Don't see how that solves anything, you could, in theory just make a algorithm that executes exactly whatever orders you want (heck, you can have it execute orders from an csv stream you yourself control).

It is a good point, but controlling each order in real-time or with some signal is not the intent here. Any extraneous non-market input should be blocked. I guess it would be rather easy to limit all input parameters to the market data and history() from Quantopian.

A preset algorithm that only reacts to market information available to general public is what I am after. The regulations are against insider trading, and potential misuse of non-public information.

I know, but if you can switch algorithm at will, you can always make an algorithm that will allow you to trade that particular stock (say you use morningstar data to filter the size and sector of the company along with the price of it).

And since algorithm makers on Q can be anonymous, I don't think it would be too hard to use Q to execute inside trades like that. And I don't think a Q ran algorithm would qualify as a manager.

Raghavendra,
I like your concept, but I don't think it is the current direction. From everything that has been communicated, Q is doing a hedge fund for accredited investors only. Certainly, one will be able to invest in the fund if they are an accredited investor, I believe. I'm not sure if they're planning to do an LP setup, or giving access to accredited investors directly from IB accounts.

Back to your original concept, I always thought it would have been very interesting if Q would setup as an automated, online RIA (open to non-accredited investors). But, they do not seem headed down that path (for reasons I can understand). However, I think Q should work with enterprising quants who want to perhaps start RIAs using Q's platform (of course) so that ideas, like yours, can become available to non-accredited investors. Here's a thread of a similar discussion: https://www.quantopian.com/posts/if-i-were-q-i-would-dot-dot-dot

Hi Jeff,
Thanks. I read your post. I think your idea is really good.

Take a look at this link Global Autotrading This is a web-site that offers something like what I was suggesting Q could do. They have setup an agreement with IB to auto trade customers accounts. I currently have an account with them, but their fees are outrageous. The fees you were thinking of charging would be orders of magnitude cheaper.

I am also considering offering my own algo (once I am successful myself with it) to other investors via some means within Q. I thought there were regulations meant to protect non-accredited investors from high-risk active trading type of arrangement. From what I remember reading about this some time ago, these regulations would prevent us (hedge funds) from advertising publicly or on any web-site etc. Do you think you can get through the regulations to be able to offer these services to non-accredited investors? I think a lot of people would be interested.

Raghavendra, I think your idea is good and very exciting. I don't think you need Global Autotrading. In terms of advertising performance (to non-accredited investors, or accredited investors), you should talk with a securities attorney, as this is something to be very careful about with the SEC. My understanding is that it is possible to advertise performance. HOWEVER, you have to be 100% accurate and truthful in any advertisement of performance. It is tricky because if you have several clients, one client's net performance could be slightly different from another's because of a fill, or fee depending on account size, etc.....So, perhaps you'd want to always be careful to run consolidated performance in IB of all client accounts (your own personal account included) and save a copy of this performance report so that you could show the SEC in an audit what you based your advertisement on.

My opinion would be that you consider the following (and again this is just my opinion):

1. Register as an online RIA with the SEC, list your algo strategy in the form ADV so that everyone is clear on what you're offering, just a written general description, obviously not the code (by the way, you'll have to take the series 65/66 to become an RIA)
   
2. Use IB (because it is low cost and can integrate with Q well)
   
3. Automate your RIA in all possible ways (e-sig client agreements, etc)
   
4. See if Quantopian will work with you on a low cost way to link your algo with each client's account
   
5. See if Quantopian will give you some free hours with their attorney regarding advice on how you want to advertise your Quantopian algo performance to attract new assets
   
6. Advertise your Quantopian algo track record online to attract new client assets using accurate IB net performance reports
   
7. Purchase errors and omissions insurance
   
8. Register as an LLC (in a state that offers max protection) to protect your assets
   
9. Partner with an academic institution to publish something showing research of why your algo is better than MPT and advertise it

Again, just some thoughts and my opinions/brainstorms. Best of luck!

Excellent list Jeff - as an RIA, is it the case that you can still only market to accredited investors, or is it wide open?

Last I heard Q could not hook up to SMAs/advisor sub-accounts but I bet that could change if folks actually start trying to make this happen. I am not sure how many of us are considering going down this route?

Hi Simon
If you only charge a management fee of total assets, then wide open.
If you would want to charge a performance fee, then accredited investors only.

On sub accounts, yeah that needs to be worked out. Or, maybe Q will let RIAs link algos to each client account for very nominal fee?

Is there a book that goes over all the mundane details of starting up an asset management business like this? Every time I think about it, I get overwhelmed by details of off-shore feeder funds and all kinds of things. Also, I live in Canada, so who knows what additional complexity that brings.

Last time I investigated it, I think it was much easier to become a CTA than an RIA for securities...

I have never looked into Canada. RIA in US is not too difficult or expensive to setup. If you need help in US or would like a partner in US for RIA, let me know.

Thanks Jeff. It is an excellent list of steps! I am guessing you are already an RIA? Does being an RIA impose more restrictions on what you can trade in your own personal account? Further, I suppose having one "good strategy or algo" would not be enough to be an effective Investment Adviser since one would have to cater to different "risk/return" profiles of the clients.

My view was that Q can have a few RIAs as consultants who would be in a position to advise investors (accredited or non-accredited) on which algo or which combination of algos to choose from based on the risk profile of each algo and the client's risk appetite. Then there would be the algo contributors who are not necessarily RIAs, but would benefit in some way (a small percentage) from the total dollar value their algo is trading.

My original thought was to be an algo contributor to start with. You have got me thinking about being an RIA.

Interesting thoughts/points, Raghavendra.
I used to be a broker at a large wirehouse, left, then started a small RIA (pre-Quantopian), shut it down because I wanted to re-think the strategy. Now, I'm interested in doing an on-line RIA using Q's platform. Yes, you're correct. As an RIA, you have to be careful and certainly not trade against your clients. You cover this point in your form ADV. And, if you trade one of your algos with your personal money, it can trade along with your clients. Again, there's verbiage in the form ADV to cover this issue.

In terms of "one good strategy or algo"....this is an interesting conversation...conventional wisdom would say to offer clients several algos based on their individual risk tolerence, based on an on-line list of questions they answer, and they would get put into a conservative, moderately conservative, moderate, moderately aggressive, or aggressive algo. This would be a very prudent way to offer investment choices to clients. On the other hand, if your algo manages risk well, could you offer just one algo to all investors? Perhaps. I think it is possible. Chances are, you're obviously not offering typical MPT portfolios like every other RIA/Investment Advisor/Broker out there. You're going to be offering something different that probably is diversified, but it probably changes/pivots as the market changes. Sometimes it will be conservative, sometimes it will be more aggressive. So, depending on your algo, you might be able to offer one algo as it will manage risk well, thus fulfilling the fiduciary standard you need to uphold as an RIA.

In terms of RIAs as consultants (as you described)...it sounds nice, but it seems like an extra layer of fees that the clients need to pay and I'm not sure it would be worth it. Then, the compliance around that is tricky, I believe.

If you're ever looking for a partner, let me know...would love to chat with you further. Best of luck!

This discussion of legal quandaries and requirements is one compelling reason in favor of Quantopian's plan. The way I have seen it described, Q will deal with all the legal responsibilities of the investing, and algo writers will simply provide a software license to Q (while keeping ownership of the algo code). This makes it very low risk, so that algo writers can focus on generating alpha, and they will be paid based on algo performance.

I think this same legal model could be used if Q were to offer the hedge fund to retail investors (although they have stated that retail is not in their plans.)

I believe Q will offer fund to retail accredited investors (not just institutional). But, I'm not aware that Q has a plan to serve retail non-accredited investors, at this time. So, I believe one of the only ways for a quant to serve non-accredited retail investors (using Q's platform) would be starting an RIA. Yes, more difficult than if one becomes manager in Q fund. But, possibly more opportunity if quant starts their own RIA. Depends what quant is looking for long-term.

Jeff, thanks for the clarification; I am still learning the terms of this industry. :)

I believe Q will offer fund to retail accredited investors

On https://www.quantopian.com/posts/june-contest-rules-update-its-all-about-that-beta-star, Fawce includes "family offices" as one of the target markets for the Q hedge fund. I'm not sure if I'd qualify a 1%-er family office as "retail" although I suppose the money could be directed by an individual family member and not an institution (although from a legal standpoint, it would be the family office directing the money, right?).

Having been with Q since the early days, their approach is to be selectively open regarding their business (understandable, I suppose, but not absolutely necessary). As far as I know, the decision to launch a hedge fund did not involve the user base, so while they see the comments here, I would not expect them to jump in and start chatting substantively about their thinking. They value feedback, but it's not exactly a public conversation between business partners (again, understandable). The sense is that at some point, they came to an agreement with their VC's that a crowd-sourced hedge fund would be their business, targeting the obvious large markets. They are marching to that directive. If Fawce & Co. sit down with their VC's and say "Oh, by the way, we have this other idea we'd like to play with. We have some users who want us to support more of a retail model, so we are gonna devote some of your precious venture capital to that effort, taking away focus from the crowd-sourced hedge fund idea" the meeting might not end on a happy note.

With the 'pure alpha' hedge fund concept, I do get the sense that Q has shut out some users from access to outside capital, which is too bad. But, it may just be the reality of the business; otherwise, they'd be trying to squeeze blood out of a turnip and just fail.

IB has good support for RIA accounts, and Quantopian can be used with them. Quantopian can be connected to sub-accounts to run algos. The usual rules apply - one algorithm per sub-account. You could certainly create an RIA and use Q as the platform.

Sometimes the honest answer is: I don't know yet. I like to be crystal clear about what we are doing, so I prefer to say nothing when I haven't decided yet.

To start, our fund will be for institutions. I don't know yet if we will offer something for individual accredited investors in the future. We won't be able to offer our fund to non-accredited investors under current regulations. Even if regulations change, I don't know yet if we will try to reach retail investors.

I do not understand the statement "Q has shut out some users from access to outside capital".

Hi Fawce, thanks for clarifying accredited institutional investors only, at this time.

In terms of "one algo per sub-account"...as someone trying to create a business model using Quantopian's platform for an RIA, what would be the cost to link an algo to a sub-account? Personally, I'm interested in trying to setup a very low cost RIA, so I'd like to please understand how much Q would charge the RIA? Is % of assets possible? Or, choice of per account monthly fee or % of assets?

Right now, it is free. I don't know yet what we will charge for platform use in the future. I do know that more assets on the platform is very good for us, so I would want the cost to work well for your business.

We have hedge funds and RIAs using that platform for their businesses already.

I do not understand the statement "Q has shut out some users from access to outside capital".

Hello Fawce,

The basic idea is that you are going after a very specific capital market, where a pure alpha product is required. But money is invested in many ways, right? Most mere mortals don't invest in hedge funds directly, but have money in stock mutual funds/ETFs. Say one of your users comes up with a nice long-only asset allocation algo that would be well-suited for a retirement portfolio? Is it in the game plan to assist that user in attracting capital? Or say somebody has a speculative algo that just tries to make a killing on a single stock or go bust? Would you try to match him with capital? Or maybe one of the many technical analysis based algos published and discussed on this forum without a market-neutral hedges? How could those be matched with capital? Others to consider are on Jess' list, http://blog.quantopian.com/5-basic-quant-strategies-implemented-by-the-quantopian-community/.

Make sense now? Basically, you've said, "If you can write a low-beta algo with high returns, we can match you with capital. Otherwise, we're not interested, at least for now." Or am I misreading?

Anyway, if the big money is with the pure alpha, and that's where you think you'll have success, then it's all good. But it does seem to rule out other approaches that could also land money in the pockets of Quantopian users.

Best wishes,

Grant

Long-only high beta strategies are certainly fine for savers and so on, but it's worth pointing out that beta is very cheap these days; SCHB charges 0.04%. If such a long-only strategy has some alpha, then one can simply neutralize your beta with a short and you're done.

The fact of the matter is that the majority of technical analysis strategies with high beta are effectively throwing darts at index constituents and as a result perform as one would expect random samples to perform. If the strategy does have alpha, you may as well attempt to isolate it, because if nothing else, you can then start to dig deeper into why.

If the idea is to attract some of the Wealthfront/Betterment crowd with AlgoBuilder derivatives or something like that, there's nothing stopping any enterprising RIA from doing so with Quantopian right now, today. They'll just have to make sure they can squeeze a living out of 15bp or something like that, but considering the platform and data is free, there has got to be some people thinking about it. I am not sure Quantopian would really want to step in to attract assets for such folks though, since in the grand scheme of things, the platform has the potential for hedge fund returns and hedge fund fees.

All this leads me to the following question. Currently, members of the Q community can link their IB trading accounts to their own algo that is currently running on Q. I presume no "accredited or not" check is involved in this case. So if this is merely extended to allow users choose to run not only their own algo, but potentially other developer's algo as well then it would address the issue I suppose. To do this, the algo developer needs to be able to share his "Buy/Sell Signals" without necessarily sharing the underlying code.

So what is the concern in this case.
1. Is it because running someone else's algo opens up the issue of compensation for the algo developers?
2. Is it regulatory, in that one (non-accredited investor) cannot choose to trade a black box algo with his own will ?
3. Is it because there is no current mechanism in Q to share algo signals, without sharing the code ?
4. The development cost of supporting this feature is too high ?

I would think when one is trading real money (using say some one else's algo) or even ones own algo, it would be justified for Q to charge a technology or maintainence fee if cost is the concern.

If such a long-only strategy has some alpha, then one can simply neutralize your beta with a short and you're done.

Then why wouldn't Quantopian be interested in long-only strategies? They could just neutralize the betas, right?

I can't speak for them, but I presume because they are trying to avoid writing order-generating code themselves, and the algo author who knows what she's trading is in a better position to choose the most effective hedge. They may ultimately still need to manage residual beta, who knows?

I don't really understand the objection to the beta restriction...

Raghavendra,

1. Is it because running someone else's algo opens up the issue of compensation for the algo developers?
   
   1. Is it regulatory, in that one (non-accredited investor) cannot choose to trade a black box algo with his own will ?
      
   2. Is it because there is no current mechanism in Q to share algo signals, without sharing the code ?
      
   3. The development cost of supporting this feature is too high ?
      Blockquote

1. I don't think there's much $ compensation in this, right??
   
2. I think regulatory hurdles exist in terms of Q needing to become RIA, I believe...and algo developer may need to become RIA possibly?
   
3. I don't think there is a mechanism.
   
4. I don't see the ROI for Q...unless I'm missing the revenue potential here.
   

Grant,

I'm of the belief that eventually every form of investing, and every investor segment, ends up being a lot more automated. For Q, serving institutional investors is the best place to start. That goal depends on us building the best platform for creating investment strategies.

thanks,
fawce

If the idea is to attract some of the Wealthfront/Betterment crowd with AlgoBuilder derivatives or something like that, there's nothing stopping any enterprising RIA from doing so with Quantopian right now, today. They'll just have to make sure they can squeeze a living out of 15bp or something like that, but considering the platform and data is free, there has got to be some people thinking about it. I am not sure Quantopian would really want to step in to attract assets for such folks though, since in the grand scheme of things, the platform has the potential for hedge fund returns and hedge fund fees.

Simon, this is very well said. I understand Q wanting to focus on the hedge fund for revenue/profit maximization. But, if we really want to DISRUPT the wealth management industry, then I think, in parallel, Q should keep encouraging quants to setup RIAs for non-accredited and accredited retail investors using Q's platform. There is the potential for thousands of Quantopian quants to start RIAs where these RIAs could collectively be managing TRILLIONS in 10 years. Then, some of this money could flow into Q's hedge fund. From a social responsibility standpoint, average investors need access to the algos quants are writing on Quantopian.

From a social responsibility standpoint, average investors need access to the algos quants are writing on Quantopian.

I agree 100%, but none of this happens if they run out of runway. I need to read that hedge fund book, so I know what is involved in becoming an RIA, I feel like running SMAs in IB with Quantopian should be technically very simple, but I am still worried about the compliance aspects of it all.

Simon, please remember that you can outsource compliance for a reasonable monthly fee so that you don't need to worry about becoming a compliance expert. These companies could also help you get setup.

I don't really understand the objection to the beta restriction...

@ Simon,

It's all about the target customer market. The game plan for Q is to go after the 'pure alpha' institutional market. But traditionally, my sense is that the Wall Street game is to basically fool traders/speculators and investors into thinking that they aren't throwing darts, that they have less risk than they actually do, and that whatever fees/commissions will be negligible compared to the gains. So, ethics aside, it seems like Q has ruled out the traditional market. I have to think that there are lots of people flush with cash who aren't interested in or capable of writing a Q algo who may have an idea they'd like to try, or would be willing to try someone else's idea for a fee (i.e. license or buy an algo from another user). In fact, regularly, there are folks who post to the Q forum looking for assistance. But there is nothing in the Q business model for experienced users to make a few bucks giving them a hand. One could imagine a decent Q coder getting 5-10 algos out there, at $100 per month each. That's $6K-$12K in supplemental annual income. That's not to say that the Q hedge fund is a bad idea and I kinda understand how they landed on it, but there are other models that potentially would be more open, entrepreneurial, and peer-to-peer. With the contest, low-beta, institutional money, etc. it sorta feels like Q is calling the shots, versus a wide open community-based organic effort. Lots of cool stuff on Q. But this part, not so cool, in my opinion.