Witnessing HFT’s return

Back in February, I was working in a trading platform for a high-speed shop. We were working for more than 8 months until we hit such a great returns.

After several months, I had permission to share (anonymously) what 2 hours of a real high-performance trading looks like. And wanted to share this post to give an idea of what kind of trades and how fast/frequent they are.

The platform was created from scratch, using lightweight C++ code. I was especially focused on the architecture of the software, something crucial to achieving high-speed processing. After platform was done, was the turn of creating, testing and adjusting the strategy.

The result is a system with excellent returns on the forex market. And by the chart below you can see by how much.

Several things to look at:

·        Winning ratio is 80% (at least in these 2 hours)

·        A total of 109 trades has been made over this period

·        Average trade, last 206 milliseconds

·        Note how transaction cost affect the system

o  Blue line is without cost

o  Orange accounting transaction cost

·        Attempts to close means how many times we tried to flat our position but couldn’t (because of execution, or price movement). Every time there is more than one attempt, the trade is a losing trade. So, that’s the proof of how important execution quality and speed is.

Summary of the collected trades

Equity curve, with and without trading costs

The actual trades

You can find the excel sheet showing all this at this link http://www.slideshare.net/silahian/what-2hs-of-a-real-hft-trading-system-looks-like

Conclusion, execution speed is priority number 1, then the quality of execution and last liquidity.

ENJOY! And Share if you like it!

 

Ariel Silahian

http://www.sisSoftwareFactory.com/quant

https://twitter.com/sisSoftware

Keywords: #hft #quants #forex #fx #risk $EURUSD $EURGBP $EURJPY

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.