Data infrastructure

Back to Community

Data infrastructure

posted

I've used the Bloomberg API, both the C++ and python implementations, for the past decade to extract data to do research, build and run quant trading models. For years there was no better solution. The Quantopian data infrastructure pales in comparison. However, at almost $25,000 a year plus predatory contracts I am always looking for alternatives, part of the reason I have been checking out Quantopian. Unfortunately, the gap between Q and B is enormous; limited data access, a small slice of the global investable universe in a single asset class, rudimentary coverage of even US equities. It is too much to expect any real improvement in my lifetime.

Has anyone checked out Money.net? The NYT had an interesting article about them; I have a call later today with one of their techies about large scale data usage. At $95 a month, the cost is small and if the API is half as good as Bloomberg's, Money.net may well be worth considering.

To all you serious quants out there, what is your favorite data platform?

10 responses

Dov Caspi

Did you check www.quandl.com

Sally Sanford

Thanks Dov. I did use Quandl extensively in the past. They are pretty good but the free data is limited, although easily accessible from both R and python, my languages of choice. The premium data is actually expensive if you want multiple asset classes, but still not as expensive as Bloomberg. I agree Quandl is pretty good.

Dov Caspi

What about:
www.tradingview.com - mostly for fun - I like it very much - Its a platform for writing indicators and analysis

Sally Sanford

I have extracted prices from yahoo and google directly both end of day and intraday. I can extract all end of day history for the 5,000 or so nyse/nasdaq equities in about 40 minutes and 20 days of one minute bars in about 10 minutes neither of which Quandl would let me do for free (2000 API calls per day+ sloooow). There is a wicked cool way of creating a backtest universe of securities for both the S&P 500 and Russell 3000 without survivorship bias, but Quandl is missing many prices of the dead companies.

My main strategy uses over 2000 signals for 10,000 global equities with a machine learning algorithm. I just can't see this working in an interactive test environment.

I'll look at tradingview, thank you for the link!

p.s. I guess this topic is not "interesting"!!!!

Dov Caspi

Hi
In case you are using deep learning and you need lots of history and in high resolution - I am not familiar with free data providers.

But - what platform are you using for running deep learning code ?

I am asking because I am the system architect of a company that wrote a computation platform for deep learning and I have some questions - can you send me your email ?

Thanks Dov
[email protected]

Sally Sanford

Hi Dov,

I sent you an email this morning, let me know if you did not get it.

Sally