Since the Quantopian announcement Phasing Out Brokerage Integrations, have there been any developments by other parties to bring retail algorithmic trading to the masses? I've decided to look into Alpaca. Are there other good options (that are aimed at Quantopian-style algorithmic trading) that scale down to minimal starting capital?

Also, I'd be interested if any current Quantopian users still find Quantopian useful in trading their own money, and if so, how?

13 responses

What kind of program will you be adopting Grant? I find some sort of cut down version of long short using two or three fundamental factors attractive but I suppose to be practical one would have to limit the universe considerably to perhaps a few hundred shares.

Would you be seeking to clone Quantopians optimisation routine? Or would you be going for something more modest?

Also for my money I would want something running on a third party server, preferably the brokers but I don't think IB offer anything like this. I must say Q did become a lot less fun after it cut out the possibility for personal trading.

Also a lot of interesting, amusing odd ball characters bailed out.

It has all become rather sterile. But maybe that's good for Q's business?

@ Zenothestoic -

Just trying to understand the options at this point.

QuantConnect has a nice platform. But their backtest is really really slow. I heard from their CEO that they are working on fixing that. They don't have an optimizer. They do have forex, and futures too.

Ah good point. Now all I need is a copy of the algo on that other thread which is up for sale and will buy me the universe in the next few years.

@Grant,

I am right now in the same boat. The route I am taking is renting a server on the cloud, purchasing data from www.intrinio.com and building my own infrastructure to integrate with brokers.

I must add that I found this to be the best option because it is not tied to a platform and the alpha is completely in your control.

Data Kris
Are you offering this as a service to others or is it just for your own trading?

Cynically speaking I wonder whether you would end up making more by offering a paid service for others than running your own algos?

@Satya, can you give me an idea as to prices of these datafeed? Say, just US stock prices and fundamentals only, how much will it cost you?

@James - https://intrinio.com/financial-data/us-fundamentals-financials-metrics-ratios-stock-prices (around 44$per month) @Zeno - I have been approached by someone for a 6-month evaluation (out of sample). I need to build this infra. Not trading my own capital. I do trade small amounts but on zerodha (Indian stocks). @Satya, thanks! Good luck building your infra. Perhaps not what you are looking for, but nevertheless quite interesting is the crowd-sourced Apiary Fund. Their business model is somewhat different to Q's. It is, shall-we-say, "less intellectual" and focuses on very practical hands-on trading skills, specifically with FX, but nevertheless the practical insights that one can gain there may also be extrapolated to other (leveraged) markets. They have a well-structured step-by-step training program that starts off very easily with extremely basic trading skills and then progresses steadily from there. The participant can either pay for training OR do it for free, and there is absolutely NO necessity to pay for training. If you look around on the internet, you may see one or two comments to the effect that "Apiary is a scam". AS far as i can see those are NOT valid comments and have no more basis than the sort of "Q is a scam" comments made by the inevitable stupid & discontent people. I started participating in the program myself about a year ago, have been very pleased with it, have seen no evidence of "scam", and there is no cost except for a monthly live FX data feed fee. At the end of the program they have something more-or-less equivalent to Q's "allocations" in which, if you can actually trade well enough on a practice account, they will then put up the money (up to max$ 250k), you the trader get to trade with it and to keep most of the profits. (I am still not quite there yet .... still improving my live trading skills). Traders can either choose to trade manually watching the screen, or alternatively write & use their own algos, which in their context they call "Expert Advisors". These, if you want to use a mechanical / algorithmic approach, you must write for yourself. The language for that is predominantly C# with the addition of some specific trading function syntax as per the MT4 language. Although in many ways quite different compared to Q, personally i think they actually complement each other nicely, especially if you like real-time live trading and want to develop that as an additional skill set.

We are moving to QuantRocket, which only is interfaced with Interactive Brokers API. You can use data from IB or other sources (Quandl). They have their own proprietary backtesting metholodogy called Moonshot but also support Zipline, although Zipline Live Trading is only planned for Q2. QuantRocket you can also setup locally on your own hardware which might be good for IP reasons.

I have been using Alpaca. I had trouble installing pylivetrader (their implementation of zipline) due to some weird dependency issues. So I've just been using their REST interface. It's easy enough.

My experiences so far have lead me to the conclusion that Quantopian is quite poor at accurately modeling fills when dealing with retail-levels of capital. That's something to keep in mind.

Alpaca does not currently support shorting stocks, but they are working on introducing it.