Learn to live with algo trading

Equity markets have entered an era where machines are rapidly replacing humans. Programme-driven trades account for around 70 per cent of equity trading in the US and for 40 per cent in Europe.  We in India are not far behind with one-third of trades in both cash and derivative segments of the National Stock Exchange driven by such programmed orders.

While many of these programmes use algorithms that execute orders through the day, speed is vital in one subset of algo trading known as high-frequency trading (HFT). In HFT, the programme that smells out opportunities and executes them the fastest, scores. Execution time is measured in milliseconds, or one-thousandth of a second. One buy and sell transaction could take just 10 milliseconds and in the race for being the fastest, traders are moving their terminals as close to the exchange servers as possible.

But as the world awakes to the growing size of algorithmic trading, dissenting voices are also becoming louder. A series of mishaps involving malfunctioning algos have not helped. The regulators are in a quandary on whether to put these trades on a leash or to let them have an unfettered run.

Understanding algos

What do these algorithms do? Some programmes slice a large order and place them through the trading session at the right time and rate so that the price is not unduly affected by the order. Another kind of algorithm identifies price anomalies between price of an asset over different platforms or deviations from typical statistical relationship between assets and executes trades to exploit these differences. These programmes act like conventional arbitrageurs. There are again algos that follow a set of technical parameters to buy and sell.

The above are the more innocuous kind of algos. There are some roguish programmes that sniff the outstanding orders in other systems and take advantage of these. There are algos that make bluff trades so that other programmes reveal their intended trades. They then go on to cancel these bluff trades.

The fascinating aspect is that these algos are constantly mutating to trump their peers.

Risks

The main concern with high-frequency trading is that bug-free programmes are rare. Errors in software that spews hundreds of trades every second can cause severe loss as was seen in Knight Capital recently. This company made software to trade on a new electronic platform launched by the New York Stock Exchange.

As soon as the programme was switched on, Knight Capital started losing money at the rate of $10 million every minute. At the end of 45 minutes, the company had lost $440 million.

There was a similar glitch on the Bombay Stock Exchange (BSE) in the Muhurrat session of 2010. The algorithm of a Delhi-based share broker got into a loop, buying and selling repeatedly resulting in the BSE derivative turnover shooting to many times its daily average. The BSE had to annul all the trades made in that session.

These are programme errors. There are times when a manual error makes algos of various traders go berserk. This was the case in the infamous flash crash in the US in May 2010 when the Dow plunged 600 points in just five minutes and then went up again equally sharply.  This was caused by someone attempting to place a very large order on Globex. This sent all the trading algorithms into frenzy as prices started moving down in a bid to find buyers.

When traders saw their algos behaving in a funny manner, they turned off their system making liquidity dry up. What saved the day was that the Globex platform switched off for five seconds. It gave the much required breathing space in which traders could understand the situation and reverse their trades.

Such manual errors that can send algos into a tizzy are becoming frequent in India as well. In April this year, both Infosys and the Nifty declined sharply and then recovered in purported freak trades. Such trades tend to send algorithms programmed to make the most of such prices into overdrive. The estimated loss from such trades was, however, limited to less than Rs 10 crore.

Do we need HFT?

Two arguments put forth by supporters of algorithmic trading are that these trades improve liquidity and lower bid-ask spreads. If we consider the Indian market, volumes on the National Stock Exchange have more than doubled in the derivative segment since the advent of algo trading. But it needs to be noted that more than three-fourths of the volumes in derivatives is concentrated in Nifty futures and options. Conversely, cash volumes have declined. Since it is the cash market that enables corporate capital raising, algo trades are not really helping to deepen or strengthen equity market. 

Bid-ask spreads have reduced since the advent of algorithmic trading but investors who invest for the long term would not worry much about the spreads.

Live and let live

That said there does not appear to be a strong case for banning these high-frequency trades either. Yes, there have been mishaps, but there is no evidence that they can bring the entire exchange mechanism crashing. Further, algorithmic trades have now grown to such an extent that they contribute a significant chunk to exchange turnover.

It is also obvious why the Securities and Exchange Board of India (SEBI) gave permission for enabling these trades. Foreign institutional investors who control over one-third of the volume on our bourses have now grown accustomed to trading through computer-generated programmes in other jurisdictions. Disallowing these trades in India will only make them move out of our country.

There are already some checks present in the Indian equity market that protect it against a price crash through programmed trades. The price bands that exist for securities, ranging between 5 per cent and 20 per cent, ensure that a rogue software does not wreak havoc on any stock. We also have market-wide circuit filters, trading halts once these are hit.

The SEBI proposed a series of checks in March for both exchanges as well as market intermediaries who facilitate high-frequency trades. These include price, quantity and order value checks so that any terminal putting in inordinately large orders can be switched off. Broker terminals once switched off for exceeding the margins can be switched back only manually. This helps ward off loops. Finally algorithms used by intermediaries and changes to such software needs to be ratified by exchanges.

As regulators struggle to keep up, investors have to learn to make the best of it. After all this could turn out to be the natural process of evolution for equity trading — from trading rings to on-line trading to computer-driven trading.

COMMENT NOW