Computational Investing Part I (Coursera) - Week 4

Nov 27, 2013 ../computational-investing-part-1    MOOC  Investing

Week 4

1.1

  • Sound was terrible
  • Ray Charles got a big annoying
  • Quantitive investment is based on 'arbitrage' model
    • Discrepency between price and value
  • Calculating value
    • Technical analysis - pricing and volume
    • Fundamental analysis
      • financial statements
      • P/E ratios
  • Where information comes from:
    • Price/Volume: the markets
    • Fundamentals: SEC fillings

1.2

  • Efficient markets hypothesis
    • Weak version
      • future prices cannot be predicted by analysing prices from the past
    • Semi-strong
      • no room for arbitrage
    • Strong
      • market is efficient.
      • Prices reflect hidden infroamtion
  • Some evidence that it's not true
  • Behaviour economies is an argument against EMH

1.3

  • Again, terrible sound
  • Study of how positive and negative event affected prices of related stocks
  • Drift up and down for bad news before the event
    • Why? Leakage can occur.
    • Bad news can just suggest a bad stock and vice versa
  • Upcoming assignment talking about running and event study
  • Align stocks so Day 0 is the same for all of them (??)

2.1

  • Terrible sound - microphone was way too loud
  • Understand "risk", correlation and covariance, mean variance optimization and efficient frontier
  • Portfolio optimizer
    • Given: set of equities and target return, what's the optimal portfolio?
    • Find: allocation to each equity that minimizes risk
  • Risk often refers to how "volatile" their stock is. Volatility is calculated by standard deviation
  • Ideal portfolio is high return and low risk but that's rare. You want a healthy combination - each are a trade off.
  • Harry Markowitz developed "mean variance optimisation" theory.

2.2 - Inputs and Outputs of a Portfolio Optimizer

  • Portfolio optimizer balances return and risk. Exploits information about it.
  • Inputs
    • Expected return for each equity
    • Volatility (risk) for each equity
    • Target return
    • Covariance matrix -- no idea what this is?
  • Output -- optimal portfolio
Opinion
  • Can't help but feel that a portfolio optimizer is a silly idea. You make returns based on future earnings which you can't really "optimise" for.

2.3

  • How can we have a portfolio with lower risk than individual equities?
  • Higher weight to "lower risk" stocks. Look at covariance and anti correlated stocks.

2.4

  • Efficient frontier -- Lower risk by combining anti-correlated equities
  • How to combine them

2.5

  • How Optimizer Works
    • Define variables
      • Things optimiser can "tweak
      • Vary how much to allocate to equities (weights)
    • Define constraints
      • Sum of all weights must add up to 1
      • No less than 10% in a certain equity
    • Define optimization criteria
  • Optimizer algorithm
    1. Tweak weights
    2. Check constraints
    3. OK?
    4. Call function
    5. Repeat
  • Could be a giant for loop that brute forces the thing.
  • QSTK uses an optimizer called CVXOPT
3.1
  • Overview of what event studies are
  • Slow movies. Lot's of static
3.2
  • Found it a lot easier to understand most of the lesson after completing the course work. I guess that's standard.
  • When a stock drops below $5 it changes a lot of things
    • Not listed on major indicies

Assignment

  • Fairly easy
  • Most of the data was already provided