Last week’s flash crash,which sent the benchmark Nifty crashing by 900 points,has not gone down well with the regulators and exchanges. However,for a set of institutional investors who use a lethal combination of algorithm trading (algos) and co-location servers,it came as a blessing in disguise.
According to market participants,these technology-savvy investors,who have their servers located close to that of the National Stock Exchange (NSE),were able to get their orders executed even before the index circuit filters were triggered. The algos pounced on the shares that were suddenly available at nearly 20% below the then market price,they say.
Algos refer to software programmes that execute trades as and when the pre-set defined parameters are triggered. Co-location is a facility offered by exchanges wherein a broker-member can place his trading servers close to that of the exchange. The closer the server,the higher is the speed of trade execution.
This is clearly corroborated by the intra-day tick data available on Bloomberg. Take for instance Reliance Industries that was trading at R816.70 at 9:50:59am. It fell to R682.35,the next second and more than two lakh shares were traded in a matter of two seconds. Similarly,HDFC,which was trading at R752.35 at 9:50:56am,fell to R631.25 in the next two seconds and nearly 5 lakh shares got immediately traded.
ICICI also had a similar story with nearly 2.40 lakh shares being traded within one second when the price fell from R1,068 to R866.75. An identical trend is visible in ACC,Asian Paints,Ambuja Cements,Punjab National Bank,L&T and Ultratech Cement.
Interestingly,in all these instances,the first order after the price fall has been typically for less than 100 shares. Experts familiar with benefits of algo and co-location say that this only tests the water and the moment the order is matched,the system is flooded with a large-size trade. Incidentally,the NSE has over 200 such servers co-located in their premises.
When a market order wipes off all the pending quantity in the system,this particular order,if remaining,sits at the lowest price, says the head of a firm that designs and sells algos to brokerages. While the index calculation engine computes prices,algos running on co-location servers get triggered and start selling/buying to pick up this order, he said wishing not to be named due to the sensitivity of the matter.
Experts further say that there is no way that an ordinary trader would be able to execute a trade at the lower prices in the event of such a flash crash. The lag could be only in milliseconds,but that is what makes all the difference, says an institutional dealer. While algos are working all the time from various terminals,it is the co-location,which causes more damage than algo as the speed at which it executes orders is much faster, he says.