"Has Quantopian become all you had hoped it would be?" Strikes back

I thought it was time to bring back the interesting Market Tech's topic: Has Quantopian become all you had hoped it would be?

This topic is not intended to praise or blame Quantopian but instead to listen to users ideas and comments on current and future Q features, possibilities and speculations. Mostly for fun but also because it is inspiring to listen to other people's ideas.

My view is that Quantopian has really grown to be a good research platform and the quality is increasing year by year. It is very common to find posts on academic papers, quant books, implementations on those two and serious digressions on quant and math topics. This is due to the fact that they created the Research environment and also due to the Q team effort to periodically share learning material, papers, NB and algorithms. I really like this Q approach of leading new users towards quant finance subjects and not letting them on their own.
That teaching and research approach, together with the choice of python, scientific python modules, market data, fundamental data, external vendor data and a very talented software engineering team make Quantopian a first choice for anybody who wants to start the path of quant finance and certainly make Quantopian my preferred platform.

When I look around the only alternative to Quantopian I can consider is Quantconnect. I believe that's the only alternative because it has been around for a while (like Quantopian) and it's not a "2 people company" without solid base. I personally don't find Quantconnect technologically appealing (last time I checked: no research env., .Net instead of numpy/scipy/pandas ) but I would like to highlight the different business model they created and the consequences of that choice. Quantconnect is a platform designed for running users algorithms on users' account. Somewhere in the website they claim they might connect investors to algorithm creators but I didn't see it happening. What is real is that users have to pay to run algorithms on Quantconnet (that's their business model) and this leads to the consequence that the users requirements lead the development direction of the platform. This means if users request some data (e.g. FOREX or tick/second data instead of minutely data) then Quantonnect makes that a priority. If users complain of a missing feature in the GUI that is fixed/improved/implemented right away.

The bottom line is that the business model dictates the direction of a platform and so the features the users will have. Now let's consider Quantopian case, I can see the users' needs are not prioritized: FOREX is missing, second/tick data is missing, lots of reasonable improvement requests have never been implemented (additional backtest information, editor features, source code versioning, algo in multi files, algo in sub folders etc. etc.). I am not blaming Quantopian, I understand their available resources are limited and they focus on their business development.

On my side, I will be more than happy to pay for Quantopian, as long as I don't bind my code to a proprietary platform (zipline is open source so it's ok) and as long as the money I pay are used to develop the features I need for my algorithms (at least partially). The current situation is that I wouldn't be willing to pay to run algorithms in Quantopian as their business model draws resources away from the development of features I need for investing my money. At the same time I believe Quantopian's team is very good in producing high quality products and I also like the idea of crowdsourced hedge fund. So I would really like to see Quantopian becoming a platform with equal effort on both users' algorithm needs and crowdsourced hedge fund. Also, I would be more than happy to pay a monthly fees for such a platform (and to have a dedicated virtual machine so that I won't see any backtest timeouts and memory/usage limits ;).

One last note on the crowdsourced hedge fund. While I agree with Quantopian's strict choice of algorithm selection for their fund, I still find interesting the Grant's idea of build a fund out o hundreds of algorithm (without strict selection criterias) with a small allocation each. I wonder why Q cannot adopt many different approaches: if the users had access to all community algorithms performance, they would be able to try and find a portfolio of community algorithms that performs well. All in all, the Q hedge fund doesn't prevent the platform to try other options, they can all coexist. Quantopian would fund only their selection of algorithms but the users could still use all the community algorithms (as black box, not the source code) for their research and maybe they could submit their portfolio of algorithms in a special competition aimed to find the best "algorithm aggregator". In the first place is Quantopian that taught us to think big and innovative ;)

Maybe Q has already proven to themselves that the $10M/algo route is the only way to go at this point. At QuantCon 2015, I had a chat with one of the Quantopian principals and there was mention of figures like $100M/algo, to which I responded that it would be a full-time job at that point--no longer crowd-sourced. It is probably the practical thing to do, to try to get off the ground. Q is stepping into a mature fund market, and they just need a small slice of it. Even without some new-fangled approach (which Q has with their black-box, crowd-sourcing), they'd still need 10 years or more of real-money trading, ideally through some ups and downs in the market. Back-of-the-envelope, Q has to raise a lot of capital in a short amount of time to pay their large development and support staff. They need to herd the crowd to familiar pastures.

Regarding charging for trading, I gather that there are other ways to make money off of retail traders, including payment for order flow (see https://en.wikipedia.org/wiki/Payment_for_order_flow). To my knowledge, Q has never denied doing this sort of thing, so maybe it is part of the business plan? One has to wonder what the game plan is to monetize their 80,000 users, since only a tiny fraction of them will produce for the fund.