Quick trades for a quick buck

25 July 2013 | Tim Phillips

Cover

Tim Phillips

Blog Author | Freelance writer


On October 5 2012, India was the latest country to experience a 'Flash Crash', as its national stock exchange 'nifty' index fell by more than 15% in a few minutes. The cause was a series of huge orders, processed automatically according to preset algorithms. Nobody knew it was about to occur. Nobody understood why it happened. Nobody knew when it would stop.

On October 5 2012, India was the latest country to experience a 'Flash Crash', as its national stock exchange 'nifty' index fell by more than 15% in a few minutes. The cause was a series of huge orders, processed automatically according to preset algorithms. Nobody knew it was about to occur. Nobody understood why it happened. Nobody knew when it would stop.

India's stock exchanges, like most throughout the world, now operate by the microsecond: to do this, it allows large brokerages to pay to put their trading servers at the same location as the bourse's main servers. Co-lo close to stock exchanges is a booming business too, because speed of access to a bourse's servers has become a competitive advantage for stockbrokers who trade in this way - they pay more to get closer, and the most to be closest. Proximity hosting, where the data centre is literally next door, is also a booming business as high-frequency trading becomes more prevalent around the world.

The 'efficient market hypothesis' is that, if everyone knows all relevant information simultaneously, and everyone can trade, a stock price incorporates everything we know about the company and the economy at that moment - and so (if a market is efficient and people are rational, two very big 'ifs') it should be impossible to make money trading stocks unless you have some relevant knowledge. Therefore new information that has not been 'priced in' has an extremely high value - provided you can act on it before other people do. Automatic trading analyses this new information almost instantaneously, using algorithms, leading to occasional flurries of trades. But those trades require that buyers are matched to sellers. Being first in the queue matters, because chances are that their algorithms create similar trading patterns to yours.

Microseconds matter in this world, and some data centre specialists know how to shave microseconds, and have the real estate too. For example, in the US, hosters such as Savvis have found this to be a way out of the commodity trap for co-lo, in that almost all co-lo is basically the same. Savvis claims that electronic trading already accounts for between 65% and 73% of total US trading volume: 'Being even a millisecond ahead of a competitor on a deal can be the difference between significant profit and loss. Any lag or latency could be a deal-breaker,' it says in its proximity hosting white paper 'Cutting Through the Noise', which explains the model in detail.

There's an argument that encouraging automated trading through proximity hosting may destabilise trading, as feedback loops of automated trades provide wild swings in stock markets. The Indian 'Flash Crash' is not the only one of its type, or even the largest. But you can also argue that this is simply a business using leading-edge communications technology to adjust to a new environment and that, as algorithms become more complex, they will become less similar. In which case, there is a sustainable business in proximity hosting.

It may be one of the few parts of the data centre business that can never be commoditised, because not everyone, by definition, can be an equal distance from the required server, and the speed of light is the same wherever you are.

In India, regulation threatens this advantage. After the Flash Crash SEBI, the stock exchange regulator, released a short discussion paper 'to provide greater equality and fairness in order handling to the participants that do not use co-location services'. Its theme was that proximity hosting is a step back to the days of open outcry trading - where seniority was an advantage for the biggest firms because they were better placed in the trading pit. It’s a market that gives an unfair advantage to a few powerful traders, some argue, which is anticompetitive. In February 2013 the paper shows that, at the Indian National Stock Exchange, 74 per cent of equity trading and 94 per cent of equity derivatives were traded from the co-lo computers.

SEBI’s suggestion is that, in future, two separate queues for electronic trades could be formed: one from the co-lo computers, and one from other sources. The time stamp on the trades would be used to settle trades in order.

But a global end to the advantage of co-lo or proximity hosting would not be an end to the advantage enjoyed by the rich and powerful – because already, some of the market data that feeds the algorithms used for high-frequency trading is released to premium subscribers a few seconds or minutes early. This economic rent-seeking on behalf of the data owners is clearly a zero-sum game eventually, because anyone who can pay for early data will be forced to subscribe, and then be back where they were at the start – except for small investors, who continue to miss out. The people who compile the data will see higher profits, at least.

If bourses globally decide to follow India and regulate the proximity advantage of automatic traders, then for the telco business it’s one less way to make a margin in hosting. Algorithmic trading based on a small time advantage may still trigger a cataclysmic flash crash one day soon, but at least no one will blame the co-lo companies afterwards.