Back to Community
How much capital is assigned to a strategy?

Pardon any ignorance in this post, I have started looking at Quantopian in the past week and am beginning to develop some sample strategies after following the various tutorials. I can see the value in the provided libraries to cut down a lot of code.


I am wondering how much capital is assigned to a strategy if it chosen to be live traded?

This question is based on two considerations:
1) A strategy may require a minimum amount of starting capital
2) To assess whether it would be viable to expend a lot of time and effort within the Quantopian platform, one would want to know that a yearly return off an effective strategy would be significant. i.e. with $10,000 capital and 10% annual return (=$1000) the developer would only get $100 of that at the 10% that is paid to the developer

Thanks for your time

S

9 responses

Even I am interested in this information. Even if an algorithm is allocated 1,000,000$ and it makes 10%, the developer gets $10K per annum which does not compensate for the time and effort.

Thanks for confirming your interest Pravin, hopefully this will add its validity and prompt some informed answers

At $1M it very well could end up a break-even endeavor for some. Say you have alternate means to make $50/hr. on the side (or in a full-time job). Do I put in 200 hours to develop an algo for Quantopian, which may not even get selected (and I'd need to wait at least 6 months for out-of-sample paper trading results to find out). Or do I put in the 200 hours for $50/hr., with 100% certainty that I'll get paid?

The other issue is that a scalable long-short algo, with hundreds of stocks potentially, is not amenable to retail trading by its author (unless he's loaded, in which case, why would he be writing code?). So, if I don't get selected, I can only try to seek out another source of institutional funding, which will probably require a lot of work.

At $10M or more, it is another story, unless you can make $500/hr. (without engaging in criminal activity).

In the end, it probably only makes sense as a fun, free hobby, with a potential big pay-off adding to the allure (like combing the beach with a metal detector).

For Q/Point72, my read is that it is also a nice recruiting tool. What better way to search the globe for top talent! I have to imagine some Q virtual managers will become real ones.

Hi Grant, thanks for the reply.

I understand what you are saying and it seems realistic. So with your 538 algorithms, (if you don't mind me asking) have any been selected and are you making a profit? I assume from what you have stated above it is not a livable profit if so?

"Fun free hobby" sounds about right. My banker friend was talking about the huge turnover of algos and analysts at Millennium. Churn and burn. Seems to be the name of the game.

I have to say, if I wanted to raise capital (god forbid) I'd probably rather the old style CTA TF approach. Still seems to be working well for some of them.

A far cry from the highly leveraged picking up dimes off the railway track approach of these huge leveraged players.

Anyone doing Numeral?

That is an interesting perspective Grant. The way I see it is that you have to be fully committed to make a decent algo else with a half-hearted attempted its performance could be anyone's guess. Unless Quantopian has a secret mantra where they can take average algorithms and then mix them together to generate robust returns.

@ Steve -

I have not been selected (aside from my early fame as the winner of the first contest, which was a burning of $10K of Q capital). That said, it would be kinda fun to be selected, just for yucks. I'd like to think I have a shot, if I plug away at it. It is an interesting challenge, and I continue to learn about this area of finance. Kinda cool in my book.

@ Pravin -

I'd like to think my effort isn't half-hearted (definitely hacked, though), however, it is a challenge to keep up with the Q API changes and additions, and to research from the ground up. As you suggest, it is a pretty big commitment to do it right. We are amateurs in a professional domain.

@Grant-

Assuming you know what you're doing, your 538 algos without being chosen seems somewhat of a good measure of getting chosen. As well as maybe the overall difficult of creating a good algo.


On the notions put forward by people above, it seems a better approach to develop knowledge and gain experience writing algos for a different platform. i.e. if you wrote an EA in MQL4 for MetaTrader4 and ran it in the cloud you could get investment from outside source a lot easier (i.e. prop firm). If you develop Quantopian it seems to be a risk that you'd get used to the tasty libraries and add a learning overhead in the future when you were to convert your algo to another platform. Any thoughts on this?

Working backwards: they have $250MM in investment. The way they currently manage algos is pretty labor-intensive, requiring separate accounts and each algo running its own order handling logic. I don't think they'd want more than 50 of those running. So, if you get selected, an allocation of $5MM seems reasonable to expect. The fact that they recently switched the contest to $10MM of paper capital is a good indication that that is the ballpark they are thinking too.