Program trading – the computerised exchange of vast sums of assets – has champions and critics in equal measure.
It has been blamed for some of the market's biggest excesses and some of its blackest moments in recent history. But it also helped provide a key source of liquidity through the financial crisis and beyond.
For retail traders, it has been both a godsend and devil made. Bid-ask spreads have narrowed, making profit (and, indeed, losses) easier to come by and broker commissions more difficult to charge.
But at the same time increased bouts of volatility can be exacerbated by program trading and it was considered responsible for the 2010 "flash crash".
What is program trading?
Modern definitions of program trading refer to computerised or electronic dealing, but more specifically techniques such as algorithmic or high-frequency trading (HFT).
Complex computer algorithms, or "algos", are used to trade vast quantities of shares or foreign exchange by just a few points, even fractions of points.
Buying and selling occurs in nanoseconds and because of the large volumes traded at a time, huge profit can be made on the price moving by just a single pip.
Who's doing it?
Mainly large institutions indulge in program trading.
Buy side firms such as hedge funds and mutual funds, are involved but more often than not it is specialised HFT firms, and banks’ proprietary trading desks.
It is a highly-specialised sector of financial trading due to the technological skills needed to devise algorithms that recognise certain patterns of trading.
It is also a high cost enterprise as the hardware that cuts down latency – the time taken to complete a single trade – is dependent on the placement of servers, routers and other necessary equipment, as near as possible to the location of the financial exchange.
If such equipment can be placed adjacent to the exchange's servers – a process called co-location – latency can be improved and HFT algos can react to changes in prices a second or so ahead of the investing masses.
Exchanges, however, charge millions of pounds for such privileges.
When it goes right
Given the distinct advantage presented by co-location HFTs and other algo traders can make millions on tiny blips in asset prices thanks to the reduced time it takes to act on the information and complete the trade.
As technology becomes ever more advanced, with the introduction of cloud computing and machine learning/AI, the human factor – which is responsible for the vast majority of trading errors – can almost be eliminated from trade.
Although the cloud is not particularly useful in executing program trades, it can be an enormously useful tool at the beginning of the process.
Using cloud's massive computational power to design and stress test algorithms, possible errors can be eliminated or compensated for.
Critics say program trading has few intrinsic benefits.
Shares are rarely held long enough to aid investment and it is merely an information arbitrage, but the industry has always countered with its importance as a liquidity provider.
But when it goes wrong
Philip Stafford, the maestro behind FT.com's Trading Room, reminds us here of how spectacularly wrong program trades can go with his top four trading disasters.
1. The flash crash of 2010 – Philip calls this one the "daddy of them all".
On 6 May – election day in the UK – markets were already jittery about the mounting Greek debt crisis.
In just minutes the US Dow Jones Industrial Average slumps 9% after an algorithmic trading strategy used by a US mutual fund to execute a $4bn sell order triggers further selling by HFT firms.
By the time prices regain some stability, some of the losses are recovered and the Dow settles the day 3% lower.
2. Knight Capital 2012 – The electronic market maker's lesson in how to lose $440m in half an hour.
The firm's electronic trading software goes crazy and starts buying high and selling low on nearly 150 stocks.
Knight's share price loses more than 60% – its market cap now worth less than the $440m loss – and is only saved from bankruptcy after a $400m lifeline thrown by a group of investors.
3. HanMag Securities 2013 – Small Korean brokerage bankrupted by program trading.
A trading accident due to "faults" in its network programs causes the firm to lose $41m in one day.
Given that HanMag's market capitalisation is only about $19m, the firm is sent skidding into bankruptcy.
4. British pound flash crash 2016 – Sterling slump triggers sell orders.
The pound makes a fairly ordinary-looking 1% loss.
Because this is during low volume Asian trade, however, market dysfunction is amplified as options traders sell sterling to cover their positions. Losses are extended and trigger program trade stop-loss selling.
The pound falls as much as 9%, depending on whose data you use, but as more experienced London traders come on line it recovers to finish the day just 1.5% lower.
Mitigating potential losses caused by flash crashes
Sometimes the market will fall so rapidly, there's little you can do.
However, using limit orders and stop losses may help prevent some of the most severe effects of a flash crash.
Limit order: this will execute a buy order at a price deemed acceptable by the investor. It is not a flawless measure, but can prevent the risk of volatile market behaviour sharply raising the execution price of your orders.
Stop loss: the simple stop loss will sell once the investor's designated sell price is reached. It is not always possible for the broker to hit the exact level, however, and in highly volatile markets prices may drop so rapidly that greater losses are sustained.