High Frequency Trading (HFT) and Colocation

I often get asked “Do retail traders get squeezed out of the markets due to co-located automated robot-trading”? The answer is nearly always no. High Frequency Trading refers to automated algorithmic trading. The type of trading done by client-side computer servers in a either full or semi-full autonomous fashion. Co-location refers to co-locating these client-size computer servers closer to an exchange’s matching engine for increased speed and throughput, oftentimes to seek a difference of just a few milliseconds.

In this blog post, I will attempt to answer why co-location doesn’t hamper the average retail trader and why the average retail trader shouldn’t be so worried about it. To be clear, I am not trying to compare automated high frequency trading versus human trading, but rather why the addition of colocated automated trading does not impact manual trading strategies.

Let’s first discuss various types of latencies incurred, from most to least:

  • Human latencies: 250-500 milliseconds. This is pure human reaction time – from the time your eyeballs read something on the screen to the time your fingers react.
  • Server reaction latencies: 5-50 milliseconds. This includes operating system latencies, application level latencies, and hardware latencies.
  • Network latencies within Mumbai (non colocated): 10-30 milliseconds. This is the latency incurred if a server were at a non-colocated site, say an office premise or a off-site server hosting facility within Mumbai. It is the time taken by a data packet to travel from the server all the way to an exchange’s core network switch.

As you can clearly see, human latencies are 10 times more than the high frequency trading latencies involved. A server, even at an off-site, can react to a price fluctuation and place a resulting trade 10x faster than a human being. This fluctuation becomes even more steep once you get into optimized hardware and software. An optimized server setup can be designed to react within single digit millisecond response time nowadays. A human being simply cannot compete with that.

The addition of colocated high frequency trading simply removes this last network latency of 10-15 milliseconds by allowing the automated algorithm to be situated within an exchange’s building itself. So the network latency becomes almost nill. In other words, a colocated server algorithm can react to a price change and issue a resulting trade within single digit millisecond latencies end-to-end, whereas a human being will take 250-500 milliseconds. The addition of co-location simply removed an additional 10-30 milliseconds of overall reaction time.

It’s all about the strategies

In India, the NSE, BSE, MCX and MCX-SX offer co-location facilities to host automated algorithms. These high frequency trading algorithms must be approved by the respective exchange(s) prior to deployment. As of now, SEBI doesn’t permit retail traders to employ algorithmic trading. Although we hope to see this change in the future as per international standards, this doesn’t mean retail traders are necessarily disadvantaged. The types of strategies these automated algorithms deploy (co-located or non-colocated) are oftentimes super latency sensitive arbitrage opportunities. Strategies like Cash Futures arbitrage, Spread arbitrage, Delta Hedging and Index arbitrage are essentially reserved for high frequency trading. Retail traders trade at a time frame above that. Technical and Fundamental analysis do not rely on split millisecond time frames, but are often times much better suited towards longer time frames (for example, see Pairs Trading). At the same time, high speed automated arbitrage ensures that markets are efficient and that prices converge as they should – these are all ultimately positives for the retail trader.

A retail trader should focus on strategies that he/she sees developing on their screen. If you can see it register on your screen, it is all after low latency arbitrage, and it’s thus safe to say you can make an informed trading decision on it.

Ravi Kumar

Ravi Kumar

Ravi's background is in Computer Science. He loves technology and is a true tech geek. Prior to RKSV, he worked with several prominent technology firms in the US and managed his own proprietary trading company. In his free time, Ravi enjoys producing music, creative arts and working out regularly.

  • Joe Brunner

    Great article man… saying HFT effect Retail investors – is like saying Magnus Carlsen effects a couple of guys playing Chess in their garage. There are so many strategies and so many ways to trade them (low, mid, high latency) that the # of combinations insure that many people can win (big) at every tier. In any market or industry, its easier to “blame the popular winners” than to do your own homework and be better.