week 10 Flashcards
o Algorithm Trading
The use of computer algorithms to automatically make certain trading decisions, submit orders, and manage those orders are submission
o Represents the computerized executions of financial instruments, including stocks, bonds, currencies, financial derivatives etc.
o An algo is basically a set of instructions for accomplishing a given task
o A trading algo simply defines the steps required to execute an order in a specific way to achieve some specific goals
o Math or beat a specific benchmark
o Minimize overall transaction costs
o Seek to trade more opportunistically
o A set of instructions for accomplishing a given task, there is a wide range of algorithms each with distinct goals
Why do we need AT?
o Main aim is to do a good job in the implementation stage of investment cycle
o Helps to control our emotional/behavioural biases
Impact-driven Algo- TWAP
o Time weighted average price algo
o An average price based algo
o The benchmark is an average price, which reflects how the asset’s market price has changed over time
o Based on a uniform time-based schedule
o Unaffected by any other factors, such as market price or volume
o Uniform vs randomized TWAP
Randomised approach allows more flexibility in the trading strategy, although it can increase the risk of missing the TWAP benchmark
Impact-driven Algo- TWAP ISSUES
Subject to signaling risk, too uniform
Poor execution due to not considering of market conditions (price change, liquidity change)
Solved by randomized TWAP order
• Need to calculate the cumulative execution rate and compare it to the randomized TWAP orders
• Does not solve poor execution but solves some signaling risk
Impact-driven algo- VWAP
o Volume weighted average price
o An average price based algo
o Benchmark for a given time span is the VWAP
o Approximating the TWAP is simply a matter of trading regularly throughout the day, VWAP needs to trade in the correct proportions, which are based on the day’s trading volume but is unknown beforehand
o Common solution
Use historical volume profiles which assumes that the day’s trading volume follows a similar pattern to the historical profile
If the historical profile is based on sufficient data, this may be a reasonable assumption for many liquid assets
Need a large amount of data
o Not affected by market trading
o Resolves the issue of liquidity but does not resolve the issue of unfair price change
o What if the market is highly fragmented?
Depends where you trade, so if you want to trade a lot -> go to ASX for example and then you look at the historical data for that
Impact Driven algo – POV
o Percent of Volume
o This algo tries to ‘go along’ with the actual market volume
o Need to pre-determine a participation rate e.g. if a million shares of ABC trade in a day, then a POV algo tracking a 20% participation rate should have executed 200k of those
o Unlike TWAP and VWAP, where a trading schedule may be pre-determined, for POV algo the trading schedule is dynamically determined
o Does not help predict the volume
Always need to be watching, therefore you will either a) go ahead of target b) fall behind target
Impact driven Algo- POV explanation
o In the table, the market volume is 3k in this 15minute window -> we would like to trade 600
o By 8;23, when 1700 of ABC has already traded we would like to have seen 20% of this, or 340
o Rather than respond to every single trade, we can set trigger points -> we catch up at 825 by filling an order for 400
o Need to adjust our participation rate slightly to account for our own trading
o Actual participation rate = 1/(1-Participation rate) -1
Choosing among trading algos
o What factors affect your choice?
Intended benchmark (arrival price? Close price?)
• Minimise transaction cost
Level of risk aversion
• Want to minimize cost and risk
• If risk averse, they tend to pay more to give up risk (certainty), purple line
o Higher risk aversion level
• Balance between expected cost and timing risk
Desired trading goals (min cost or min risk of trade-off?)
