All Forum Posts by: Jason L.

I believe you're misreading the context of that chart. "Within 5% of sales price" is not the same thing as "within 5% of the median value for the entire metro". Averages are often misleading. On a more granular basis, the Zestimate prediction might have a harder time predicting houses that are more expensive (fewer comps, more days on market, higher price volatility) than ones in smaller price ranges where the AI has more to learn from and where price volatility will be lower. 

The way you are interpreting it would be to say that even houses that sell for $150k are having $30k in error, which would be 20% and not 5%. The error would actually only be $7500. I would point to my post above for an explanation on what the Zestimate is trying to accomplish and how it actually works.

I have a background in real estate and data analytics so maybe I can shed a little light here.

It's important to remember the context of what you are looking at. A Zestimate is an output from a machine learning algorithm. It is not a broker estimate. There are many pros and cons to this.

Pros:

--A computer can parse and aggregate millions of rows of data in seconds. Imagine if you asked your broker to do a CMA analyzing every single house in your zip code on 100 different data points. How long do you think it would take your broker to do that? Weeks? Months? That's what the Zestimate is doing for EVERY house in the country, every single day.

--A computer can change complexity of a question without much sacrifice for time or efficiency. (Using the example above) now let's say after your broker had taken those data points on every single house there is that you only want a CMA done using 3/2s with a two-car garage. Your broker would likely need a few more weeks to manually check off which houses he had already looked at that met that criteria and then re-calculate the value. An algorithm can make this change in seconds.

Cons:

--An algorithm lacks the "finer touch" of a human being. The computer can parse more information, but it can only really look at what data is readily available to it (namely macro level trends). A computer can't parse out things like condition (house A has granite countertops while house B has formica), location (none of the neighbors on house A take care of their lawns, but all of the neighbors for house B take perfect care of their homes), or even "personal details" ("the seller is desperate to move and will probably sell for under asking if you close immediately"). That's the value of a human touch.

--An algorithm will generally try to fit a distribution as a whole rather than predict each of its parts individually. Say there are 100 houses and you plotted them all on a graph. In regression, you would try to find the continuous line that goes through all 100 points while minimizing total error from the line to each point (hence why in regression it's called the "line of best fit"). The idea is to find a line that could scale to best predict future houses that would join that distribution. 

If an algorithm tried to make a prediction by focusing on each data point individually, then it would come up with a model that is far less complex, more biased, and won't scale well to future data points. This is known as overfitting the line. For Zestimate, it'll predict houses it already knows about perfectly but won't be able to predict unknown houses very well.

This picture is kind of what I'm talking about. The one on the left is more like the Zestimate and the one on the right is the biased, overfit model...

So the Zestimate is not designed to be a spot-on estimate of your house. It's designed to predict with relative certainty based on the distribution of every house in your market.  In other words, the Zestimate is designed to predict every house "pretty good" as opposed to predicting one house "perfect". In that sense, the Zestimate is very good at predicting values (and from an innovation perspective, downright extraordinary). However, it's more of a jumping off point than anything. You should use it as a starting point in your decision making process, but you should not use it to base buying decisions off of without a closer look of your own.

 To piggyback off John's questions a little, make sure you follow up by asking about details. Ask them specifically how the flip they said they did went for them (how long did it take? what did you learn from it?). Ask about the performance of the rental property (how they bought it? do you use a property manager? what kind of tenants did you end up with?). Don't just ask yes or no questions that anybody could easily lie about. Make sure they actually walked the walk.

Don't be afraid to leave money in the deal. If after a refinance you had to leave $10,000 of your money in the deal but it conservatively cash flowed for $2000/year, then is a 20% return really not worth your time? The latter is a far more common scenario than the "BRRR -- Buy houses for free -- CLICK HERE" examples that most of the blogs and books talk about.

I finished my MBA 2 years ago. It undoubtedly opened many new doors in my career that would never have been opened with just a Bachelors. It was a bit of a gamble up front because I paid for it myself entirely while pretty much all of my classmates had tuition reimbursement through their employers (for instance, a few of my classmates worked for Disney,  who pay their full-time employees 75% of their education costs. That's just one company though. Amazon pays 95%, even for their lowest level employees). 

While I paid a pretty big cost up front on tuition, it did bring my wages up considerably (and I expect that gap to continue to widen even more over the next few years than if I had stuck with the path of just my Bachelors). In many ways, this payback period is not all that unlike buying a rental property. Think of it like investing in yourself rather than investing in a house, but the upside is far more uncapped (you can only raise the rent so much, but career growth at any given time can theoretically go parabolic).

My suggestion to you would be to look for the route my classmates took. Finish your undergrad, apply to employers that have a tuition reimbursement program (hint: big corporations generally have these systems in place), work for a year or whatever the qualification period is for those programs, and then apply to dental school. You might need to be a little strategic about who you work for because ideally you'll need a somewhat flexible schedule to make both work. It'll take a little longer to finish, but at least your employer will foot most of the bill for you.

Why not do both? 

Go to school part-time and do real estate part time (you probably won't have a full week's worth of work to do when you're starting out anyway). Go to your local community college and take 9 credits per semester (most incoming freshmen take 18, so you'd only be going half the time). Community college is a great cheat code for obtaining a degree because most of the credits will transfer to your state schools and yet the credits cost a fraction at CC. So it's a pretty affordable option to start earning college credits, and it would buy you 2 years as a real estate agent to determine if it's the path you truly want to go down.

I have what I would consider to be a vanity name for my LLC. I'd probably change the name to something "neutral" if I could, but it shows you can name it whatever you want.

The one warning I would give is not to have an LLC name that involves "money", "investments", "capital", etc if you plan to be tenant-facing with it. I might just be making this narrative up in my own head, but I feel it would be kind of awkward for the tenant to respect me as a landlord while at the same time having to write rent checks to an entity that makes overt mentions of being funded by said rent checks.

Random person: **Googles "How do I invest in real estate when I have no money to invest"**

Google: "Become a wholesaler"

Random Person: **Makes new thread on BiggerPockets: "How do I find cash buyers for my wholesaling business?"**

BP Investors: "We have money. Please add us to your list."

Random Person: **Makes new thread on BiggerPockets: "Why doesn't anyone want to buy any of my wholesale deals?"**

BP Investors: "Are you sure your deals are really deals?"

Random Person: **Googles "How do I find real job"**

Repeat.

Does anybody rent a room in their own house that they currently live in? I don't mean a separate rental you own. I mean your current primary residence. 

I just bought a new 3/2 primary. I'm a single guy, so I basically have a guest bed and guest bath that are going to be fully furnished yet never get used otherwise. I do work during the week and I'm not sure about having strangers in my house when I'm not home, but I was considering the idea of trying to open up the room for weekends and holidays when I'd be home. 

Any pros/cons I should consider from people who have done this? Was it weird to have other people in your house while you are home (not home?)?

I haven't tried it in awhile so I can't confirm it still works, but there is a R package called realtR that is used to read listings on Realtor.com that I have used before. Not sure it is what you are looking for but it might be worth reading the documentation.

http://asbcllc.com/r_packages/realtR/2018/introduction/index.html