Exam 2 Flashcards
Why did it become a good investment to bet against mortgage-backed securities
The default rate on the mortgages kept rising
AND
Banks were incentivized to issue more and more mortgages
Bollinger Bands is a ___ indicator
Lagging
Who was the first bank Michael approached to make him the CDSs?
Goldman Sachs
If the Information Coefficient is doubled and the trading opportunities are multiplied by 9, what happens to the Information Ratio
IR is multiplied by 6
Equation for fundamental law of portfolio management
IR = IC * sqrt(BR)
What does CDS stand for?
Credit Default Swap
What did CDS’s do in The Big Short?
Insured the bonds against failure and the insurer paid the claim if the underlying bonds fail
Following statement is true or false: Hedge Funds cannot succeed if EMH applies.
False
In Reinforcement Learning, the default MDP has an assumption of infinite horizons to overcome that, we introduce a concept of ___________ rewards. Multiplying the reward by λ raised to t. Where λ’s limits are ___< λ <=___
discounted, 0 , 1
What is the meaning of “synthetic CDO” in “the big short” movie?
A CDO that contains credit default swaps
Dyna is
a blend of model-free and model-based methods
Which attribute of Q learning is important in convergence over infinite horizon?
gamma
According to the movie “The Big Short”, which of the following is closest in meaning to a “Credit Default Swap (CDS)”?
Going short on the default of mortgage-backed securities
Given 2 companies have same Information Ratio Company A has algorithm 100 time smarter than Company B Company A trades for 20 days a year How many trades does company B need to execute?
200000
According to Bloomberg News, what opportunity similar to CDO which several large banks have started selling since 2015?
Bespoke Tranche Opportunity
Transition function is a three dimensional object given by T[s,a,s’]. Suppose we are in state “s” and take particular action “a” then the sum of all the next states “s’(s prime)” we might end up in should sum to be __
Always 1
In order to reduce risk while maximizing returns, portfolios often allocate stocks based upon _____ correlation in the short term and _____ correlation in the long term.
negative, positive
For Grinold’s Fundamental Law, what is Information Ratio (performance) similar to?
Sharpe Ratio
If a stock is said to have good relative strength, it means:
The ratio of the price of the stock to a given market index has trended upwards
Which stock properties would best be considered to comprise an optimal portfolio with higher combined returns and lower risk than either of the individual stocks.
Stocks ABC and DEF with covariance value of -0.9.
What is an issue when we buy or sell stocks with most significant price changes?
Certainty of price change
What is information ratio?
Sharpe Ratio of the excess return
True or false, AAA tranches have higher risks than BBB tranches.
False
Assume beta represents the market return and alpha represents the residual return. Which answer below exactly contains the factors that Information Ratio is related to?
mean(alpha), std(alpha)
According to the movie, what happened in the second quarter of ‘07 that caused mortgage defaults to skyrocket?
The adjustable rates on the mortgages kicked in
What is NOT a valid way of solving a reinforcement learning problem?
Transition Iteration
True or false, the EMH thinks the following: The market can be beaten by efficient trading practices
True
using Mean Variance Optimization technique, what are the inputs required
Expected Return, Volatility, Covariance, Target return
Stock options can increase the ___________ of a portfolio
Leverage (as the investor can keep more cash on hand up to the expiration date)
What is the root cause of 2008 financial crisis?
hedge fund trading with derivatives
The end goal of the Markov decision problem is to find a policy that ____ total rewards
maximizes
The lowest possible Sharpe ratio of a portfolio that contains at least two stocks is
lower than the lowest Sharpe ratio of any one stock in the
portfolio
How will information ratio change if adding additional cash to the portfolio?
The information ratio will decrease.
Stock A and stock B each gives a 12% annual return. Daily returns standard deviation of stock A is 0.0027 and stock B is 0.0031. If you have $100 which one of below 4 option s would be the best investment strategy to invest the $100?
$100 in Stock A
Instruments
price of AAPL stock, which takes $300 as the premium. The stock price
moved from $115 to $120. What is the maximum return available?
9700
Commodities
has options on besides stocks
Can a portfolio lie above the efficient frontier?
No.
The efficient frontier represents the lowest risk portfolio for a given level of return.
The assumption that prices will adjust rapidly to new public information represents which form of the Efficient Markets Hypothesis?
Semi-strong
What is risk?
standard deviation of historical daily return
Which schema is model free?
Q-learning
In Q-learning, what do s, s-prime, a, and r stand for?
s = initial state, a = action, s’ = new state, r = reward
Information Ratio IR is the Sharpe Ratio of:
Excess Daily Return
For the update rule of Q table, a higher value of alpha and a higher value of gamma indicate which of the following?
The previous value of Q is preserved less; The learning process is more quickly; The later rewards are valued more
Which form of the Efficient Market Hypothesis states that stock prices are affected by ONLY the past price information.
Weak
What dies pi represent in Markov Decision Processes?
The current estimate of the policy
The information coefficient (IC) of a manager is 0.2 . The manager makes monthly bets on 27 stocks. What is the information ratio(IR) (or performance) of the manager?
3.6
What portfolios lie on the efficient frontier?
The lowest risk portfolios for each level of return
Grinold’s Law formula
performance = skill * sqrt(breadth)
Which form of the Efficient Market Hypothesis prevents profiting from Technical Analysis?
Weak & Semi-Strong & Strong
Among four strategies of options, which one’s maximum loss is theoretically unlimited? Therefore, we should manage this strategy very carefully.
Write Call
Where did Michael Burry work?
Scion Capital
In Reinforcement Learning, which can either be a STATE or a REWARD?
Daily Returns
What is an example of exogenous data for an airline stock?
The price of oil
Purpose of the efficient frontier
show the range of least to most efficient portfolios
The efficient frontier contains ___
the portfolios with the lowest possible risk given a
particular amount of return.
Risk Adjusted Reward
AKA Sharpe Ratio = Expected return divided by standard deviation. For single-table bet example it’s, 20/31.62 = 0.63
If you take this and multiply it by square root of 1,000, you get the Sharpe Ratio of the 1,000-table bet mathematically. So IR = IC * Sqrt(Breadth)
Three lessons re: Sharpe ratio
- One, higher alpha generates a higher sharpe ratio.
* Two, more execution opportunities provides a higher sharpe ratio.
* And three, sharpe ratio grows as the square root of breadth.
IR is like Sharpe ratio of ___
excess returns
Skill is represented by ____
Alpha
Simons and Buffet have same IR. Simons algorithm is 1/1,000th as smart as Buffet’s. Buffet trades 120 times per year. How many trades must Simons execute?
IC_buffet * sqrt(120) = IC_simons * sqrt(x)
IC_buffet * sqrt(120) = IC_buffet/1000 * sqrt(x)
1000 * sqrt(120) = sqrt(x)
1000^2 * 120 = x
120,000,000 = x
What is risk?
Volatility (standard deviation) of historical daily returns
MVO inputs
- expected return
- volatility
- covariance (a matrix between the asset and all others)
- target return (anywhere from the max return asset to the min return)
MVO output
-set of weights for portfolio that minimizes risk and achieves the target return
For any return level, there is ___
An optimal portfolio
What does the efficient frontier mean?
It means that there are essentially no portfolios outside of it, and that any portfolio under it is suboptimal in some way (either higher risk or lower reward than could be achieved with different weighting)
In practice how is the efficient frontier used
Mainly to visualize where their portfolio falls
If you draw a tangent from the origin to the efficient frontier, where it hits is ____
the max Sharpe ratio for that portfolio
In RL, if gamma == 1, it’s ___
infinite horizon. Closer gamma is to 1, the more we value future rewards. The closer it is to 0, the less we value them.
What are S, A, T, and R in Markov Decision problems?
S is the potential states, A are the potential actions, T is a transition probability, R is the reward function.
What is the goal of RL?
The goal for reinforcement learning algorithm is to find a policy, pi, that maps a state to an action that we should take, and its goal is to find this pi such that it maximizes some future sum of the reward
Q[s, a] = ?
immediate reward + discounted reward
pi(s) = ?
argmax_a(Q[s, a])
pi*(s) = ?
Q*[s, a]
Advantage of model-free approach like Q
can easily be applied to domains where all states and/or transitions are not fully defined. As a result, we do not need additional data structures to store transitions or rewards.
Why use Dyna in trading?
Using just Q doesn’t allow you to take a random action to learn a good strategy because you lose money! Need to simulate.
Three components we add to Q-learner for Dyna
The three components are, we update models of T and R, then we hallucinate an experience
Options: Call
The right to buy
Options: Put
Option to sell