Back to Community
hedge fund comments

I've tried to follow your hedge fund thinking, and at this point, it seems like you are needing a handful of sorta institutional-grade algos that you can fund at $5M or so each, versus an innovative crowd-sourced fund with widespread participation (but at lower levels of capital per algo). If you are "looking for very consistent positive returns month over month and year over year" then in my mind, you are looking for effectively high-yield bank CDs (return like (1+r)^n). One of your goals, as I understand, is to have uncorrelated return streams, but in the limit you are talking about, all of the strategies will be highly correlated in their returns--they'll all just go straight up! Yes, it'll be crowd-sourced, but in name only. If you could figure out a way to get perhaps 10,000 of your 45,000+ users on board, and get cash flowing to them, then it'd actually be something. There would be a relatively low threshold to get $5K-$10K of working capital, that would grow as real-money results flow in (or decrease as the algo falls apart).

By the way, in the spirit of transparency and crowd-sourcedness, do you have any outside investors lined up? Or are your VCs willing to supply an initial round? And I've never seen any potential investors post to this site? Where are they? How do we know that you are looking at the right market of investors? Maybe there is a cohort of investors that actually doesn't want consistent returns, and would trade risk for potential reward? Do you ever get inquiries along these lines, or do all of your candidate investors want CD-like returns? I understand that the "pure alpha" scalable institutional money is what you are aiming for, but I have to figure that there are 1%-ers who would like something more exciting for the speculative portion of their holdings.

26 responses

I'd also be curious if you've gotten any feedback from potential institutional investors on the black-box nature of the algos that would be in the fund? Are they comfortable with this approach?

Grant, you've got several questions bound up in those two paragraphs. i'm going to try to tease them apart.

First, I want to again say that I don't think that it's productive to think of the hedge fund as a high-yield bank CD. I've tried to point out before that the expected returns of those two instruments are different by an order of magnitude. I'll expound on that: No one would ever borrow money to put it into a high-yield CD. The structure of the market is that the borrow rate is higher than the CD rate; it's a losing proposition. But if we're building the hedge fund right, then it does make sense to borrow money, in the form of leverage, and invest in the hedge fund. We expect to have institutions that, for instance, put $100M in capital to work with our hedge fund, and choose to use 3X leverage so that the fund is actually managing $300M in investments for them. I don't see that happening with ETFs.

The next question is whether, at limit, will the strategies all be highly correlated. I suppose, ad extremum, we could be confronted with a host of algorithms that each go up exactly .05% per day. If that happened we'd have to develop other tools to figure out the degree of correlation. In practice, though, the world is a lot messier. Algos go up a a few points one day, down the next, and up again. An algorithm can be quite predictable over a long period, but that doesn't mean it's going up every day, exactly like every other desireable algo. The fact that they go up differently means we can evaluate their correlation.

Next I want to tackle is crowdsourcing. Your first sentence characterizes the choice as one of "institutional grade algos versus an innovative fund." I think that's a false choice. We're actually doing both. We are building a quant hedge fund from crowd-sourced algorithms. That is innovative, despite my lack of modesty in saying it so baldly.

I agree, it would be both neat and awesome if we could get 10,000 of our community members allocations from the hedge fund. Unfortunately, we don't have that many members writing high-quality algorithms that we can invest in (yet). You can download the contest CSV and look at the entries. There aren't 1000 good algos, let alone 10,000. If we get there, we'll fund them, and that will be the thrill of a lifetime for us all.

You don't quite say it explicitly, but I think you're saying that it doesn't count as crowd-sourcing if we use algorithms from a few dozen people rather than a few thousand. Does Kaggle count as a crowd-sourcing solution? I think that's a semantic argument, and probably not a very interesting one.

We have not signed any institutional investors. (Also, to be clear, we are not soliciting nor advertising for them either.) We've had informal talks with a number of people in the industry, and the response has been very positive. When you get down to it, a lot of these deals boil down to "what's the price?" We're doing a lot of work to lay the foundation for the fund so that when it comes to sign the deal, we get a price for the money that we like. That day isn't here yet.

I think there are a lot of parallels between the institutional money and VC money. Any of the questions you ask apply to VC and institutional investors alike. Why aren't they members of the community? I can't speak for them, but my guess is that because it's not their community. We're a community of quants, not institutional investors. How can you (or I, for that matter) know that we're building a product they will invest in? That's not something that is provable beforehand. We are, I assert, building something that has never been built before. You can never be sure it will work until it already has worked. What you can count on, though, is that our interests are aligned. We only make money if the community makes money, and vice versa. Are there investors out there who are looking for something like us, but not exactly us? There certainly are. However, we can't be everything. One of the best ways to make a startup fail is to lose focus and to try to do too much. We've talked to a lot of smart people, collected a lot of advice and experience, and we've made a choice. The response we've gotten supports that choice. If the data starts running the other way, we'll evaluate it when we get there.

Finally, your question about the black-box algos. It hasn't been an issue with investors yet. It's weird, but they are used to investing in black boxes. I'm not too worried about it, either way. As we negotiate contracts with algorithm authors, I see granting the right for someone to review the algorithm code as one of the potential terms of the contract. We will see how that one shakes out.

Disclaimer

The material on this website is provided for informational purposes only and does not constitute an offer to sell, a solicitation to buy, or a recommendation or endorsement for any security or strategy, nor does it constitute an offer to provide investment advisory services by Quantopian. In addition, the material offers no opinion with respect to the suitability of any security or specific investment. No information contained herein should be regarded as a suggestion to engage in or refrain from any investment-related course of action as none of Quantopian nor any of its affiliates is undertaking to provide investment advice, act as an adviser to any plan or entity subject to the Employee Retirement Income Security Act of 1974, as amended, individual retirement account or individual retirement annuity, or give advice in a fiduciary capacity with respect to the materials presented herein. If you are an individual retirement or other investor, contact your financial advisor or other fiduciary unrelated to Quantopian about whether any given investment idea, strategy, product or service described herein may be appropriate for your circumstances. All investments involve risk, including loss of principal. Quantopian makes no guarantees as to the accuracy or completeness of the views expressed in the website. The views are subject to change, and may have become unreliable for various reasons, including changes in market conditions or economic circumstances.

Investors trust fund managers. Well, investors trust fund managers' records. Without a record, you're nothing. Q-Fund won't have a record for a year or more. Not a record that major money will consider adequate for establishing a trust relationship at least. Now, what fund managers trust to create records (the algos) that investors can be sold, that is what is being tested here. Can 100, 1000 or more anonymous quants build algos that a fund manager can assemble into a system which will prove out through a sustainable, viable, vettable record? That is the experiment that Q represents. So far the record is, well, the record is non-existent.

Yet so far the diversity of algos seem to have hit an evolutionary bottleneck. A viable ecosystem relies upon diversity to ensure shocks to the system don't destroy the sustainability of the system. Lack of genetic variation spells evolutionary dead ends. The more and more restrictive the algo purity tests become the less and less resilient the system becomes as genetic material gets filtered out. Can the Q-Fund manager take such a rarefied species of algo and build a robust, stalwart profit system? We'll see, in about a year or more.

Not been done before? Not that has been publish publicly, that we've all heard of -- so far. Maybe not. I can't imagine the likes of Citadel or its ilk not having serious quant competitions -- internally -- that would have accomplished effectively the same intent that the Q is shooting for. Hundreds of top-notch quants all vying for a spot in the limelight? Yeah, it's been done before. But only the Q will, that is may, provide a window into the process. For all we know though, when the million$ hit the Q-machine the curtains may fall and all the winning quants become hired guns, the website gets internalize, the forum goes dark and the Q's VCs insist on zipped lips and high walls. Money trumps all good intentions.

(Oh, happy Saturday Jonathan.)

Say Q could offer a product that returned X% year-over-year, with zero volatility, and could almost guarantee that investor's capital would be safe. It'd smell like a CD, right? And it would be very attractive to investors. So, the more the Q product behaves like a CD, the better. In fact, if I look at Pravin's algo on https://www.quantopian.com/posts/the-process-of-naming-the-september-prize-winner, it has some choppiness, but it basically follows a straight line, just like a CD would. It looks like noisy compound interest to me, (1+r)^n, with some variability in the return, r. You are rewarding algos with CD-like returns, and if you build a fund out of a bunch of algos with CD-like returns, it'll look even more like a CD. Am I missing something? Wouldn't the ideal pure-alpha fund look like a high-yield CD (returning >10% per year, adjusted for inflation, etc.)?

I'm not exactly sure what the world's first crowd-sourced hedge fund would look like, but it doesn't feel like you are on the right path. As Market Tech points out, it seems you are asking the crowd to develop algos similar to what hedge funds must ask their R&D staff to be working on all the time. So, there's no innovation there. The innovative part, I suppose, is that you are getting the R&D tools out to the masses, at low cost and at a global scale. My impression (perhaps wrong) is that you are being pretty darn conventional in what you are asking users to develop with those tools. The risk is that you'll end up with the same thing that every other hedge fund produces. But to Market Tech's point, maybe that's what you desperately need at this point. In a year or so, you'll need to put together a slide pack that shows that you turned $25M into $27.5M, and that all indications are that the trend will continue. The easiest way to get there is to put $5M into 5 algos that are low-risk, conventional approaches. The question is how do you engage all 45,000 users and get to $10B? Then you'd be doing something new.

Regarding participation from the money side, you must know where your first injection of capital would come from (sounds like VCs). Why not eliminate the middle man, and just have them tell us what they need to put up $25M (or whatever the number is)? Or at least share the slide packs that you are presenting to them.

...negotiate contracts with algorithm authors, I see granting the right for someone to review the algorithm code as one of the potential terms of the contract.

This doesn't sound like a approach that will scale to 10,000 crowd-sourced algos, with "managers" coming and going on a daily/weekly/quarterly/yearly basis. It sounds more like the Q platform is a recruiting tool for a handful of highly skilled quants, who can write high-capacity institutional-grade algos that you can pour money into. In my mind, if you are truly crowd-sourcing, there would be no way to negotiate on an individual basis (although I could imagine a standard form, with the choice to share code or not as a button to be clicked).

your relentless skepticism is out of place with the rest of the board

Hello Tybalt,

Believe it or not, I'd like to see Q succeed wildly (Dan and others know this). If there is something I've said that you think is bogus or if you have a different spin, feel free to comment.

Grant

As an entrepreneur I can tell you that you'll never get anywhere if you ignore or try to shut down the Grant's of the world who ask the probing questions and apply some rigor to the thought process. That type of thinking is what gets you ahead in big organizations (I did work for the government for some time), but it's the kiss of death to an entrepreneur and the reason why startups can take on and win against big organizations.
To add some skepticism of my own, I actually took a several month break from Quantopian because of the exact phenomenon Grant and Market Tech are describing. Tightening the parameters so much that they end up looking for this holy grail of 10% nearly risk free yield in a 1% risk free rate world using nothing but equities. And not even all equities because some are apparently bad. Trading on minute bars. With completely unreasonable commission rates. And a fill model that doesn't match reality. Maybe they can find one or two algorithms that can achieve this, but so far they haven't. Even if they do, as Market Tech points out it will likely be a case of multiple discovery of the same underlying phenomenon, which doesn't give you robustness.
I don't see that Quantopian has done any investigation of the portfolio effect of multiple, completely different algorithms that don't necessarily independently meet these crazy, must be long and short, must trade all the time, must have super low beta... type requirements. Not the anti-gaming rules, I'm all for that, the "ensuring CD like returns" rules. Not only has Q never mentioned doing this type of analysis, but there are only probably two people on Q's staff who would have the background to do it. It's still a software company, which is great for me because I get a ton out of the free platform. However the software part will only be around until the VC cash is gone or the company because self-sustaining, so we all have a vested interest in them reaching that goal.
The portfolio concept is fundamental to modern finance, and its a shame that Quantopian is shutting down such a promising avenue in search for a much more elusive goal that has a far lower chance of success. It's completely throwing away the concept that it's far more likely that you'll find 10 algo writers whose independent strategies, when combined, produce the holy grail, than the chance that one algo writer will strike on all 10 of those concepts and put them together in one algo. Or the even smaller chance that you'll have tens or hundreds of individual algo writers who can do this. The idea that when you combine an algo with a huge positive beta and one with a huge negative beta, you could end up with very low beta, for example. Even better you combine 5 very positive and 5 very negative beta's. The idea that an all long fund and an all short fund, when combined, could produce the same results as an enforced long/short fund.
Bottom line, as I've been saying for about a year, Q really doesn't know what they don't know about finance. They're all super smart, no doubt about it, and it's a trap I recognize that just because you're smart in one area doesn't mean you automatically are in another, even if its related. The danger here is that they'll paint themselves into a corner where a bunch of their promising algo writers leave, or at least take a break from the site, they don't find the holy grail with what remains, and they run out of money before they can pivot. I sincerely hope this doesn't happen.
I don't want to be the complainer without suggestions, so here are a couple. Maybe now that we've stepped away from monthly contests, you hold two 6 month contests, once with all the crazy parameters and one with only the anti-gaming rules? Even better, you put together a platform that allows "crowd portfolio managers" to pick from any combination of the contest algo's to put together an algo that itself is in the contest. That way you don't even have to hire someone with portfolio selection experience, you can use the crowd which is in line with the company's DNA.

Thanks Kevin,

Yeah, my concern is that Q is on a "me too" path with respect to their business strategy, which means that they'll have nothing in the end, if they are looking for strategies that match what is already out there on the market. There needs to be a significant differentiator, which just doesn't leap out to me (but I'm not in the financial field, so maybe the hedge fund industry is shaking in its boots).

Grant

Hi Dan,

So do you have a commitment of $25M or so for the hedge fund, to get it off the ground? And if so, what do you need to show to get the money? Or is this whole thing still a dream? Market Tech is likely correct that for any big money, you'll need something like a 3-year track record, so we are a long ways off to get to any scale. You seem to suggest that you haven't started engaging institutional investors (since they'd say "Let's see your track record" and you'd be lucky to get a free cup of coffee from them). So, I think there is a consistent theme here.

What's the story? You are asking community members to put time into writing super-duper algos. Do you have money lined up? Or are we still in the "pull ourselves up by our own bootstraps" phase?

On a separate note, do you ever get inquiries from individuals (or maybe "family offices") with money they'd like deploy to an algo on Quantopian? If so, how do you respond to them? Why did you decide to go the hedge fund route, versus more of a peer-to-peer model, where Q acts as a facilitator/intermediary rather than "The Man" (in this respect, you've taken a conventional "Wall Street" approach versus something new/innovative/open)?

Grant

I work for a large hedge fund and amongst other things am involved in seeding new managers. so on the condition that I will answer what I can, please feel free to ask me any questions.

Thanks Tybalt,

I guess the main question is does Q have a viable business strategy with their hedge fund concept, as it has been laid out? If there are shortcomings, what are they and how could they be remedied? Will they end up having any advantage in the hedge fund marketplace? Will their product be unique?

This is not to be overly pessimistic, but if they are headed toward a cliff, then they should realize it.

Grant

One thing to consider is the product, the story, and who it will appeal to.

Institutional investors have v different metrics to retail investors. Both are valid ways of attracting capital.

Another is the type of fund they wish to launch, how often will subscriptions be allowed (daily, monthly etc), a 40 act fund, a UCITs fund, an offshore fund?

What they need to do, and I believe they are trying, is putting together a suit of super safe algos that they can show to somebody with capital to deploy and convince them they will not lose al of their money.

This game is about building a real track record for 1 to 3 years, and then growing assets at various intervals that opens you up to different tiers of investors.

Whether I deem their product viable or not is not really the question, the fact they have raised finance a few rounds says there is the appetite to at least TRY!

IMO they will have no problem raising a few million to deploy using the selected Algos. I know of a lot less sensible ideas getting money!

Thanks Tybalt,

Yeah, the "putting together a suit of super safe algos" is the part I'm actually worried about. Does it sound any different from what you would do in the hedge fund you work for? If they end up making and selling the same lemonade as everyone else on the street, then what's the advantage? I thought the safety part would come from pooling together lots of algos, with a little money deployed to each. Or maybe I don't yet grasp the crowd-sourcing part.

Grant

Is the difference not simply that they are drawing their talent from a population which is different from street professionals? Why is that not difference enough?

I feel like you have some expectation of them to be patrons of the quant community (beyond offering free tooling and data), when I think their role might be better described as talent scouts...

Yes, true. I just can't quite put my finger on what will be different in the outcome. Maybe Q's costs will be lower, so they'll be able to offer a better deal? Or the performance will be superior? Or the risk will be lower, due to greater diversification? Or street quants will be more creative that professionals, and uncover unbeknownst gems? In the end, as a start-up, Q has to have a unique edge, right? What will it be?

To understand a new or foreign paradigm I often rely upon finding an analogous situation. For instance, what if we were talking about paintings as art -- Artopian.

Initially, Artopian would offer free easels , paints and brushes to any who care to participate. They claim that painting should be democratized, that Artopian will be the advocate for the underprivileged artist. That they will build a studio where they will take the best impressionist, expressionist, realists, abstract, etc. paintings and display them for the world to see and for art connoisseurs to peruse and to buy. What a great idea!. They will certainly be a unique art studio.

But time goes by and because the funders of Artopian, for whatever reason, feel that they would rather sell only realistically rendered painting that they change the offering to their open studio to only allow realism paintings. Painters of impressionism, expressionism, and abstraction will not be allowed to display their art in the studio. They can still use the free tools and supplies, but they are banned from the studio.

Bummer. And we thought they were going to be the champion of young artists everywhere. Now only artists of Norman Rockwell'esque style will be shown in their studio. So much for a democratic representation of artists.

Ehhh, I'm not much of a realism painter. Forget it.

And we thought they were going to be the champion of young artists everywhere. . . . So much for a democratic representation of artists.

This is the part that I think is a problem. At the end of the day, their purpose is to sell art, and that is what finances the free easels and paints. If realism is what sells, it's churlish and frankly ungrateful for an impressionist to begrudge Artopian for not trying to sell their art. The photographers might appreciate using the free easels as well, yet don't feel so entitled that they are upset Artopian doesn't sell photographs. It's the impressionists and other non-realist painters specifically that feel persecuted or dismissed, arguably because they have unrealistic expectations.

Boring! "Announcing the Quantopian Crowd-Sourced 'Me Too' Pure Alpha Fund!" Doesn't sound very sexy, but with a cheap frame, maybe it'll sell at Wal-Mart and Target.

arguably because they have unrealistic expectations.

Those expectations having been set, originally, as a broad representational spectrum of art styles. And only later narrowing down to an exclusive method.

Expectations change, sure, but when they get set up front, and changed thereby filtering out most participants, one can only expect to see some rebuffed responses.

Oh I am not surprised to see them, I just disagree with them! :)

Reportedly, the bar ain't all that high to get Q hedge fund money. Sharpe > 0.7, 7% annual return minimum, beta +/- 0.3 or less, long-short in some fashion, and maybe a few other constraints. 10 year performance that isn't too ugly. There may be hope for you, Market Tech. Why don't you whip something up? Get out your paint brushes and easel, and slap some splotches on that canvas!

I'm afraid I would only embarrass myself. Besides I only know how to paint impressionist style and as we know, that's not what cuts here here anymore.

Market Tech,

You sell yourself short! Although, if the payout is unclear, it may not be worth the effort. In the next 2-3 years, there could only be 5-10 manager slots for Q hedge fund capital, unless there will also be seed money at $25K-$250K for more speculative algos that aren't institutional grade "stat arb" ones.

Grant

I don't see it as anything to worry about. Safe means you stay in the game, and while you stay in the game you earn your fees and can research more and more algos. :)

Tybalt,

Yes, if there is real money just waiting to put money into Q strategies that meet certain criteria, then by all means, strike while the iron is hot. And maybe there is no path to money for the Market Tech's of the world. Maybe it is obvious to you and others in the field that pure alpha CD-like strategies will win the day. But if it is obvious, then there is a risk, since it'll be obvious to everyone in the industry, and Q won't have a differentiated product (although Q may have more flexibility in fees charged to investors). On the other hand, I guess there is a real potential to very quickly get to 30 or more high-quality strategies that could then be mixed and matched into a much better offering than any other start-up hedge fund could pull off, with only a couple smart guys/gals and a few support staff. In the end, I guess that this is the Q "big idea"--scouting for talent globally, and being able to employ them quickly and at low cost. The threat to Wall Street is that talent scouting and acquisition has been innovatively globalized by Q on a grand scale. Cool. Nice work, Fawce & Co. See, I ain't so negative.

Grant

Honestly the differentiation is the story to begin with, the outperformance, or not, will come later. The story is enough to get people interested I am sure.