Before class

Go to ledatascifi.github.io and then to chapter 1.2

1. Make sure you have a github account

2. Have git and github desktop on your computer

Hello everyone!

Today, lets

1. Finish all of your CAPM projects for all your finance classes for the rest of your college life in 3 minutes
2. Do an auction to learn how to save $381m per quarter
3. Learn how we're all connected
4. Get our hands dirty!

Ambitious, but fun; This class in miniature!

Motivation

Where $V$ is the NPV of your education, $B$ is biz/econ skills, and $T$ is the technical skills covered in this class:

1. $\delta V/\delta B > 0$ (good choice in your major!)
2. $\delta V/\delta T > 0$, and $\delta T$ is achievable
3. $\delta^2 V/\delta T \delta B > 0$ is important but overlooked

On #2: Who has estimated beta for a stock in FIN323, FIN328, or FIN335?

On #3: Let's play a game!

• Raise your hand if you've ever bid in an auction

Terms of the auction

1. I will give a $20 bill to the winner
2. Bids in $1 increments
3. The winner pays what they bid, and the 2nd place player pays what they bid

What did you learn from that auction?

There are only 2 ways to avoid "death spiral":

1. The Star Trek solve: Turn win-loss into no-win: split the $18 gain from incrementing the bid
2. The War Games solve: "The only winning move is not to play"

The two solutions have commonalities:

1. Plan before doing: "Measure twice, cut once"
   • Think about big pic! Especially important on coding tasks where it's easy to get lost in minutiae...
   • What's the economics of the situation?
2. The problem dictates the tool! I'm very excited about skills you will learn in this class... but when you have a hammer everything looks like a nail!
   • Here, valuation isn't the tool you need! It's game theory.

"That's silly!" Fine.. let's look at a real auction

Well... thousands of them. Enter: Zillow.

1. Zillow's plan: Predict price Zillow can sell the house for, offer to buy for less, <do stuff quickly>, sell
2. Profit hopes: liquidity provision, price appreciation
3. Economics:
   • Have to outbid other buyers (Winner's curse)
   • Home owner knows most about the house! (Asym. Info)

Outcome of Zillow's plan

Outcome of Zillow's plan

Nov 2, 2021: Zillow (articles here and here)

• Lost $381m in 3Q alone flipping homes because algo was "unable to predict prices"
• Cuts 25% of workers
• Loses about $10B market cap in the week

Good valuation wasn't a good enough tool

Let's define a model, and let Zillow's valuation be 2x better than homeowners:

1. True value of a house is $v$, with $v \sim U[250,000, 2,500,000]$
2. Owner's guess $v*(1+e_o)$, with $e_o \sim N(0,10\%)$
3. Zillow's algo predicts $v*(1+e_z)$, with $e_z \sim N(0,5\%)$
4. Because of winner's curse and the transaction cost owners face moving, Zillow is only able to buy houses when it offers more than 20% above what the owner thinks the house is worth. Talk about adverse selection.
5. If Zillow buys the house, they subsequently sell the house at the true value.

What happens in this model?

So, if Zillow looks at 100k houses with values uniformly ranging from 0.25m-2.5m, and the above steps play out...

(You can figure this out with under 30 lines of python)

It buys 6k houses and loses $300m!

"But hindsight is 20/20!"

Except... we didn't need hindsight. An NBER working paper from 2020 discussed the issues this business model faces

• Adv. selection is key, limits scope for profits
• Profits plummet when liquidity is low (downturns)
• Better valuation models don't help!
• Winner's curse increases with competition
• Front running: Zillow's valuations are public knowledge!

What do we learn?

Recall, my three initial assertions:

1. $\delta V/\delta B > 0$ (good choice in your major!)
2. $\delta V/\delta T > 0$, and $\delta T$ is achievable
3. $\delta^2 V/\delta T \delta B > 0$ is important but overlooked

The Zillow case study illustrates the last one.

The Zillow team got obsessed with the technical skills and how good their pricing model could be. A little attention to the business forces could have saved them a lot of money!

The goal in this class is to both

1. add some powerful skills to the bottom of your resume ($T \uparrow$)
2. think about how business settings impact the use of those skills ($T*B \uparrow$)

So that you get the good jobs, but avoid losing $381m!

Onboarding

If that sounds good (great?), let's continue:

1. What are we working towards?
2. Icebreaker - chain
3. Syllabus, norms, and expectations
4. Getting going

What are we working towards?

Skills you can benefit from over your career, yes.

A grade? Sure!

But that's not what I'm talking about...

The penultimate assignment is about glory

The LeDataSciFi LeContest

The winners get "etched" into history.

About me