What Is Quantopian?
Quantopian is a crowd-sourced hedge fund that offers a platform for users to build and test trading algorithms, for free. Once the algorithms go live, they can then form part of the hedge-fund, where outside investors put capital into the algorithms they think are best. Quantopian’s self-declared goal is to unite algorithmic talent with financial backers, helping people who otherwise wouldn’t have the financial support to make their algorithm a reality.
It sounds like a great idea — and indeed it is. Quantopian is one of several sites that are now available for this purpose. Sentdex.com, for example, is a sentiment indicator which uses algorithms to aggregate investor opinions about stocks and sectors. Quantopian uses both sentiment and fundamental data, allowing users to design models based on anything they like.
Once upon a time, anyone wanting to build a framework of this kind would have had to do it from scratch, but now democratising quant strategy research is becoming big business. Companies like these are still small, but they are growing, and accessible to anyone interested.
Quantopian users can write algorithms in Python using the site’s Interactive Development Environment. They can start out using algorithms from the community, or code from scratch. These codes can then be backtested against the site’s stored 14 years of data about US stock prices and other fundamental data. It’s free to test how the strategy would have performed historically, and can be continually altered until it is to your liking. At this point, the algorithm can be deployed to live trading.
Quantopian doesn’t leave you to your own devices, either — there’s a community that encourages members to share ideas, discuss code, ask for help and share data. It’s entirely free, algorithms are kept secret, and they remain your own intellectual property. You can share an algorithm on the site, at which point it will no longer be secret, but will remain the property of the designer, not the site.
How is Quantopian Different from Sentiment Data-Based Websites?
On Quantopian, a huge amount of data is available to the user. The primary data available is fundamental: the site has price and other data from all US stocks from January 2002 to present. It also incorporates third-party data, which includes news sentiment, VIX, earnings calendars, etc. Some of these are free; others require the user to pay a monthly subscription to access the full set.
Sentiment analysis, or sentiment data, involves classifying things based on a more subjective approach. These kinds of data involve trying to assess the “feeling” of bodies of text — in theory. In practice, sentiment data analysis will involve some kind of data mining, like Google crawling, to work out the general attitudes on the internet towards particular investment fields and similar.
There are a number of ways in which text can be analysed for sentiment data, but there are two major ones. The first one uses sentient fields, or word strings, not necessarily in context. “Good” words will have a score, and “bad” words will have a score, and the ultimate resulting score will indicate whether the text is positive or negative in terms of the thing being investigated.
Perhaps a better model is the Natural Language Processing approach. This one tries to actually understand sentences and context, which means the machine analysing the text must have some understanding of grammatical principles. This means certain parts of speech have to be “tagged” so that the actual language of the text can be understood.
This doesn’t sound as if the data would necessarily be very useful, but in actual fact, most machine based systems actually run at around 80% accuracy already. Sentdex uses sentiment analysis, for example, to compare sentiment towards stock prices to actual stock prices. Sentiment analysis determines whether texts are negative or positive towards certain stocks and markets, and also the degree to which this is true. Ratings websites like Amazon, Google, IMDB, etc, offer excellent databanks of user ratings that can be applied to algorithm processes to show user sentiment towards a certain topic.
Sentiment data based websites, such as Sentdex, pride themselves on using this model to determine how positively or negatively certain markets and areas are currently viewed. On Quantiopian, the sentiment data aspect of the site is much smaller, and in order to use significant amounts of sentiment data, you would need to be subscribed to some of the paid databank services.
Users of Quantopian have generally been impressed by the IDEs, although they would probably be difficult still for someone without development experience. You cannot expect to simply show up and write an algorithm if you have never done so before. However, for those who have rusty experience, the community is a great way to find help.
Quantopian users need to be familiar with Python, based in the open sourced Zipline library. But those who have coding experience found that the IDE was very intuitive, and that the amount of support provided was excellent. There is also a significant amount of real money trading support, through Interactive Brokers.
The Future of Trading?
As a website, Quantopian is relatively simplistic. While you do need to understand code on some level to use this site, it still represents a means of opening up algorithms to people who otherwise would not have been able to access them. You need to understand code to use this site, but you do not need to be a professional coder. Quantopian offers the money and tools to people who know a lot about certain markets: they can then use the algorithms and fundamental data on this website in order to apply their skills to the stock market.
You may not be a professional coder, but if you have never thought about algorithms before, Quantopian may be a good place to start. A little knowledge on this site, with its huge community support platform, can go a long way. If you know a lot about a certain market, but had not known before how to use this knowledge to help with investing and investment brokering, something like Quantopian would certainly be worth investigating.