Bill Faruki interviewed by Nigel Woodward, Global Director of Financial Services, Enterprise Solution Sales, Intel Corporation
http://www.hyperrig.net/pdf/bkgTechTradeLCJune09.pdf
Neotick is your bleeding edge technology partner that boasts the most advanced technical capabilities, subject matter expertise and solutions customized to give you unparalleled edge in today’s world and the future of trading.
Blogged by Bill Faruki CTO, Neotick, Inc - bfaruki@neotick.com - www.neotick.com - +1 312 884 7604
Thursday, April 14, 2011
Wednesday, April 13, 2011
High Frequency Trading & Algorithmic Trading
High Frequency Trading has been hitting the headlines more and more over recent months. Why? Because it involves huge sums of money and more than a little intrigue.
It first came to the attention of the public when an ex-Goldman Sachs trader was accused of stealing the firm’s proprietary code for an algorithmic trading system, a system that was able to generate multiple small profits by systematically trading electronic markets at lightning speed, in effect a goose that could potentially lay golden eggs day in and day out.
When high frequency trading came to light amongst the chattering classes, it immediately caused a furore. Accusations of market manipulation and even illegal practice flew left, right and centre. Why, asked the more vocal members of the investing public, should we be disadvantaged just because our brokers and fund managers don’t have the same sophisticated trading algorithms and market access speed as the big proprietary trading firms?
These sorts of questions are being increasingly asked and they don’t seem to be going away. As a result, a number of proposals are now on the table to see if the playing field can be leveled to assuage the clamour of the investing public.
Exposing the Myths
However, there are many misconceptions and misunderstandings about high frequency trading. The tendency amongst many is to take a somewhat simplistic view, a view that high frequency trading is nothing more than a sophisticated method of “front running” the markets. The purpose of this article is to explode some of the myths and to present a more balanced viewpoint by exploring the topic of high frequency trading in more detail, investigating what exactly it is and how it works.
The Algorithms Behind High Frequency Trading
Behind any high frequency trading strategy is a set of algorithms, usually developed and programmed by highly paid computer boffins at proprietary trading firms. The terms algorithmic trading and high frequency trading have now become somewhat synonymous, although they do not actually mean the same thing.
These algorithms used for high frequency trading take into account a range of criteria to determine when and where to trade, including: the price of a security; the size or liquidity available; various timing considerations (e.g. how quickly can an order be executed, or when exactly should orders be placed to ensure the strongest chance of execution); how likely an order is to be filled (the “fill ratio”); and the overall costs of each transaction.
All of these factors are taken into consideration when firms are developing and deploying trading algorithms and each individual factor can have a significant impact on the overall profitability of a firm’s high frequency trading strategies.
Price and Size
The factors of price and size are closely linked, because together they determine liquidity. Price is obviously one of the most important criteria because it is always advantageous to execute at a favorable price. But if you are only able to execute a small amount at that price (i.e. if the liquidity is not there) then any potential profits will be limited.
Time
For most algorithmic trading strategies, time is also an key factor. Speed of execution is for example hugely important with high frequency trading, because the faster you can get your order into the market, the more chance there is you will be able to hit a quoted price. But algorithm timing itself is also an important factor. Many algorithms take into account the timing of when big buy side orders are likely to be in the market (many buy side order are executed towards the end of the day for example).
Accordingly, a number of algorithms incorporate a TWAP (Time Weighted Average Price) component, to calculate the average price of execution across a specified time range (e.g. the last hour of trading).
Fill Ratio
The fill ratio is the percentage likelihood that an order will be executed. Many algorithms use statistical analysis and historical benchmarks to calculate the fill ratio percentage and some algorithms will even guarantee that their fill ratio will be above a certain level (e.g. 50%).
Transaction Costs
The full costs of performing a transaction also have to be taken into consideration with any algorithmic or high frequency trading strategy, because if those costs are more than the profit per transaction, then the algorithm will be unsuccessful (either that or it will need to be adjusted or fine-tuned). These transaction costs take into account not only transparent items such as fees, commissions and taxes, but also estimated “latent” costs such as opportunity costs, price movement during execution, fluid bid/ask spreads and a multitude of other intangibles and “moving targets”.
Aggression & Stealth
The algorithmic trading program itself takes into account the above five criteria and determines the execution criteria, whether passive or aggressive, and whether any kind of order slicing or waving is to be used.
The more aggressive strategies will generate market orders to hit existing bids or offers, enabling them to absorb liquidity at a given price. Passive strategies on the other hand, will generate limit orders either joining the current bid or offer, or a number of ticks below or above the current spread.
Slicing and waving is the practice of splitting a large order up into smaller chunks and releasing those chunks into the market at specified intervals or “waves”. If those waves have any kind of random element, they can allow the trader to operate in a kind of stealth mode, releasing the order into the market without tipping his hand to other traders or investors, who might be watching for the large orders coming through.
Conclusion
All of the above elements are key to running successful algorithmic trading strategies. And if these strategies are being run at high frequency (i.e. multiple orders per second), additional factors around speed and latency also have to be taken into consideration, from both a pure computer/data processing and a market connectivity standpoint.
High frequency trading firms invest huge sums not only in developing their algorithms, but also in putting in place the necessary infrastructure to ensure speed of execution. But as it only takes one rogue algorithm to reverse all those profits, it is a risky game.
Neotick is your bleeding edge technology partner that boasts the most advanced technical capabilities, subject matter expertise and solutions customized to give you unparalleled edge in today’s world and the future of trading.
Blogged by Bill Faruki CTO, Neotick, Inc - bfaruki@neotick.com - www.neotick.com - +1 312 884 7604
It first came to the attention of the public when an ex-Goldman Sachs trader was accused of stealing the firm’s proprietary code for an algorithmic trading system, a system that was able to generate multiple small profits by systematically trading electronic markets at lightning speed, in effect a goose that could potentially lay golden eggs day in and day out.
When high frequency trading came to light amongst the chattering classes, it immediately caused a furore. Accusations of market manipulation and even illegal practice flew left, right and centre. Why, asked the more vocal members of the investing public, should we be disadvantaged just because our brokers and fund managers don’t have the same sophisticated trading algorithms and market access speed as the big proprietary trading firms?
These sorts of questions are being increasingly asked and they don’t seem to be going away. As a result, a number of proposals are now on the table to see if the playing field can be leveled to assuage the clamour of the investing public.
Exposing the Myths
However, there are many misconceptions and misunderstandings about high frequency trading. The tendency amongst many is to take a somewhat simplistic view, a view that high frequency trading is nothing more than a sophisticated method of “front running” the markets. The purpose of this article is to explode some of the myths and to present a more balanced viewpoint by exploring the topic of high frequency trading in more detail, investigating what exactly it is and how it works.
The Algorithms Behind High Frequency Trading
Behind any high frequency trading strategy is a set of algorithms, usually developed and programmed by highly paid computer boffins at proprietary trading firms. The terms algorithmic trading and high frequency trading have now become somewhat synonymous, although they do not actually mean the same thing.
These algorithms used for high frequency trading take into account a range of criteria to determine when and where to trade, including: the price of a security; the size or liquidity available; various timing considerations (e.g. how quickly can an order be executed, or when exactly should orders be placed to ensure the strongest chance of execution); how likely an order is to be filled (the “fill ratio”); and the overall costs of each transaction.
All of these factors are taken into consideration when firms are developing and deploying trading algorithms and each individual factor can have a significant impact on the overall profitability of a firm’s high frequency trading strategies.
Price and Size
The factors of price and size are closely linked, because together they determine liquidity. Price is obviously one of the most important criteria because it is always advantageous to execute at a favorable price. But if you are only able to execute a small amount at that price (i.e. if the liquidity is not there) then any potential profits will be limited.
Time
For most algorithmic trading strategies, time is also an key factor. Speed of execution is for example hugely important with high frequency trading, because the faster you can get your order into the market, the more chance there is you will be able to hit a quoted price. But algorithm timing itself is also an important factor. Many algorithms take into account the timing of when big buy side orders are likely to be in the market (many buy side order are executed towards the end of the day for example).
Accordingly, a number of algorithms incorporate a TWAP (Time Weighted Average Price) component, to calculate the average price of execution across a specified time range (e.g. the last hour of trading).
Fill Ratio
The fill ratio is the percentage likelihood that an order will be executed. Many algorithms use statistical analysis and historical benchmarks to calculate the fill ratio percentage and some algorithms will even guarantee that their fill ratio will be above a certain level (e.g. 50%).
Transaction Costs
The full costs of performing a transaction also have to be taken into consideration with any algorithmic or high frequency trading strategy, because if those costs are more than the profit per transaction, then the algorithm will be unsuccessful (either that or it will need to be adjusted or fine-tuned). These transaction costs take into account not only transparent items such as fees, commissions and taxes, but also estimated “latent” costs such as opportunity costs, price movement during execution, fluid bid/ask spreads and a multitude of other intangibles and “moving targets”.
Aggression & Stealth
The algorithmic trading program itself takes into account the above five criteria and determines the execution criteria, whether passive or aggressive, and whether any kind of order slicing or waving is to be used.
The more aggressive strategies will generate market orders to hit existing bids or offers, enabling them to absorb liquidity at a given price. Passive strategies on the other hand, will generate limit orders either joining the current bid or offer, or a number of ticks below or above the current spread.
Slicing and waving is the practice of splitting a large order up into smaller chunks and releasing those chunks into the market at specified intervals or “waves”. If those waves have any kind of random element, they can allow the trader to operate in a kind of stealth mode, releasing the order into the market without tipping his hand to other traders or investors, who might be watching for the large orders coming through.
Conclusion
All of the above elements are key to running successful algorithmic trading strategies. And if these strategies are being run at high frequency (i.e. multiple orders per second), additional factors around speed and latency also have to be taken into consideration, from both a pure computer/data processing and a market connectivity standpoint.
High frequency trading firms invest huge sums not only in developing their algorithms, but also in putting in place the necessary infrastructure to ensure speed of execution. But as it only takes one rogue algorithm to reverse all those profits, it is a risky game.
Neotick is your bleeding edge technology partner that boasts the most advanced technical capabilities, subject matter expertise and solutions customized to give you unparalleled edge in today’s world and the future of trading.
Blogged by Bill Faruki CTO, Neotick, Inc - bfaruki@neotick.com - www.neotick.com - +1 312 884 7604
Subscribe to:
Comments (Atom)