Hi Vasily, great analysis. I agree with every point you made.

I use Quantopian. Like you, I do not intend to participate in any contest for the simple reason none of my trading methods will comply to such requirements.

However, it is still a valuable free strategy development tool with minute, split, and dividend adjusted price data with fundamentals. It includes delisted, bankrupted, and merged stocks.

I look at it as a library of trading programs, code snippets and ideas on how to do and not do trading procedures. When you develop in isolation, it is good to have your ideas challenged or corroborated in some way.

And like you, I do not like to have my programs on their machine. It is why I water them down, or erase them after execution. I know it is not enough, but...

i gave up on q the night they gave me the wrong directions to the meeting in dallas. i think the real jewel of trading software is amibroker myself.

IMHO committing to a highly volatile and leveraged marketplace like eMinis or any other futures for that matter (or even US equities) in automated fashion and without utilizing non-linear instruments (I. e. options) is akin to driving on interstate with only an accelerator pedal but no steering wheel. After all, you know you wanna get _there_ and not some silly sideways ...

i have tried myself to use q but the lack of functions leave you wanting for a more complete platform. it's a cool concept but not something you could invest time and confidence in yet. i never liked or trusted the backtesting at all, but i guess time will tell watching the funds performance from the sidelines.

It's interesting how vogue has changed in the last 6-7 years. I do remember that when I insisted on any quant strategy been based on an economical foundation, I was constantly laughed at in, say, 2010-2011. I also do remember how I was suggested Quantopian as an example of a "pure quant" platform. It's really funny that now they post literally my requirements to trading systems.

Guy R. Fleury, yes, surely Q brings value (what I explicitly mentioned in PS). Probably my essay was a little bit too critically-biased but the intention was to say that Q should screw down its requirements for its own good. E.g. this ridiculous requirement not to correlate with others can be reformulated something like "if some good strategies correlate strongly, a capital will be spread among them".

BTW, some of my critics was (partially) retorted in Forum on Q (https://www.quantopian.com/posts/using-zipline-locally-vs-quantopians-platform), e.g. "Quantopian will not look at an algorithm (or other content) without permission from the owner". Personally for me it is still not enough, however, it is better than nothing.

Anton Pavlov, Great metaphor about (lack of) non-linear instruments and driving without a steering wheel!

It seems to me that most of the Quantopian's requirements are taken from a schoolbook for FRMs or alike. They are taught that if you have uncorrelated strategies then you mix them up in a portfolio and sell to excited laymen as a "structured product", or how they call it today. So, they are taught to work with uncorrelated strategies, whatever it may stand for (it's always fun to ask them how they calculate correlation — most of them even don't know the definition). Same with other requirements.

Vasily, yes, I agree. Like you, I don't find it enough to say: “we will not look at your code without your consent”, when all the facilities are just a few keystrokes away to do so. I am apprehensive of that. Corporate policies can be bent so easily.

Yes, there is intellectual property at risk, even if they currently say they won't look at the code unless requested by the author. We have no way of checking on this. But, I can still go with that for the moment and take protective measures.

Quantopian, like any other trading software, can be a poor, good or great tool. The difference will come from the user's ability to extract what he/she wants from it. I have already cloned a few strategies there, modified them a little toward my own trading methods to make them literally fly. So, for me, it promises to be a great development tool. In the months to come, I want to transport some of my best EOD systems to the minute level. Doing so will improve performance.

I say Quantopian is disconnected from the market they try to target, which the list of issues in the article illustrates. They recently got a 24M$ funding from Andreessen Horowitz, and can only say WOW ...

Soren, you could add another WOW. The also got 250M$ from Point72.

Actually, there is a counterpart version of Deng Xiaoping' Black Cat White Cat saying: it does not matter it is a black rat or white rat. Provided it is not caught by the cat, it is a good rat!

I took a brief look at Quantopian, and agree with your points. It is an interesting concept, but one that in its current form seems more suited for university students looking to "play around" with quant development as a learning experience, rather than something a pro would use as a resource.

Tried and then quit!

Their frugal limit of the CPU time your algorithm is allowed to consume rules out any deep backtesting on large scale. However, as their fund grows they may ease the current restrictions.

They should have several market-corralated algorithms and I guess they are looking for counter algorithms to balance them. They are clearly not interested with the similar algorithms that they are currently using.

There is no free lunch; they provide FoC data, provide easy platform and many webinars. I reminds me Richard Dennis's turtle set-up.

Serious traders should focus their own properitery solutions/products.

Vasily, I think you make a lot of valid points here, especially the one regarding uploading your trading strategies. If your strategies are any good then sharing them with any company and its employees is crazy, no matter who they are.

Trading strategies have to be kept a secret, they should never be shared with anyone outside of your most trusted group. There are many other risks to cloud based platforms that I could think of that go beyond those listed in the terms of use.

IQBroker is a new and free desktop alternative that can connect you to quality historical data and portfolio level backtesting, optimization and algorithmic trading using C#, VB.NET, R.NET and F#. You can backtest your trading strategies privately and never share them with anyone.

It is a heck of a lot easier to teach trading to a developer than it is to teach programming to a trader. Therein lies, I believe, the greatest challenge the Quantopian model will face.

Edge and alpha are extremely elusive buggers, and when you throw on top of it the requirement that users have to 'code,' whether it's Python, PHP, Ruby, C# or whatever the next hot language might be to emerge, to express and test ideas, the pool of trading talent is lowered drastically.

We've be building expert trading communities (whether it be fully automated system developers or rules-based, discretionary, independent retail traders) for over 15 years, and the amount of toads we have to kiss to uncover a single prince(ss) is staggering. And that's with our users submitting signals using whatever tools, platforms, or languages they choose.

Roderick, yes. But, the programming is not that hard. You have one big do while loop where you have a stock selection process to which you set conditions to buy or sell. That's it. Therefore, any decent programming language could do the job.

Look at the Quantopian problem from another angle. A software developer can cost $50 - $100k+ / year. And you are not sure if something useful will come out of it. While Quantopian has over 100,000 volunteers that will code, on their own time, at practically no cost. And probably get something useful. They will be able to select the cream of the crop. No need for interviews, no HR, no responsibility. I see it as a winning formula for them.

The individual participating in this is another story. The competition is fierce and the tools made available lacking. But still, it opens up opportunities.

I think the people who are capable of making algos already make them, for themselves​. It us a nice site for students to play around with but not for professional grade trading. they do not allow you to connect to your own database which is a major problem. You are better off writing your own system than trying to jump through their hoops.

There is an old saying: those who cannot make money, teach! In order to provide a real workable solution one has to define the problem right! Regarding oil, for short term and long term we are dealing with political and supply/demand. Political drives the psychology behind ( volatility), and supply/demand drives the trend! Needless to say a lot of oil smuggled from those oil rich countries and members of OPEC way below the OPEC price itself. Quantipian is nice but they couldn't define the problem right, yet!

A paid trader is part of the staff that processes orders he is forced to process . It is not a day trader trying to make a buck. If you trade your own capital and process your own orders, what do you bring ? You can exchange speed against risk or take a position in the limit book. A limit position is a short option you did not get paid for. I traded on the staff on Amex and Nyse. The money comes from the huge flow of orders we get. The place we work in is dirty, noisy, and our income is stable. Can anyone comment on this ?