Thanks Quantopian and fawce! This is excellent and indeed big news!

The move will certainly incentivise a diversity of ideas, retain the quant talents and IP vested in Quantopian, and really "level the playing field" when implemented globally!

Cheers

Karl

This is something what I thought about, too. This is the right way to go Q!
I have talked with some advisers about Q and will inform about Q to more professionals in the financial industry. There are infinite ways to combine data to find alphas in Q; unlimited possibilities and potentials!

Bravo. Way to go.

There are currently over 12 million backtests and 6 million algos on
Quantopian.

Those are impressive numbers. A lot of those algos just became valuable.

A list of potential clients general interests would be welcomed.

I am sure a lot of talented people here will be able to tailor old or new algos to their needs and specifications.

Congratulations and thanks to Fawce & to Quantopian. This is indeed excellent news !

One suggestion to attract more institutional investors; Open a new contest to meet the needs of an investor. Q should ask what investors want; expected return, risk appetite, any specific strategy, etc. Some investors might be interested only in long positions; some might be serious about low systematic risk, etc. Then, Q set the rules for a contest and Q authors will build algorithms that meet the specific needs of investors. We will provide more valuable and customized services/products for investors and it will attract more investors, more funds and more talented authors coming in Q.

Excellent suggestion James. An ongoing Q Contest and separate individual investor contests. More work for Q but then more reward for everyone :-)

@James M
Good comments. At the moment all we have as a guideline is the current Q need for the sickly neutral everything under the sun. We also need to know how to submit algos - currently they will be rejected unless they fit the neutral everything mold.

Or will Q just look at the results of our algos even if not submitted?

In which case I would prepare a few algos especially with the larger investor in mind and just leave them sitting on the server.

In the vain hope that sometime, somewhere, someplace, someone might contact me.

Thanks for the kind responses, everyone, we are very excited about these developments too.

@James: That is a great point and the challenges appear to be the ideal tool to provide targeted guidelines for what institutional clients need. So one could definitely imagine e.g. a long-only challenge.

Great news! Thanks for sharing.

great! feels like Q is moving to the direction of being a competition platform like Kaggle but only focusing on trading

hope in the future, there will be one-off competition not only using tearsheet but actually live paper-trading as well

and hope there will be a QuantCon

John/Thomas,

It would also be fantastic if Quantopian could mark algorithms as having looked at and rejected for licensing e.g. due to performance, risk, or correlation to existing strategies. The lack of a feedback loop makes it difficult to know if one is on the right track.

Excellent news! I am very excited about these new challenges, and I think that's what people are waiting for here.

Nice suggestions James, and I totally agree with the idea of having more feedback on algorithms.

Great news! This certainly opens up more opportunities for fund authors that are not within the Q specific framework. Kudos to Q management for expanding their horizon on behalf of the community. Keep it up and interesting!

Very good news! Exciting to see that Quantopian is working actively to find ways to bring value to both third-party investors and the community!

Great news! I've sent some concerns privately to [email protected], but feel free to address those here if it makes sense.

I'm also curious how datasets play into this. Would all algorithms be eligible regardless of dataset?

I'm concerned that with the axing of the paper trade feature, Quantopian will no longer have the infrastructure required to deliver position information to the 3rd-party clients at the earliest feasible opportunity.

Does that mean the positions that 3rd-party hedge funds license will suffer the same artificially-imposed 2-day delay as the Q Fund? I hope not, since not every hedge fund is operating under the same constraints as the Q Fund, so this potential design mistake would cost them some serious alpha and/or force them to pass on otherwise viable strategies. That would be a serious buzzkill, especially since the problem is easily fixed.

To be honest I am and always have been deeply skeptical of the whole "democratization of Wall Street" line.

Let me be upfront and say that I do find this forum useful and I am grateful for the provision of free price and fundamental data. Quantopian's return is presumably to get a whole lot of ideas for free, although I guess it does have to pay for algos which have been devised and squirreled away in secret. It seems to be the case that Q monitors all algos and can approach the owners if sufficiently interested.

But as I have said before, it all amounts to slave labour for the hapless punters who play here. The prospects of ever getting paid for any of the endless hours people put in here is, to say the least, remote.

Fair enough, Q has spent a lot of money and produced a lot of useful features.

However think how much more money they would have had to pay out if they had hired all these people to produce all these algos. Allbeit that 90% of the algos produced here by the hapless punters are not fit for purpose and hence valueless to Quantopian.

Its all rather like the Mechanical Turk project or freelance writing - or indeed slave labour at Amazon. Or these people like Medium who get people to contribute articles in the vain hope they might eventually collect rewards. The internet has spawned an entire economy where a whole bunch of determined operators like Comrade Bezos are able to get people to work for next to nothing.

Frankly, from my perspective, I would rather pay a monthly fee and be able to use Quantoipian to automate my own trading (as per Quanconnect) rather than chase the rainbow of ever getting paid for anything.

No offence to Mr Fawcett or any other Q worthy, but that is just the way I see it.

I was approached a year or so back by one of the world's largest Hedge Funds: send us you algos and we just might pay you for them if they turn out ok. To which, needless to say, I responded "no thank you".

I am sure no one here or knows me will be surprised by my attitude or my usual rambling nonsense, but to my way of thinking much of the internet works on the basis of getting something for nothing.

Just as you are doing?

I see some parallels with Google, Facebook, Youtube etc. If you build the #1 platform and then integrate a marketplace that can be huge.

@Tony Morland

Well, as I say, I would be happy to pay for a live trading service as per Quantconnect. And if I understand it correctly, Quantopian sifts through all of our algos anyway and so has the opportunity to licence anything you produce here.

So I guess we are all providing something in return for the platform, even if we do not wish to enter into the contests.

Don't forget also that we are providing considerable value to Quantopian in building a community. The systems and code snippets we provide all help to build an interesting environment which attracts more people.

That sort of contribution should not be undervalued.

So, no. I do not consider I am getting something for nothing.

However it is definitely NOT worth my while designing whatever systems Q's investors want. Particularly the current tedious neutral everything model. These sort of models simply do not fit in with my mentality or trading style and bore me.

If I were paid to design such models that might be different. But I would not care to design models for free which do not fit in with my own objectives and which I would not want to trade for myself simply "on spec". In the hope I "MIGHT" get paid for my work.

I have never been one for "competition", that horrible word seemingly invented by the ghastly Margaret Thatcher. Competition in this context involves working on something which does not interest you in the vague hope that you might get paid something for it some time down the line. Amount of payment and duration of payment undefined.

@Zeno / Anthony, i understand some but not all of your stoic logic here.

Certainly i agree with you that contributors of systems / code snippets / new ideas / interesting viewpoints do all indeed add a lot of value.

If you would like to pay for a live trading service as per Quantconnect, why not suggest this seriously to Q?

I know that you don't like market-neutral models and, for my own personal use outside of Q, i don't either. I do however learn a lot from working with them here in Q, so i don't find them boring at all, especially when they gradually tick away a few bucks into my bank account.

But as i understand it, Q is now intending to open up the scope much wider to include all manner of other practical models / types of algos as well. Soon there will be lots of other things to look at besides what you call "the current tedious neutral everything model". Don't you find that interesting?

Re your comment about competition & Big Maggie, i think i understand part of the reason for your even-more-than-usual grumpiness this week. As another of my British friends put it, BJ is like "a Trump that came out of the oven only half-cooked" :-(

Cheers, best wishes.

@Tony
Well at least we won't end up in the Gulag as we would have done with Corbyn...I despair of all politics.

But yes, it will be interesting to see whether Q comes up with new interesting models. Unfortunately for them it will require a great deal of work. For a start they will need to bring back something more akin to the old Help Documentation, and of course the current screen attached to back tests will need rethinking. Currently it blacklists anything outside of the very strict parameters which are presumably required by big Steve Billions or whoever the current backer is.

But you are right - I am being a grumpy, miserable old Scrooge. Who knows, perhaps the various Ghosts of Christmas will be visiting me later in the month in an effort to correct my ways?

Quantconnect has access to your code if you participate to competitions or put it on the alpha market. Tha alpha review team has the right to look at the code before accepting the algo. At least they are upfront about it.
What about quantopian? As users, how can we know if our algos are not visible to the system administrators, since they are on their servers?
I am not talking about legal stuff, but what is the technical impediment, if it does actually exist, which prevents the system administrators to read the code, and how we could ever discover it.
I think it is impossible for us to know, it is like emails. If the system adiministators want, they can read them all.

@Berbard I should not feed the troll, but should tell you this; your irresposible comments can hurt innocent people's feelings. Quantopian has been generous to share its knowledge and resources.

There is only one way that is free from "trust" issue: use your algo in your own system to trade and never put it in any website/system. Most people generally do not have enough money, knowledge, and resources. That's why we need Q.

I prefer Quantopians' way because I trust Quantopian team. Q said it cannot see my code and I trust Q because Q trust me. Q's way of picking stocks for its portfolio does not require to see any codes.

I think it is stupid to give your permission to others to see your secret code. Once others see it, it's no longer a secret. In Q, your secret code will be secret because Q cannot see it unless you give Q a permission to do so.

@James M, i agree and fully support your comments. Q is indeed very generous in many ways.

Fortunately there has not been as much negativity in recent times as there was in the past, when I became very tired of reading all of the nonsense that some people used to post about: "Can't trust Q; They are a scam; They don't really pay up any of the money they claim to, etc, etc". For my part, i have always found Q staff to be consistently helpful, i have never seen any evidence of impropriety of any kind on the part of anyone in Q, and all payments promised have always been prompt. So i have a lot of reasons to trust Q and no reasons at all to distrust.

@Bernard, you are correct that reading of emails by governments, hackers, etc does occur and, unless you are very skilled, it is indeed "impossible for you to know" if they did. But that is not a sufficient reason for you to make similar inferences about Q, or to post such unnecessarily negative comments here. But if you believe that your code is really so good and so secret that you are worried about someone in Q trying to read it, then either send it to Renaissance Tech or just enjoy the fruits of your wonderful code yourself and don't post it at all!

This started out as a very positive and encouraging thread by Fawce about a wide range of potentially great new opportunities for us, thanks to Q, encouraging us to ask more about that. Please let's not just drag it down.

@Bernard: I believe all code is encrypted and by default not accessible by anyone, including sys-admins/super-users. Sure, I’d expect there are master decryption keys available, but I’d also expect there to be “break-glass procedures” in place, requiring at least two people to access them, should a need arise. A bit like the keys to the nuclear launch codes. :)

My question is theoretical not moral. I am not asking if you trust Q, but from a theoretical point of view, how can we check the fate of our code. When I say check, I mean check it personally, without having to rely on someone else. My question does not imply any moral judgement on Q, I just wanted to know if I can check the secrecy of my code. This is a game theory problem if you want.
From my perspective, the answer seems to be not. We cannot check it; we can only trust it. This does not mean I do not trust Q, it just means I cannot check directly who has access to my code. IP is 99% value of quantitative finance, so this is not a small detail. If any breach of secrecy will ever happen(error happen everywhere, also in nuclear reactors), i do not think we will ever know about it(no radiation will be emitted..), unless something like cambridge analytica will happen, which I strongly doubt. These risks are inherent in any internet business models based on data, and the only solution is the one which was suggested above, i.e. not to have any account. In fact Reinassance does not have any Q account I suspect . Does it mean they do noy trust Q?
I think it would be interesting to come up with some new platform architecture where the code is store locally, and only some low level instructions are sent to Q servers where the data is stored. That would increase much more the IP protection, and put security in the hands of the user. Basically it would be like sending encrypted emails, encrypting them before sending them.

With love and trust

BP

@Bernard,

Yes, you are correct that in many instances in life: "We cannot check it; we can only trust it"
This applies to things like credit card transactions, especially in some countries where i have worked, and lots of other situations too.

In practice, how could you actually "confirm the secrecy of your code", unless it is only on your own secured computer, never connected to the Internet?
The reality is that other people are usually not going to be bothered even trying, unless you have an exceptionally high return like Renaissance, which was generated by the combined work of a team of more than 30 world-class PhD specialists. Most individual's people's algos never even get close enough to warrant worrying about it at all.

Personally i don't really care if anyone at Q reads my algos .... even though i trust them not to :-)

With genuine best wishes for your trading / algo writing. Tony

@Bernard -- This is getting way off topic and probably should have been its own thread.

We just have to take them at their word that they cannot access our code without our permission and trust that a current or former employee would blow the whistle if this were a deception. That would be such a scandal, it would be over for them if it happened. Nobody would use the platform any more, and they would have lawsuits on their hands. So you have to trust that they wouldn't take such a risk, especially since they don't need to. They have too much at stake.

As for building out an entirely new offline system with encryption to protect against stealing of datafeeds and to keep users from cheating, A) this would be too much of a distraction for Q to implement, even if B) it were possible, which I don't think it is.

@Virdian what you said was what I wanted to say :)
@Tony thank you for sharing your experience
@Joakim 100% agreed

Yes, it is getting way off the topic. Let's stop talking about any topic that is not related to this thread.

@Thomas thank you Thomas. I'm looking forward to a long-only challenge! BTW, your webinars helped me to learn and understand more of what Q is seeking. Thanks :)

Yes, agreed. I moved most of my off-topic comments to a separate thread, and i'm glad to head back to Fawce's original theme. Thanks :-)

In the spirit of getting things back on topic, I have more questions:

• What kind of timeline are we looking at? Will this launch in the coming months?
  
• How exclusive is this program? Is it open to any hedge fund or are there specific requirements?
  
• What is Quantopian's fee structure for this feature?
  
• Does Quantopian continue to act as an intermediary between us and the hedge fund (host our algorithm and deliver trades)?
  
• If so, how much of a delay is there between pipeline data updating and the hedge fund receiving the trades/positions? (As mentioned previously, it seems Q just killed the only feature they had for delivering these in a timely manner. Q should provide a function we can call from before_trading_start where we submit portfolio weights for the day, so that data can be transmitted to the hedge fund with minimal delay.)
  

Currently, Quantopian uses an automated system to review the output of all backtests and contest entries in order to identify promising algorithms to license. When we find something interesting, we contact the author and work with them to ensure the algorithm is a good fit.
If it is, we then offer a license agreement which includes the opportunity to earn royalties. This new arrangement with third-party investors would work very similarly. We would allow limited access to anonymous backtest results for them to analyze.

Based on this can you clarify:
- Does Quantopian 'read' and analyze backtests of even those algorithms that were not submitted to the contest? If so, that would be unexpected behavior for me as I thought that algos and backtests were protected IP of the creator, unless explicitly shared with Q. Why would Q do that and is there a way to opt-out?
- Is the plan to share such backtests with 3rd party investors before getting consent from the author or after?
- And finally if I wanted to license algos from Q, what would be the best way to go about that?

@Vladimir, have a look at the terms of use, especially this section:

Additionally, in accordance with Quantopian's efforts to support the creation of high-quality strategies in which to invest, we may review the results of your backtests and the performance and other "data exhaust" of your Content (such results, performance data, and other data exhaust, collectively, "Performance Data"). Such Performance Data includes, but is not restricted to, returns and risk measurements of any backtest executed with the Content. We may generate additional Performance Data or other information from your Content for the purposes of (a) verifying that software platform changes do not change or break your algorithm or (b) analyzing standardized performance results. You further agree that Quantopian may reproduce, store, and create derivative works from your Private Content or other Content necessary to evaluate your algorithm. Any testing, evaluation, reproduction or other use will be done with the spirit and intent of protecting your Content. We may also share anonymized and/or aggregated Performance Data with third parties. Any such external or third party sharing or usage of Performance Data will not be authorized in the extent that sharing or usage of Performance Data would reveal the details and workings of the underlying Content.

With regard to your Private Content, you grant Quantopian a worldwide,
sublicensable, assignable, royalty-free, fully paid, perpetual,
irrevocable license to use your Private Content as described in this
Agreement, including as discussed in this section. Further, with
respect to your Private Content, you represent that: (i) you have the
necessary rights to grant Quantopian the license to use your Private
Content as described in this Agreement; and (ii) the Private Content
does not violate any of Quantopian’s Terms of Use. The limited license
you grant to Quantopian regarding use of your Private Content
continues even if you stop using the Site and Services.

Thanks Joakim. Terms or not, this came as a surprise to me as I fully trusted that backtests I generate that I do not submit remain private. Why that happened is another matter and I guessed I was eased into it by repetitive 'algorithm privacy' mantra from Q.

Additionally, in accordance with Quantopian's efforts to support the creation of high-quality strategies in which to invest, we may review the results of your backtests and the performance and other "data exhaust" of your Content (such results, performance data, and other data exhaust, collectively, "Performance Data").

What if I am not interested in creating strategies for Quantopian to invest, and I signal that by not submitting those strategies to the contest? How is reading the output of my strategies in alignment with these terms, and ultimately Q's mission?

What if I am not interested in creating strategies for Quantopian to invest, and I signal that by not submitting those strategies to the contest?

You could store your algorithm code outside the platform and delete it here on Quantopian and we won't evaluate or consider it further, or use our open-source backtester Zipline https://github.com/quantopian/zipline for complete off-platform usage.

Could you let us know the full list of "other data exhaust" that Quantopian and 3rd parties are able to see on private algorithms that are not submitted to contests?

I understand being able to view metrics that is typical on a tearsheet (returns, leverage, turnover, sharpe, drawdown, etc) but being able to see end-of-day holdings data as John mentioned is surprising and troubling.

For one, since Quantopian has guided users towards developing medium term alpha strategies that execute near the close, seeing end of day holdings is basically same as being able to see all the trades. Revealing the holdings/trades of a systematic strategy is significant information and I think there are many users that would have a problem with that, especially to third parties beyond Quantopian.
I believe full transparency and some user control over the data exhaust is important and hope you can clarify.