High-Frequency Trading (HFT) is essentially trading that takes place on incredibly small time-scales involving algorithms that place orders into the market to take advantage of small price discrepancies between assets between exchanges.
In order to put the speed that trading takes place into context we have to begin with a breakdown of the "second" (our base unit of time):
| Fraction | The Number of Seconds | Prefix | Symbol |
| unit | 1 | ||
| tenth | 0.1 | deci | d |
| hundredth | 0.01 | centi | c |
| thousandth | 0.001 | milli | m |
| millionth | 0.000 001 | micro | µ |
| billionth | 0.000 000 001 | nano | n |
| trillionth | 0.000 000 000 001 | pico | p |
Generally, HFT deals with four fractional units of the second: seconds, milliseconds, microseconds, and nanoseconds.
Contextually, blinking takes around 300 milliseconds, but to further put these units of time in perspective, let's pretend that we want to trade between the Chicago Mercantile Exchange (CME) and the New York Stock Exchange (NYSE) - although the "trading" actually takes place in data centers in New Jersey, not New York:
The time it takes for a signal to travel down the fiberoptic cable from Chicago to New Jersey and back currently takes around 12.5 milliseconds - 4.1% of the time it takes you to blink.
In the early days of HFT it used to take twice as long to do this, but new fiberoptic cable networks in combination with firms who dedicate themselves to HFT infrastructure has dramatically sped-up this process.
Statistics:
- In 2012, according to a study by the TABB Group, HFT accounted for more than 60 percent of all futures market volume in 2012 on U.S. exchanges.
- In the United States in 2009, high-frequency trading firms represented 2% of the approximately 20,000 firms operating today, but accounted for 73% of all equity orders volume.
- The Bank of England estimates similar percentages for the 2010 US market share, also suggesting that in Europe HFT accounts for about 40% of equity orders volume and for Asia about 5-10%, with potential for rapid growth.
The Role of High-Frequency Trading:
Like any trading strategy, the ultimate purpose of HFT is to make money. This process however can be loosely split down into two categories:
- Market making and liquidity provision.
- Proprietary trading strategies.
Market Making and Liquidity Provision:
When there's an industry wide hiccup and people question the purpose of HFT, the industry responds with this as their go-to argument for their relevance. Without branching off on a huge tangent, market making is the process of creating an environment where trades can be facilitated in the exchange.
This process is called "Liquidity Provision" you are making a liquid market where trading is easy for investors.
e.g. If I want to buy 50,000 shares in Tesco, I need to get them from somewhere (they don't just appear in the exchange), so market makers will create an order book with prices at which they will buy Tesco shares and prices at which they'll sell Tesco shares. Thus allowing me to trade Tesco shares.
Now would be a good time to add that "trading volume" and "liquidity provision" are not the same thing. From an ethical perspective, I feel that if the HFT form of trading is meant to truly provide liquidity in the primary sense of the word (market making is synonymous known as primary trading), then there ought to be a level of market risk undertaken by the market maker/HFT firm. Market makers traditionally lose out on between 50-70% of the deals they make with clients, but will sacrifice these losses to take advantage of the order-flow they can see as a result on the proprietary side of their bank - if you undertake no risk as a result of having system advantages then there is clearly sign that the market is not fairly pricing deals at the exchange.
It's worth noting before I go on that stock exchanges may pay, or charge people to trade at the exchange. HTF firms will therefore usually opt for exchanges where a liquidity rebate is provided for market makers - where they get pad to make markets there.
It's also notable that electronic trading has brought the costs of trading down significantly over the last twenty years (mainly through drops in bid/ask spreads), but it's valuable to acknowledge that electronic trading and HFT are not the same thing.
Proprietary Trading Strategies:
When you hear the term "proprietary trading", this basically translates as "taking risk with a firm's money to make more money".
e.g. If i buy SAB Miller shares, because I think they'll go up tomorrow, this would be classed as a form of proprietary trading strategy.
In the context of HFT, proprietary trading strategies can take many forms, with the following three being a couple of examples:
- Arbitrage
- Front Running
- Predictive strategies
Arbitrage:
This strategy involves taking advantage of mis-pricing of assets in two different places (in the US there are currently 18 different exchanges - six more if you include futures-only exchanges). This can take place through several strategies including, but not limited to:
- Event Arbitrage - certain events reoccur and these can be taken advantage of.
- Statistical Arbitrage - exploiting relationships between two linked products (e.g. a currency pair and a forward contract on the same currency pair).
Front Running:
I'm going to cover this in more detail later, but this basically the idea that HFT firms can see what orders you're going to place in the market before they reach the exchange and take advantage of this relationship.
Predictive Strategies:
These can vary hugely in their underlying strategies, but they usually involve the use of quantitative mathematics to predict milliseconds into the future and trade upon these predictions.
High-Frequency Trading Infrastructure:
There are effectively three steps involved with creating an HFT network:
- System Owner End
- Fiberoptic Network/Microwave Tower Network
- Exchange/Data center
System End:
This is ultimately where the strategy begins its life; where the roots of the logic behind the system is based.
This part of the process is often, but not always, led by programmers, traders and quantitative mathematicians who develop and build the strategies to be executed on trades at/between the stock exchanges. Once developed these strategies are then placed in data centers, where server banks will execute the instructions of the algorithm.
Now in reality, many firms collocate their systems close to the stock exchanges to limit the time lag caused by having long fiberoptic cable networks, so the system end may in reality be pretty close to the exchanges it operates on (in some cases literally right next to the exchange's matching engine).
Fiberoptic Line/Microwave Tower Line:
In the United States, where HFT has been more prevalent, the systems leaders have tended to employ individuals involved in physics and telecommunications networks to develop this part of the system.
In some cases the step of laying fiberoptic cable has even involved tunnelling through mountains, the purchase of car parks and agricultural fields and other residential assets so that perfectly straight fiberoptic networks can be laid down - the travel of a fiberoptic network from one side of a road to another can cost as much as 100 nano seconds of time lag.
In order to trade even faster, many HFT firms across the world have been purchasing microwave towers. The map below was taken from a presentation from McKay Brothers co-CEO Stéphane Tyč, showing a map of certain HFT microwave networks in the UK.
Exchange End:
Here is where the system developed executes and trades. If you went into an electronic stock exchange you would really just be entering a server bank that has at its heart a "matching engine", which is where buyers and sellers are paired off with each-other. The HFT firms will have their server banks and data centers close to these exchanges (often right next door in the case of large centers), or sometimes right inside the exchange - many exchanges in the United States can make as much as $70,000 a month for selling proprietary lines to the exchange for HFT traders operating their servers from directly inside the exchange.
Some server banks held near the stock exchanges by the HFT firms will even be tier 4 facilities - they have two of everything inside the data center so that if the system fails the secondary system can takeover and avoid downtime. Other firms are know to have two separate data centers so that if a system failure occurs at the primary data center, the second one can continue as normal.
At the exchange is where the role of the traders in the initial design of the algorithm is essential, because there are over 150 order types used at exchange level in the United States. These order types will be another determinate of the speed at which the HFT system can operate, because certain order types will have precedence over other at the exchange level of the system.
Some order types may focus on:
- The size of trades on the book on each side (bid/ask).
- Allowing the HFT firms to withdraw 50% of order when the trade is enacted.
- Post only orders (the system only trades when a rebate can be gathered from the exchange).
If a liquidity rebate is provided at/between the exchanges where the HFT firm operates, you'll find that HFT firms will have bias to trading here, where they can structure themselves so that the losses they take on making markets at the exchange is outweighed by the money they gain form the liquidity rebate.
Secrecy:
Secrecy has been a key component of the HFT industry since its inception. Primarily this is because if you know how another HFT's system works then you can design yours to take advantage of their system's weaknesses, but furthermore HFT systems don't even like other players to know where they're situated inside a data center - there's a famous case discussed in Michael Lewis' FlashBoys of a firm who covered their server bank with TOYS R US branding to disguise themselves.
Controversies:
The HFT space has been creating global concern since its discovery by the wider finance industry and trading public. These include, but are not limited to:
- Front running orders via automatic market makers.
- Flash orders in the United States (this no longer occurs on major US exchanges).
- 2010 Flash Crash.
- Euronext/Binkbank "not best execution" order placement at exchanges.
- Questions over the operation of HFT firms in banking darkpools.
To generalise, the issues people have raised around HFT usually revolve around the idea that there are slow traders and fast traders. When slow traders (the man trading over the internet connection) are trading against institutions collocating inside the exchange a clear opportunity occurs where the faster player can take advantage of the slower trader.
There's also an argument that there's a conflict of interest in the industry, whereby many of the larger exchanges have shareholders who are very active in the HFT space.
Front Running:
In a nutshell, this is anticipating price movements in the immediate future in an asset and capitalising on cross-market disparities before they are reflected in the public price quotes - or basically you can see what other people are going to do before they do it and suck shares to the HFT firm using an automatic market maker at a low price and then sell them back to the original buyer of those shares at a higher price.
Firms can do this because their fiberoptic networks and abilities to execute at the exchange are faster than standard trading networks used by institutions and retail traders.
This is also known as "latency arbitrage".
Flash Orders:
This was a feature of some stock exchanges where orders would enter the exchange, be flashed to HFT firms who were then given the opportunity to act on these trades before they were passed on the the wider market.
In the wider context of the stock market, this isn't too different from the initial style of trading on global markets, where an order would come into the exchange via telephone, be picked up by a specialist who would then shout out the trade to the floor before entering it into the computer - where it can be seen by the wider market.
2010 Flash Crash:
In my opinion, this is unfairly attributed to the use of HFT on global stock exchanges, but there is no doubt that HFT did have a place in creating some of the wilder price swings that were witnessed on the day when the Down Jones dropped over 9% only to miraculously recover the majority of this loss.
The cause of the Flash Crash has been directed to a very large billion dollar sale of E-Mini futures contracts through an aggressive algorithm. This algorithm instead of selling a piece of the position and then waiting for a market recovery before selling the next portion, would continuously sell contracts into the market. This selling was then picked up by momentum based HFT firms who would them either sell out of long positions or sell short into the market, thus exacerbating the problem.
For people who are interested, this is a really good documentary about the Flash Crash:
https://www.youtube.com/watch?v=aq1Ln1UCoEU
BinckBank vs De Giro:
One aspect about this case that's very interesting is that multiple exchanges were only legalised in Europe in 2007 - before then, this wouldn't have happened.
This was quite a large case in central Europe against BinckBank, who were not facilitating the best prices for their clients through their internal matching engine (TOM). At the same time De Giro was also accusing BinckBank of allowing flash trades and front running to occur through their infrastructure.
Dark Pools:
A dark pool is effectively a private stock exchange that stands separate to the public exchanges. The large banks who own these created the under the premise that clients may get a better deal buying/selling through a dark pool than through public exchanges, where their large buy/sell orders may be recognised by the market and act against them.
The issue here arose when claims came forward from major hedge fund mangers that in certain dark pools there was evidence of the front running of orders within the pool. Further to this, it transpired that some large banks would sell access to the dark pool to HFT institutions such as Citadel and Knight Capital Group - allowing the bank to make money taking very low levels of market risk on either end of this deal.
Is High-Frequency Trading a Threat:
This is a personal opinion rather than a fact, but while retail client orders will always be the easiest to take advantage of by HFT strategies, as a result of a lack of the features described above to reduce the time taken to execute trades, unless you are a very active trader of markets or trade in very large size you're unlikely to have a noticeable proportion of your net worth eroded by predatory HFT firms.
If on the other hand you manage large sums of money, there are still ways to avoid the ability of HFT firms to manipulate your orders:
The IEX in the United States is a stock exchange that is being set up to eliminate HFT's effect on their exchange - HFT finds it very hard to operate here because all trades are slowed down to the same speed and thus HFT intermediaries are unable to operate. Also, IEX's investors are primarily mutual funds and hedge funds, which removes some of the argued industrial conflicts of interests in the exchange.
On a more crude level, delays can be programmed into trading platforms so that orders reach different exchanges at the same time - thus stopping HFT entering as an intermediary.
