Historical Intraday Institutional Data. Rent or Buy

AnEdge

Member
Messages
57
Likes
4
I focus on automated trading of US Equities, Options and Futures. My holding times are intraday to about 7 days and I have been doing this for many years. I use R for my research and have a java framework for backtesting.

Over the years I have worked with many data vendors including Reuters, Tickdata, Activ Financial, CSI, and Nanex. These vendors all have good historical data for research and backtesting…even if it is expensive. There is another tier of vendors whose data is much cheaper but with bad quality data…I am not going to name firms as I do not want this to become a mudslinging thread.

A couple of months ago I read about a new provider called QuantGo on a blog. QuantGo has a cloud platform where I can rent access to years of institutional global data at good prices as long as I used a virtual computer instance in the cloud. For example, five years of US Equities Trade and Quote (“TAQ”) data is $250/month and three years of Options (OPRA) 5 minute OHLC bars is $75/month. The cloud platform has multiple data vendors, the two services I listed here are from AlgoSeek and they have tick and bar data for most countries from TickSmith.

I called them and signed up for what they call a Virtual Quant Lab(“VQL”) which is my own private cloud where I can start/stop my computer instances. They provide a website console to manage my instances and it is easy enough, though it does mean I have to wait a minute before I can login into my computer. I have been running both windows and ubuntu servers. As their platform is built on Amazon AWS, I can run any AWS instance type. I have full control (admin and root) of each computer instance and have setup my testing software including some data that I download from other sources on the net.

For testing and research I have been downloading 2 years of trades and quotes data for 300 US stocks which I then pre-process before using it in R. Downloading one day of data TAQ Equities data is very quick (30 seconds for IBM) and it takes 4 minutes to download one month(January 2014) of IBM . Apparently I can get a direct connection over fiber which will give me dedicated bandwidth instead of downloading over the internet but this costs $1,000 per month and is for institutional clients. It is important to note that I can download my research results and derived data but the original tick data must stay on my cloud computer instances (they do some type of network monitoring based on bandwidth to enforce this).

Overall it has been pretty good, previously I had boxes on my network and would remotely log into them and now I just remotely login into my boxes in the cloud. I have to remember to stop them when I am not using them, as you are charged per hour of computer usage. I have been using an m3.medium instance size (equivalent to a desktop) so my hourly cost is $0.16. The data is real institutional data (I compared it to another data set). The AlgoSeek data services stopped for one day which apparently was due to a network issue and I had problems starting my computer instances another day which was an AWS issue. QuantGo team has always responded to my emails.

In summary, if you are ok with remote cloud computer instances the this gives access to years of data for fixed monthly fee.

Does anyone else have any insights into QuantGo and are there any other services like it ?
 
Last edited:
Top