Analysis of High Frequency Data

Algos, strategies, code
1

Hi All,

We did some analysis on High Frequency Data (not available for retail traders) where data comes in every 2-3 microseconds (1 millionth of a second).

Just to give some perspective

  • Each day close to 1 TB of data is generated by the exchange
  • Data is disseminated across almost 30 IP addresses
  • The total bandwidth required is 10GBps (normal house internet is max about 300 mbps - some have 1 Gbps)

So what can we say about intraday trading ?

  • Profitable trades occur only in 20% of market time. So in essence - if your in market time where your positions are generating PNL is more than 20% of your investment horizon, you need to relook at your strategy

  • Roughly direction of market is determined in time steps of 120 microseconds - so for a trend to develop - it takes within a second - whereas market data comes in 1 second intervals for most retail trading platforms

Anyways - it takes years of development and fantastic technology professionals to build a platform that can handle such speeds of data. Hence so rare.

2

Is ‘HFD’ in any way useful for retail traders?

3

Yes ofcourse. Being retail just means you are not an institution - but you can trade it if you are able to process it and work with a broker who has colocation servers deployed.

4

What’s the use case?

5

Thats for you to research :slight_smile:

6

Let me guesstimate (don’t have the opportunity to analyse). HFD is unusable for the discretionary retail trader. And not of much consequence even for the retail algo trader.

7

Yes absolutely useless for discretionary trader

For retail algo traders - I think quite difficult if not impossible - as someone who has the computational infra to process is most probably not going to be a single person trading - more of a team with multiple competencies.

That said, any group of people as AOP or LLP want to trade - will still be classified as a retail trader. Typically institutional trader means financial institutions.