Today’s hyper competitive trading world necessitates ultra-fast trading systems, but while speed is good. Speed with reliability is better.
"Experience is a good teacher. From the knowledge shared from the past of trading the community; the significant risks are implemented by most institutions"
In today’s technology intensive world, ultra-low latency trading is the new normal in the competitive Institutional trading world. So, what is next?
Most of the technology solutions either use FPGA’s on Layer1 switches or techniques like TCPDirect offered by Solarflare to settle down around 100 ns latency benchmarks. India’s leading stock exchange has rolled out a feature called Gateway Router which connects to one of the exchange’s matching engines randomly, thereby removing the deterministic edge which could be potentially enjoyed by minor infrastructure difference in any. In this fair playing HFT field, the next important technology challenge is building an Institutional Grade HFT Risk Management system.
All the pre-trade risk checks mandated by regulators are kept inline while placing orders. The broker will have limited visibility of the trading algorithm and risk management of a client to assess the risk level. This pre-trade risk has to be augmented with additional near real time HFT risk management systems to protect the broker from financial loss arising due to technology glitch, the risk arising from the slow human intervention, extreme market volatility due to market events, hidden risks embedded in the portfolio of the client and risks detected from trade surveillance of exchange.
In HFT trading the actual risk has to be assessed with a conservative approach considering the exchange throttle available, the speed of exchange connecting network, the round trip latency for an order to be acknowledged by exchange, and the potential in-flight orders and trades in the wire. Hence this forces the risk manager to consider trades and open orders as also as potential trades. Another key requirement of effective risk management is acting on a risk alert on time. A human risk manager can’t act in micro seconds to manage the situations. Hence, the risk manager decision and the intelligence have to be programmed into the Risk Layer to get the situation in control. There is also risk arising from technology failure, malfunction or slowness. These situations also has to be considering accommodated while building the additional near real time parallel risk management system robotic risk management followed by a second level of human verifications is an absolute must.
Experience is a good teacher. From the knowledge shared from the past of trading the community; the significant risks are implemented by most institutions. But more importantly upgrading the risk management system constantly from the learning obtained from close observation of day to day activities is important. Changing complex software also is a challenge which can be tacked by robust automated regression testing and quality assurance processes.
Edelweiss Capital Markets and Advisory has uniquely positioned itself to handle the challenge by investing in cutting edge technology and built the product called Tradeweiss Trading Platform. The platform has a simple abstracted interface because of which if a client integrates once, he can trade all venues in India. The team also has a fully automated RMS suite with 90,000+ scenarios and 200+ race conditions. It has near real time margin management and surveillance capabilities to ensure the client and organisation is covered from all known risks and pauses trading intelligently when it enters into area of unknown risks so that humans can intervene.
It is said that innovation isn’t worth a thing, unless it is combined with reliability. Tradeweiss offers this edge by combining it all – the Edelweiss way.
Though point notes
Ignorance of risk because strategy writer looks at profit, but do not focus on loss creating circumstances like hardware failure, key field corruption due to software glitch, movement of market while millions of dollar’s open positions during a software recovery, limitations from the exchange check control check points which get triggered, Margin risk
Risk reduction mode software latency inside matching engine
In a 10Gbps exchange connectivity with a throttle of 1000 Msg/sec, 1000 orders can be placed in approximately 500 us(theoretical limit due to network serialisation)