over for better deals, since a small group of investors cannot snatch up all of the deals (otherwise why would this forum exist haha).

Another argument might be that you can never tell when a neighborhood is going to crash in its property values, or when people will migrate to a nearby one that is flourishing. If this were the case, then your model wouldn't be effective anyway. Presumably you can also protect against this with the aforementioned neighborhood growth trends.


1Reply
223 views

Most Popular Reply

Bob S.
Real Estate Consultant · Cleveland · Member since 2020 · 6k+ posts · 3k+ votes
2y
Quote from @David D.:

I am curious if anyone has employed the simple strategy I was thinking of for investing in any kind of property:

1. Use something like ATTOM API to get historical sales snapshots, or just all the historical sales of the property class for a metropolitan area. 
2. Train a regression model using square footage, bathroom count, or more advanced features to predict the price of the sales. 
3. Get the error of your predicted housing price (if you're using price per square foot, convert to housing price) down to $10K-$40K or so. 
4. Find properties with characteristics that predict they should be selling for 2 standard deviations above what their actual price/value is (or basically just properties that are selling for way below what the model predicts they should sell for).

The challenge here is of course step 3. I just started experimenting with Downtown Detroit with ATTOM's "sales snapshots" and I haven't gotten that kind of performance yet. However, I'm sure many people have, especially if they have images or floor plans of the houses/other features and lots of data. 

I apologize if this is all just fairly standard financial modeling. Very key thing here, I don't have a finance/economics background, and am just getting started in this area, but this struck me as a good strategy to use. Also looking at neighborhood trends. My worry is that this wouldn't work because any undervalued property has its price for a "reason", e.g. it is already at its equilibrium price by the time you've determined that its "undervalued", and in fact you won't be able to flip it for whatever reason for a huge profit.

I don't understand how that could be the case though. What seems more intuitive for me is that these are the homes that just happen to have so far been ignored by other investors, or else passed over for better deals, since a small group of investors cannot snatch up all of the deals (otherwise why would this forum exist haha).

Another argument might be that you can never tell when a neighborhood is going to crash in its property values, or when people will migrate to a nearby one that is flourishing. If this were the case, then your model wouldn't be effective anyway. Presumably you can also protect against this with the aforementioned neighborhood growth trends.


Everyone is trying to reinvent the wheel. Run comps, drive the area, buy right, reno right, rent/sell right. Yes, it's that simple. I read post after post, I have been researching this area for years. Really years WTH are you doing? It took me 2 hours YES 2 hours to realize my market was where I wanted to be. My 1st day, I locked up two duplexes'. Since then, lets just say I have done one or two. It's not rocket science, and I did not need any special calculators or fancy bells and whistles. RE is just math. Know your numbers. 
All the best  

3Reply
See this reply in the discussion

49 Replies

Jump to latestLatest
  • Carlos Ptriawan
    Member since 2019 · 7k+ posts · 4k+ votes
    2y

    this is what privy/propstream does but you can just simply use redfin

    0Reply
  • Drew SygitBusiness Member
    Property Manager · Royal Oak, MI · Member since 2012 · 12k+ posts · 9k+ votes
    2y

    @David D. there are challenges with what you want to do, specifically in the City of Detroit:

    1) You'd have to do this for each of the city's 185 or so Neighborhoods. Zip codes are too big.

    2) You'd have to factor in Neighborhood population density somehow. Not aware of a reliable source for this info - and please don't think US Census Bureau is the answer.

    3) How do you account for "property condition"?

    [email protected]248-209-6824
    Logical Property Management4.9458 Reviews
    WebsiteView Page
    1Reply
  • Jonathan GreeneBusiness Member
    Real Estate Consultant · Madison, NJ · Member since 2016 · 6k+ posts · 7k+ votes
    2y

    There are companies with billions of dollars in war chests doing this all the time. There is no reason to think that this will help you invest in real estate. You are basically trying to take the real estate out of real estate.

    Predictive analytics are best used as guides to the work. You sound (I could be wrong, but probably not) like you want to use your background to eliminate some of the guesswork, but real estate investing is learned by being at properties, talking with other investors, going to meetups, etc.

    Predictive data for investors is wholly flawed because it can't see inside of the houses.

    [email protected]Podcast Guest on Show BiggerPockets Daily 667: 5 Amazing Benefits Multifamily Investments Offer (That Single Family Homes Don’t) by Gino Barbaro
    Zen and the Art of Real Estate Investing59 Reviews
    WebsiteView Page
    1Reply
  • David D.
    OP
    New to Real Estate · Member since 2024 · 19 posts · 3 votes
    2y
    Quote from @Drew Sygit:

    @David D. there are challenges with what you want to do, specifically in the City of Detroit:

    1) You'd have to do this for each of the city's 185 or so Neighborhoods. Zip codes are too big.

    2) You'd have to factor in Neighborhood population density somehow. Not aware of a reliable source for this info - and please don't think US Census Bureau is the answer.

    3) How do you account for "property condition"?


    1) Here's one neighborhood level strategy. Take 48201 (midtown Detroit). I could divide up into neighborhoods and take the homes with the greatest difference between predicted and actual values in each of the neighborhood level models. I'm not sure why this would be better than the original strategy though. I'm skeptical you'd get enough neighborhood-level sales.

     2) I just found out ATTOM has 4 levels of neighborhood data, even down to residential subdivision, and pop density for each! https://www.attomdata.com/data/boundaries-data/area-neighbor... . Excited to try modeling at this level (they also do 5 year projections).

    3) One strategy might be to do the same thing we just did with sales, but for property condition. I plug in known properties, fit a curve as in sales (baths, pop density, and so on), and try to predict the property's condition based on those features within an error rate (except this time you'd be predicting the probability of a condition, as opposed to sales price). Another simple idea is to just have an expert assess the condition of a few of the best properties you find with the original strategy. 

    Another idea I had was to just use standard research out there on building systems like the air conditioning, wiring, plumbing and so on, to estimate how much time they would normally take to require repairs.

    Here's a paper that discusses "energy modeling" for single family homes that is pretty recent that is an example of what I'm thinking of:

    https://www.mdpi.com/1996-1073/12/8/1537

    0Reply
  • David D.
    OP
    New to Real Estate · Member since 2024 · 19 posts · 3 votes
    2y
    Quote from @Jonathan Greene:

    There are companies with billions of dollars in war chests doing this all the time. There is no reason to think that this will help you invest in real estate. You are basically trying to take the real estate out of real estate.

    Predictive analytics are best used as guides to the work. You sound (I could be wrong, but probably not) like you want to use your background to eliminate some of the guesswork, but real estate investing is learned by being at properties, talking with other investors, going to meetups, etc.

    Predictive data for investors is wholly flawed because it can't see inside of the houses.

    This makes sense. Basically what I'm doing is saying "can a simple model and strategy get me 70% or more of the profits that a very experienced investor with 20 years of experience would get?". The idea that I'd need to know the internals of the home makes a lot of sense. You might think that 50% or more of the time I'll lose money on the investment because the model doesn't explicitly include those. However, then you could just pull a bunch of floorplans/wiring diagrams and photos of the home from an MLS and use those as features to predict price.

    I’m also not 100% sure that model would be better than a “no internals” model, especially if the no internals one uses neighborhood points of interest and other geospatial stuff.

    also, I just heard the latest episode of your podcast and it had some helpful flipping advice! The strategy here I think would be to just do a cosmetic remodel to a home that is highly unlikely to need any other improvements. Since I’m looking at apartments mostly, that also makes things kind of easier.

    0Reply
  • Jonathan GreeneBusiness Member
    Real Estate Consultant · Madison, NJ · Member since 2016 · 6k+ posts · 7k+ votes
    2y
    Quote from @David D.:
    Quote from @Jonathan Greene:

    There are companies with billions of dollars in war chests doing this all the time. There is no reason to think that this will help you invest in real estate. You are basically trying to take the real estate out of real estate.

    Predictive analytics are best used as guides to the work. You sound (I could be wrong, but probably not) like you want to use your background to eliminate some of the guesswork, but real estate investing is learned by being at properties, talking with other investors, going to meetups, etc.

    Predictive data for investors is wholly flawed because it can't see inside of the houses.

    This makes sense. Basically what I'm doing is saying "can a simple model and strategy get me 70% or more of the profits that a very experienced investor with 20 years of experience would get?". The idea that I'd need to know the internals of the home makes a lot of sense. You might think that 50% or more of the time I'll lose money on the investment because the model doesn't explicitly include those. However, then you could just pull a bunch of floorplans/wiring diagrams and photos of the home from an MLS and use those as features to predict price.

    I’m also not 100% sure that model would be better than a “no internals” model, especially if the no internals one uses neighborhood points of interest and other geospatial stuff.

    also, I just heard the latest episode of your podcast and it had some helpful flipping advice! The strategy here I think would be to just do a cosmetic remodel to a home that is highly unlikely to need any other improvements. Since I’m looking at apartments mostly, that also makes things kind of easier.


    This is a similar model to Open Door or the other companies that bought a ton based solely on data and analytic,s and not real estate acumen. They have lost more money than anyone can count. The model will seem like it will work, but what happens is that a small market shift tanks the entire algorithm in one month and then you can't sell anything.

    [email protected]Podcast Guest on Show BiggerPockets Daily 667: 5 Amazing Benefits Multifamily Investments Offer (That Single Family Homes Don’t) by Gino Barbaro
    Zen and the Art of Real Estate Investing59 Reviews
    WebsiteView Page
    0Reply
  • David D.
    OP
    New to Real Estate · Member since 2024 · 19 posts · 3 votes
    2y
    Quote from @Jonathan Greene:
    Quote from @David D.:
    Quote from @Jonathan Greene:

    There are companies with billions of dollars in war chests doing this all the time. There is no reason to think that this will help you invest in real estate. You are basically trying to take the real estate out of real estate.

    Predictive analytics are best used as guides to the work. You sound (I could be wrong, but probably not) like you want to use your background to eliminate some of the guesswork, but real estate investing is learned by being at properties, talking with other investors, going to meetups, etc.

    Predictive data for investors is wholly flawed because it can't see inside of the houses.

    This makes sense. Basically what I'm doing is saying "can a simple model and strategy get me 70% or more of the profits that a very experienced investor with 20 years of experience would get?". The idea that I'd need to know the internals of the home makes a lot of sense. You might think that 50% or more of the time I'll lose money on the investment because the model doesn't explicitly include those. However, then you could just pull a bunch of floorplans/wiring diagrams and photos of the home from an MLS and use those as features to predict price.

    I’m also not 100% sure that model would be better than a “no internals” model, especially if the no internals one uses neighborhood points of interest and other geospatial stuff.

    also, I just heard the latest episode of your podcast and it had some helpful flipping advice! The strategy here I think would be to just do a cosmetic remodel to a home that is highly unlikely to need any other improvements. Since I’m looking at apartments mostly, that also makes things kind of easier.


    This is a similar model to Open Door or the other companies that bought a ton based solely on data and analytic,s and not real estate acumen. They have lost more money than anyone can count. The model will seem like it will work, but what happens is that a small market shift tanks the entire algorithm in one month and then you can't sell anything.


     My hope is that I would avoid this by being selective about the properties and not just having a "set it and forget it" algorithm. That is, I would let the model choose a couple of nice properties for me, and then actually see them to assess them with an expert. I figure Open Door just chose massive bundles of properties and set up a bunch of automated processes for buying, but I don't know. 

    0Reply
  • Bob S.
    Real Estate Consultant · Cleveland · Member since 2020 · 6k+ posts · 3k+ votes
    2y
    Quote from @David D.:

    I am curious if anyone has employed the simple strategy I was thinking of for investing in any kind of property:

    1. Use something like ATTOM API to get historical sales snapshots, or just all the historical sales of the property class for a metropolitan area. 
    2. Train a regression model using square footage, bathroom count, or more advanced features to predict the price of the sales. 
    3. Get the error of your predicted housing price (if you're using price per square foot, convert to housing price) down to $10K-$40K or so. 
    4. Find properties with characteristics that predict they should be selling for 2 standard deviations above what their actual price/value is (or basically just properties that are selling for way below what the model predicts they should sell for).

    The challenge here is of course step 3. I just started experimenting with Downtown Detroit with ATTOM's "sales snapshots" and I haven't gotten that kind of performance yet. However, I'm sure many people have, especially if they have images or floor plans of the houses/other features and lots of data. 

    I apologize if this is all just fairly standard financial modeling. Very key thing here, I don't have a finance/economics background, and am just getting started in this area, but this struck me as a good strategy to use. Also looking at neighborhood trends. My worry is that this wouldn't work because any undervalued property has its price for a "reason", e.g. it is already at its equilibrium price by the time you've determined that its "undervalued", and in fact you won't be able to flip it for whatever reason for a huge profit.

    I don't understand how that could be the case though. What seems more intuitive for me is that these are the homes that just happen to have so far been ignored by other investors, or else passed over for better deals, since a small group of investors cannot snatch up all of the deals (otherwise why would this forum exist haha).

    Another argument might be that you can never tell when a neighborhood is going to crash in its property values, or when people will migrate to a nearby one that is flourishing. If this were the case, then your model wouldn't be effective anyway. Presumably you can also protect against this with the aforementioned neighborhood growth trends.


    Everyone is trying to reinvent the wheel. Run comps, drive the area, buy right, reno right, rent/sell right. Yes, it's that simple. I read post after post, I have been researching this area for years. Really years WTH are you doing? It took me 2 hours YES 2 hours to realize my market was where I wanted to be. My 1st day, I locked up two duplexes'. Since then, lets just say I have done one or two. It's not rocket science, and I did not need any special calculators or fancy bells and whistles. RE is just math. Know your numbers. 
    All the best  

    3Reply
  • Bradley BuxtonBusiness Member
    Real Estate Agent · NV · Member since 2023 · 1k+ posts · 713 votes
    2y

    @David D.

    Investing in real estate is a business and buyers and sellers make irrational decisions.  Not every buyer is an investors and will not look at a property only based on numbers.  People buy because of infinite factors. People sell homes for infinite factors. This is why a machine cannot calculate for the irrationality of a person buying a house over value because they like the cute garden. The large institutional investors look at the comparable prices in an area (costs) and rents (income), they find the best properties that match the calculation then they go look at the property.  There is no substitute (yet) for the human factor.

     If you're interested in modeling for a market check out https://www.addressincome.com/address-income  They have wall street tools for main street investors. It's Nevada only right now but it might be close to what you are looking for. 

    1Reply
  • Carlos Ptriawan
    Member since 2019 · 7k+ posts · 4k+ votes
    2y
    Quote from @David D.:

    I am curious if anyone has employed the simple strategy I was thinking of for investing in any kind of property:

    1. Use something like ATTOM API to get historical sales snapshots, or just all the historical sales of the property class for a metropolitan area. 
    2. Train a regression model using square footage, bathroom count, or more advanced features to predict the price of the sales. 
    3. Get the error of your predicted housing price (if you're using price per square foot, convert to housing price) down to $10K-$40K or so. 
    4. Find properties with characteristics that predict they should be selling for 2 standard deviations above what their actual price/value is (or basically just properties that are selling for way below what the model predicts they should sell for).

    The challenge here is of course step 3. I just started experimenting with Downtown Detroit with ATTOM's "sales snapshots" and I haven't gotten that kind of performance yet. However, I'm sure many people have, especially if they have images or floor plans of the houses/other features and lots of data. 

    I apologize if this is all just fairly standard financial modeling. Very key thing here, I don't have a finance/economics background, and am just getting started in this area, but this struck me as a good strategy to use. Also looking at neighborhood trends. My worry is that this wouldn't work because any undervalued property has its price for a "reason", e.g. it is already at its equilibrium price by the time you've determined that its "undervalued", and in fact you won't be able to flip it for whatever reason for a huge profit.

    I don't understand how that could be the case though. What seems more intuitive for me is that these are the homes that just happen to have so far been ignored by other investors, or else passed over for better deals, since a small group of investors cannot snatch up all of the deals (otherwise why would this forum exist haha).

    Another argument might be that you can never tell when a neighborhood is going to crash in its property values, or when people will migrate to a nearby one that is flourishing. If this were the case, then your model wouldn't be effective anyway. Presumably you can also protect against this with the aforementioned neighborhood growth trends.



     The predictive model is actually very easy to create.
    1. You sorted out market and zip code that has positive appreciation
    2. You sort everything based on DOM
    3. Make time interval, lets say within 3 years period, you check a city that has reducing DOM within 3 years
    4. Redfin all these information
    5. Results are that Detroit is part of that city. Not surprisingly, ZHVI also has very similar data result
    6. You sorted again within the city of Detroit, what home has the lowest gap to the mean
    7. There you go, submit your bid

    But Privy already does that and few other software too. But honestly you can do the same by just downloading from Redfin data periodically. Been doing that for over ten years now lol

    1Reply
  • David D.
    OP
    New to Real Estate · Member since 2024 · 19 posts · 3 votes
    2y
    Quote from @Carlos Ptriawan:
    Quote from @David D.:

    I am curious if anyone has employed the simple strategy I was thinking of for investing in any kind of property:

    1. Use something like ATTOM API to get historical sales snapshots, or just all the historical sales of the property class for a metropolitan area. 
    2. Train a regression model using square footage, bathroom count, or more advanced features to predict the price of the sales. 
    3. Get the error of your predicted housing price (if you're using price per square foot, convert to housing price) down to $10K-$40K or so. 
    4. Find properties with characteristics that predict they should be selling for 2 standard deviations above what their actual price/value is (or basically just properties that are selling for way below what the model predicts they should sell for).

    The challenge here is of course step 3. I just started experimenting with Downtown Detroit with ATTOM's "sales snapshots" and I haven't gotten that kind of performance yet. However, I'm sure many people have, especially if they have images or floor plans of the houses/other features and lots of data. 

    I apologize if this is all just fairly standard financial modeling. Very key thing here, I don't have a finance/economics background, and am just getting started in this area, but this struck me as a good strategy to use. Also looking at neighborhood trends. My worry is that this wouldn't work because any undervalued property has its price for a "reason", e.g. it is already at its equilibrium price by the time you've determined that its "undervalued", and in fact you won't be able to flip it for whatever reason for a huge profit.

    I don't understand how that could be the case though. What seems more intuitive for me is that these are the homes that just happen to have so far been ignored by other investors, or else passed over for better deals, since a small group of investors cannot snatch up all of the deals (otherwise why would this forum exist haha).

    Another argument might be that you can never tell when a neighborhood is going to crash in its property values, or when people will migrate to a nearby one that is flourishing. If this were the case, then your model wouldn't be effective anyway. Presumably you can also protect against this with the aforementioned neighborhood growth trends.



     The predictive model is actually very easy to create.
    1. You sorted out market and zip code that has positive appreciation
    2. You sort everything based on DOM
    3. Make time interval, lets say within 3 years period, you check a city that has reducing DOM within 3 years
    4. Redfin all these information
    5. Results are that Detroit is part of that city. Not surprisingly, ZHVI also has very similar data result
    6. You sorted again within the city of Detroit, what home has the lowest gap to the mean
    7. There you go, submit your bid

    But Privy already does that and few other software too. But honestly you can do the same by just downloading from Redfin data periodically. Been doing that for over ten years now lol


     A Redfin export seems pretty useful, but don't they usually use very aggregated data? I'm not sure which export to use as I didn't find historical sales. 

    These other aspects of the strategy seem great, really reduces the chances the neighborhood is just going to tank in value or something I would think, though I suspect fitting a curve as opposed to just using the mean may be a bit more effective and less risky. I'll try it out! Would you feel comfortable sharing your approximate "loss rate"? That is, the rough percentage of the investments you make that end up costing you? 

    0Reply
  • Carlos Ptriawan
    Member since 2019 · 7k+ posts · 4k+ votes
    2y
    Quote from @David D.:
    Quote from @Carlos Ptriawan:
    Quote from @David D.:

    I am curious if anyone has employed the simple strategy I was thinking of for investing in any kind of property:

    1. Use something like ATTOM API to get historical sales snapshots, or just all the historical sales of the property class for a metropolitan area. 
    2. Train a regression model using square footage, bathroom count, or more advanced features to predict the price of the sales. 
    3. Get the error of your predicted housing price (if you're using price per square foot, convert to housing price) down to $10K-$40K or so. 
    4. Find properties with characteristics that predict they should be selling for 2 standard deviations above what their actual price/value is (or basically just properties that are selling for way below what the model predicts they should sell for).

    The challenge here is of course step 3. I just started experimenting with Downtown Detroit with ATTOM's "sales snapshots" and I haven't gotten that kind of performance yet. However, I'm sure many people have, especially if they have images or floor plans of the houses/other features and lots of data. 

    I apologize if this is all just fairly standard financial modeling. Very key thing here, I don't have a finance/economics background, and am just getting started in this area, but this struck me as a good strategy to use. Also looking at neighborhood trends. My worry is that this wouldn't work because any undervalued property has its price for a "reason", e.g. it is already at its equilibrium price by the time you've determined that its "undervalued", and in fact you won't be able to flip it for whatever reason for a huge profit.

    I don't understand how that could be the case though. What seems more intuitive for me is that these are the homes that just happen to have so far been ignored by other investors, or else passed over for better deals, since a small group of investors cannot snatch up all of the deals (otherwise why would this forum exist haha).

    Another argument might be that you can never tell when a neighborhood is going to crash in its property values, or when people will migrate to a nearby one that is flourishing. If this were the case, then your model wouldn't be effective anyway. Presumably you can also protect against this with the aforementioned neighborhood growth trends.



     The predictive model is actually very easy to create.
    1. You sorted out market and zip code that has positive appreciation
    2. You sort everything based on DOM
    3. Make time interval, lets say within 3 years period, you check a city that has reducing DOM within 3 years
    4. Redfin all these information
    5. Results are that Detroit is part of that city. Not surprisingly, ZHVI also has very similar data result
    6. You sorted again within the city of Detroit, what home has the lowest gap to the mean
    7. There you go, submit your bid

    But Privy already does that and few other software too. But honestly you can do the same by just downloading from Redfin data periodically. Been doing that for over ten years now lol


     A Redfin export seems pretty useful, but don't they usually use very aggregated data? I'm not sure which export to use as I didn't find historical sales. 

    These other aspects of the strategy seem great, really reduces the chances the neighborhood is just going to tank in value or something I would think, though I suspect fitting a curve as opposed to just using the mean may be a bit more effective and less risky. I'll try it out! Would you feel comfortable sharing your approximate "loss rate"? That is, the rough percentage of the investments you make that end up costing you? 


     you can download recent home sales using Redfin, for example : query 3/1 SF in three Detroit zip code. Put it to excel file. Upload it to tableau for nice graphic so on.

    I've been investing in last ten years using prediction and it's pretty accurate, it's not hard to made prediction based on seasonality and historical chart. Feb to July usually has the strongest activity and then flat after September.

    Problem with real estate is that ........ it is too easy, if you download the data that you interested to look for, you only have 30-50 homes and make the plot or prediction based on this input is again.....very easy.

    Also Axos has good data as well using Listing to Sold Ratio listing.

    1Reply
  • Steven S.
    Specialist · LA & Ventura · Member since 2023 · 119 posts · 60 votes
    2y

    I do this right now. To identify a good market for me, I like to see good $/sf growth over the last 2 years, and a market that is $700/sf and higher. This chart basically shows me which City's are most desirable (can do the same with Zip Codes):

    Anything to the right of 0 means the $/sf is growing QoQ on average, and the height is the current $/sf of that market. As you can see, there are a few City's that are at 5%+ QoQ $/sf growth, but lots more towards 0-3%. Instantly I can tell the hottest markets from the normal and cold ones, and separately determine which one's I am priced out of due to crazy high values.

    As for your auto-ARV feature, yes it can be done. Take this 💥just-closed renovated property for example💥:

    With my query of "22757 Charlemont Pl, Woodland Hills, CA 91364 5/4/2400-3050,0.6,300" we get an ARV Estimate that is ~4% off, and I picked this at random just scrolling through recently closed in my area

    And the Market Stats $/sf of renovated property in that Zip Code matches up pretty well, so I know the ARV estimate is realistic:

    I also set this up to run through every new-listing in my area, and generate a summary sheet of good-potential deals:


    So it's all possible to do, you just have to get your data and build on top of it. Would be happy to help if you already have lined up a source of transaction-level data for your market (a row for each sale occuring, including the Close Price, Address, BR, BTH, SF, Lot SF, DOM, Lat, Lon, Listing Description, etc)

    2Reply
  • Carlos Ptriawan
    Member since 2019 · 7k+ posts · 4k+ votes
    2y

    Nice Steven....

    And AI can give comps very fast as well these days lol :

    Based on the recent home sales data for Woodland Hills, CA, it's difficult to determine if 22757 Charlemont Pl specifically has a good ARV (after repair value) without more details on the property's condition, size, and features compared to other recently sold homes. However, we can make some general observations:
    The median sale price for homes in Woodland Hills was around $1.2-1.3 million in recent months, with prices trending up 4-7% year-over-year.
    Homes are selling relatively quickly, with an average days on market of 35-44 days. Hot homes can sell even faster, in around 15-24 days.
    The price per square foot averages around $640.
    Example recent sales include:
    $1,140,000 for a 4 bed, 3 bath, 2,228 sqft home on Oxnard St
    $1,300,000 for a 3 bed, 2 bath, 1,936 sqft home on Hatteras St
    $1,612,500 for a 3 bed, 2 bath, 2,161 sqft home on Oakdale Ave

    0Reply
  • Steven S.
    Specialist · LA & Ventura · Member since 2023 · 119 posts · 60 votes
    2y
    Quote from @Carlos Ptriawan:

    Nice Steven....

    And AI can give comps very fast as well these days lol :

    Based on the recent home sales data for Woodland Hills, CA, it's difficult to determine if 22757 Charlemont Pl specifically has a good ARV (after repair value) without more details on the property's condition, size, and features compared to other recently sold homes. However, we can make some general observations:
    The median sale price for homes in Woodland Hills was around $1.2-1.3 million in recent months, with prices trending up 4-7% year-over-year.
    Homes are selling relatively quickly, with an average days on market of 35-44 days. Hot homes can sell even faster, in around 15-24 days.
    The price per square foot averages around $640.
    Example recent sales include:
    $1,140,000 for a 4 bed, 3 bath, 2,228 sqft home on Oxnard St
    $1,300,000 for a 3 bed, 2 bath, 1,936 sqft home on Hatteras St
    $1,612,500 for a 3 bed, 2 bath, 2,161 sqft home on Oakdale Ave

    Ai Comps... lol!

    They are just pulling search results and copying the RedFin/Zillow/Realtor.com, etc estimated values/info on the pages:

    And on Redfin, the exact same numbers:

    So if you are just relying on RedFin/Zillow Estimates, then sure ChatGPT4 can help you with that, but anything more concrete/detailed and they can't (no historic $/sf of renovated homes available, only average/median, etc)

    1Reply
  • Carlos Ptriawan
    Member since 2019 · 7k+ posts · 4k+ votes
    2y

    they know about kitchen renovation too LOL :

    The property value of 22757 Charlemont Pl in Woodland Hills, CA has increased significantly over the years.The key details on the property value changes are:

    • In August 2016, the property was sold for $941,000, which translates to around $345 per square foot.
    • In June 2024, the property was sold again for $1,857,500, about 97% higher than the 2016 sale price.
    • The 2024 sale price of $1,857,500 for the 2,730 square foot home equates to approximately $680 per square foot.

    So in the span of around 8 years from 2016 to 2024, the property value nearly doubled, increasing from $941,000 to $1,857,500. This significant increase reflects the rapidly appreciating real estate values in the Woodland Hills area of Los Angeles over this period.The search results also note that the property underwent some renovations, such as a remodeled kitchen with high-end appliances, which likely contributed to the higher selling price in 2024 compared to 2016, , the property value of 22757 Charlemont Pl in Woodland Hills appears to be higher than many similar homes in the surrounding area, but is generally in line with the rapidly appreciating real estate market in this neighborhood.Here are some key comparisons:

    • The sale price of $1,857,500 for this 2,730 sq ft, 5 bedroom, 4 bathroom home in June 2024 equates to around $680 per square foot.
    • The median sale price per square foot for homes in the Woodland Hills area in April 2024 was $638, up 9.7% from the previous year according to Redfin. So 22757 Charlemont Pl sold at a premium over the median price per sq ft.
    • However, Redfin notes that "hot homes" in Woodland Hills can sell for around 4% above list price and go pending in around 15 days. 22757 Charlemont Pl sold for about 2% below its $1.9 million list price after being on the market for a bit longer at 30 days.
    • Zillow data shows the typical home value in Woodland Hills in April 2024 was $1,257,932, up 9.6% year-over-year. At $1.86 million, 22757 Charlemont Pl sold at a premium to this typical value.
    0Reply
  • Steven S.
    Specialist · LA & Ventura · Member since 2023 · 119 posts · 60 votes
    2y

    That's the same info I could find on the RedFin link for the property, it's summary is great, but it's just nothing like what the guy here is trying to do

    0Reply
  • Carlos Ptriawan
    Member since 2019 · 7k+ posts · 4k+ votes
    2y
    Quote from @Steven S.:

    That's the same info I could find on the RedFin link for the property, it's summary is great, but it's just nothing like what the guy here is trying to do


     It's the same at the end. To investigate market you just need Sale List of trends such as this:
    https://www.redfin.com/neighborhood/10958/CA/San-Jose/East-S... 

    Then if one wants to be very specific, do a query in Redfin of sold homes. Download it and put the Excel file, create a chart. Use the future appreciation by using ZHVI. Also since there are 30-40 homes only, it's easy to create mean or average statistics. Software like Privy or yours (like we discussed a few months ago) can detect if the house is flipped or not, based on the picture or recent data.

    I think something like Tableau now should have something like automatic prediction when you feed your data to them. For real estate this is just too easy as datasets are too small.

    1Reply
  • David D.
    OP
    New to Real Estate · Member since 2024 · 19 posts · 3 votes
    2y

    These are two different but related strategies it sounds like. Steven is using an ARV model that's homegrown, and ChatGPT is using Redfin's ARV model, probably with its own intuitions about repair values baked in there too.

    These are still different from my approach though, I think? I'm assuming Steven is also using a small dataset of Comps. I've been skeptical of Comps since first hearing about them. If there's only a few comps you're using, there's naturally going to be a lot of error in your valuation model, so you're at greater risk of buying a home that really isn't going to net you much of anything. 

    That's why I like this traditional ML approach, we leverage lots of data to inform our decision about which house is seriously "undervalued"

    I have a couple of questions about both of your approaches though, will be interesting to compare them!

    1. In Steven's approach, is the strategy to choose a house that has a high ARV but low current price, or to just select based on the ARV? 
    2. Have both of you found that these approaches are accurate in your own investments?
    3. Do you think an approach with floorplans (which I just found out Redfin has), would be more effective? 
    4. I'm curious about whether you've tried to model the effects of remodeling alone. For instance, let's say I remodel the home completely, have a completely different look inside of it than it had prior, as opposed to the smallest possible remodel. There is presumably a sweet spot in between that has the best "value" on average. Have you explored this? 

    1Reply
  • Carlos Ptriawan
    Member since 2019 · 7k+ posts · 4k+ votes
    2y
    Quote from @David D.:

    These are two different but related strategies it sounds like. Steven is using an ARV model that's homegrown, and ChatGPT is using Redfin's ARV model, probably with its own intuitions about repair values baked in there too.

    These are still different from my approach though, I think? I'm assuming Steven is also using a small dataset of Comps. I've been skeptical of Comps since first hearing about them. If there's only a few comps you're using, there's naturally going to be a lot of error in your valuation model, so you're at greater risk of buying a home that really isn't going to net you much of anything. 

    That's why I like this traditional ML approach, we leverage lots of data to inform our decision about which house is seriously "undervalued"

    I have a couple of questions about both of your approaches though, will be interesting to compare them!

    1. In Steven's approach, is the strategy to choose a house that has a high ARV but low current price, or to just select based on the ARV? 
    2. Have both of you found that these approaches are accurate in your own investments?
    3. Do you think an approach with floorplans (which I just found out Redfin has), would be more effective? 
    4. I'm curious about whether you've tried to model the effects of remodeling alone. For instance, let's say I remodel the home completely, have a completely different look inside of it than it had prior, as opposed to the smallest possible remodel. There is presumably a sweet spot in between that has the best "value" on average. Have you explored this? 


    Floorplan is very important.

    Also David have you try ever to use Privy or Pellego. Both software can guess-estimate the rehab cost from the pic. I check the inspection and picture, and create a range of estimates for rehab budget, for example 25k-50k or 50k-75k based on which house that needs to be remodelled. Then add 10-20k for additional unexpected budgeting. 

    For number 2. Yes. Every single time my prediction is deadly accurate. I even know when is best time to sell (Second week of February).

    Most home, when needs to be fully remodelled, would cost me around 80k-100k. So for something that low comp is 800k and high comp is about 900k. You want to buy in the price of 600-680k. One way to make remodelling works is when you buy home that has a larger lot, then your creativity could kick in to create value.

    Also if you do rehab in good market, you better rent it out for two years before selling so you can catch rent, reduce tax and make more money. If you can add additional room and/or kitchen the value of the home is higher.

    Why I said this data is becoming too easy is because, if you chase in good market (DOM < 10/15), market is efficient so much so that the best house to buy is usually home with price reduction or having DOM double than the normal. If I visited only these house I can still find good deal.

    0Reply
  • Carlos Ptriawan
    Member since 2019 · 7k+ posts · 4k+ votes
    2y

    Also Dave this is one of bit of secret lol , if you see this chart , from month-to-month perspectiveyou would notice for this market, price is up when inventory is low and price would start to drop when inventory is at peak, this is repeatable seasonality in this market.

    1Reply
  • Carlos Ptriawan
    Member since 2019 · 7k+ posts · 4k+ votes
    2y

    Also do notice when inventory is hitting new record low like In 2017, there's large spike increase in home price. Knowing these data is key to make profit very kosher and ultra-profitable. You want to buy when market is tired/stagnant and sell to FOMO buyers when inventory at ATL lol kind of same principle when buying/selling stock lol

    0Reply
  • Peter W.
    Member since 2020 · 351 posts · 329 votes
    2y

    I think your fundamental problem is that people who are going to sell 2 standard deviations below market price aren’t on mls. Even if they are on mls so one is going to end up making an offer close to market value. You can sometimes get something that’s maybe 5% below market but it’s tough.


    This isn’t a complex business (although execution can be tough). I don’t think there is a ton of juice in advanced analytics (in terms of selecting houses). In terms of selecting markets, maybe.

    2Reply
  • Steven S.
    Specialist · LA & Ventura · Member since 2023 · 119 posts · 60 votes
    2y
    Quote from @David D.:

    These are two different but related strategies it sounds like. Steven is using an ARV model that's homegrown, and ChatGPT is using Redfin's ARV model, probably with its own intuitions about repair values baked in there too.

    These are still different from my approach though, I think? I'm assuming Steven is also using a small dataset of Comps. I've been skeptical of Comps since first hearing about them. If there's only a few comps you're using, there's naturally going to be a lot of error in your valuation model, so you're at greater risk of buying a home that really isn't going to net you much of anything. 

    That's why I like this traditional ML approach, we leverage lots of data to inform our decision about which house is seriously "undervalued"

    ChatGPT & any AI is just going to give you a summary of the web pages it can find, but it doesn't have all the historic data to calculate on, it's just regurgitating a web page it found. So not useful at all (unless you literally buy based on the RedFin/Zillow ARVs)

    As for my tool, every transaction that occurs in my market area is another row in my database. I have tens of millions of rows, with data going back to 2000, for the LA/Ventura County area. It's so much that you can't open it in Excel (if you tried to... lol). 

    From this source-data, I calculate everything you see in my above reply, and built things on top of that data for users to generate their own market stats, run comps, etc. The only way to value a home is by using sales comps, if you use any other method it's not going to be as accurate. So for your tool, you need exactly what I have, a transaction-level dataset, for your market, for your tools/programs to calculate on.

    I showed you above you don't need Ai/ML to accurately determine the As-is and Renovated value. What you need is the source-data, a programmer to build you an app based on this data, and a experienced developer to instruct the programmer on how the algorithms/functions should perform these actions so they are actually useful. Without all 3 of these things readily available to you, it's a pointless endeavor.

    You mention you like the traditional ML approach, where "we leverage lots of data to inform our decision about which house is seriously 'undervalued'", what would that be to you? Giving ChatGPT a 25GB file of your source data and having it go through it? Training your own AI on your transaction-level source data?
    1Reply
  • David D.
    OP
    New to Real Estate · Member since 2024 · 19 posts · 3 votes
    2y

    I'm not sure just regurgitating webpages is how it's doing that, but that's a topic for another thread I guess!

    For data size, I'm not using a particularly large amount of transactions so far, and am getting pretty good, but still imperfect valuations of Detroit homes. Nowhere near your error rate of only 4% if that's representative of a typical error in an estimate. If you're doing that using Comparative Market Analysis, then there must be something to the technique surely. I just don't really understand how it works, because it seems to me (intuitively) like it would only introduce a lot of error into the model. I should probably read a book about it realistically!

    But yeah to summarize again, the goal here is to take a relatively standard $500,000 home, for instance, which is currently priced at $400,000, and sell it for what it's "actually worth". The main pushback on that idea from some others in the thread I sense is that there's a reason that the home would be priced so much lower, but the idea is to look at homes that don't have those defects.

    In my downtown Detroit dataset, for example, the average difference between predicted and actual price of the top 100 "undervalued" deals is $118,000. The idea is that you choose 10 of these randomly, and just visit them. The chances that all of them have some $100,000 list of defects is presumably very low, especially if that $100,000 list of defects is undetectable by an expert. Or put another way, I haven't seen any explanation of how that could be that makes sense to me. 

    Carlos mentions that you can also exploit seasonality to get comparable profits like this, and recommends some ranges to seek remodels in (thank you for sharing those!). Again, very curious to try a comp analysis informed by Gen AI, but perhaps not going to be the first thing I end up actually trying in practice haha

    0Reply
Join the conversationCreate a free account to reply, vote on answers and follow this thread.
Log InJoin for Free