A quick glance at QuantConnect's forum, and I'd say that I like Quantopian's better as an initial entry point. QuantConnect poses too much work to drill down into each sub-forum. Just tell me what's recent and active. Quantopian needs to improve their forum, but it also needs to stay approachable, as it is.

Research platform?
Sponsored conference?
Mascot (for Quantopian, see https://www.quantopian.com/assets/headshots/headshot_roge-cdb654b5ce613f1c0a7f0651098cee2e.png)?

I tried QuantConnect last year when I was trying something with tick data and second bars. Their backtester is amazingly fast but not as robust as Quantopian's.

I eventually moved to Quantopian because I can rely on the results and more importantly I can do more with Python easily than with C#.

Python > C#. I think that covers it.

I also stick with Quantopian because of python. It is good to have options though.

Wow QuantConnect is really evolving I checked them last year and they didn't look so nice. Now they have SMS notification, Mobile interface, and the platform is open-source. But community he is probably better and we have the nice competitions.
Although python is easier to start C# is a more appropriate language. First it is faster, easier to structure your code, debug.but it is mainly windows based language.

It would be interesting to compare result of two algorithms one here and one there. the one with worse performance should be the more realistic.

Python is not necessarly slower than C#. Pypy outperforms C# in benchmarks, and apperently Q uses cython, which can perform at C like speeds (granted it will just be python without the interpreter if unoptimized, so...).

The main issue with structuring at Q is that we can only have one file. I would be so happy if they had a repo feature (say I could have a directory on bitbucket or github that could run and perhaps be edited here). Putting tonns of code in C# isn't very nice looking either.

If you want to compare, you would do the same algorithm, something which rely on the market data several times a day and not just rebalance the stocks once a month on both sites and live trade it. The site that comes on top is the best.

QuantConnect have no access to morningstar or any other market-wide fundamental data. Huge pluss for Q.

The fact that Quantopian is fine with us discussing how they compare with their competition on their Community board, sometimes in ways not entirely complimentary to Quantopian, says a lot.

Super open, in public and in private. Thumbs up to Quantopian for that one! Although I'm still waiting for them to publish minutes from their meetings with VC's...

Yep same here Quantopian Rocks. Python is the way to go. I just really would like Quantopian to support Options, Forex and Futures. Also I have notice that the Backtest from QuantConnect it goes super fast. It will be nice if Quantopian could speed up the farm. But so far I am not planning to switch :D

EG

Looks like there may be some competition heating up in the open algo space.

https://www.quantconnect.com/blog/future-of-quant-trading/

The nice thing about using .NET is that theoretically one should be able to use any of the .NET languages (even Python) to write your strategies. I think that this maybe on the docket there at QuantConnect.

Fundamental data is a great benefit here at the Q. But apparently it's only a money problem. If you can cough up the cash to Morningstar you TOO can get access to their data. There are some differences, obviously, and I don't think one must use only one or the other. As both support Interactive Brokers, one could trade one's account on both platforms. I mean, why not?

Being a C# fan since dotnet 1.0 I'm going to have to exercise QC a bit to see if they can provide some of the same services the Q does. So that's two mostly open, mostly free, web based community focused trading platforms. Is this now going to become a thing?

Competition is good, both platforms are in development the they both have different focus I think one is evolving for the Hedgefund end user and the other for quants. Only if it was Java, I don't get why people get attached to .Net and Microsoft.

Unless QuantConnect is planning to set up a competing crowd-sourced hedge fund, I would think that Quantopian wouldn't view them as a direct competitor, right?

From what I read, QuantConnect will introduce you to $ if your strat is good enough. Not sure what money or how effective or if it's ever been done. But yeah, various differences in the end goal. The Q gave up on PaaS, (or more likely the VCs said, your valuation as a PaaS is not enough -- figure something else out.)

From what I hear Microsoft is going to open source all of .NET. So .NET won't technically be a MS only play. Plus QC is based on Mono so they can use all the cloud services to run their engine.

Interesting. I wonder if QuantConnect is going after the same institutional 'pure alpha' market as Quantopian, or if it would be more of an online betting model ("I'll put $100K on horse...I meant algo #7."). If the latter it could be a good complement to Quantopian. If algos could be treated legally like race horses (or slot machines or whatever), there could be a huge market, right?

Such a betting paradigm might fly outside the U.S. The UK for instance allows a more flexible "trading" environment. But here, one might think that the market exchanges and major financial players (and their lobbyists paying off Congress) intentionally suppress creative investment schemes. Take for instance the national moratorium on online gambling. Although not federally a crime, the laws surrounding online gambling effectively make it illegal. But maybe if you setup shop in the Cayman Islands and allows your investment partners to place wagers on various algos...

In the US though, at least if you're cooking with small potatoes, you have to tow the line pretty straight. Jamie Dimon and the Whalers (and their ilk) though seem to cook by different rules.

Here is a power point which outlines this exact question

http://www.slideshare.net/TipRanks/quantconnect-vs-quantopean

C#, Python and F#. Not just C# at QC. Additional resources are hands down more robust at Q. I develop on both platforms. .NET infrastructure and C# latency smokes any iteration of Python (Cython, Iron Python, etc..). Development is much easier in Py. I develop and test in Python then recode in C# for live deployment. That should explain a lot.

Python itself is nice, but it becomes a powerful research tool only when combined with numpy/SciPy/Pandas. Are those frameworks integrated on QC?

Is there some sort of Research environment on QC? Last time I checked there wasn't one.

Do you make your development on a web IDE or can you develop locally on your machine?

The 3 extra years of back test data is important. Is it fee of glitches from splits?
(Ideally I'd like data going back much further even if it is daily bars.) Does QC offer better access to things like VIX? To use VIX in Q is awkward and error prone and from what I have seen
does not work in live trading. (Quandl updates too late for Q to get the previous day's data).

Why C#.... I'm a C++ programmer. If it was C++ I would definitely try it.

On QC it's your choice.
The research environment has evolved and is very fast, based on LEAN.

@Greg - Yes it's scrubbed

If you prefer working in Py, there's no reason to consider QC. QC Python and F# are in Beta and they have A WAYS to go. [update: Python has evolved so access to higher resolution data is an advantage]

There's no "competition" really. No case for it. Both platforms are beneficial to users who have specific goals. On Q, you're obviously not developing HFT or even Medium Frequency models. Models on Q are not latency sensitive by design.

Data Resolution on QC is important for strategies that, while not latency sensitive, would still benefit from improved latency.

US Equity: Tick, Sec, Min, Hour, Daily
Forex: Tick, Sec, Min, Hour, Daily

Quantopian's "pipeline" is a powerful resource for dynamic universe selection. No equivalent to this or AlphaLense on QC. For research, there's nothing that comes close to Q but for a workflow that wants to take that research to live trading, Q is restrictive.

Having the data available for offline analysis is important. There must be some "strategic" reasoning why Q chooses not to share our back test trade logs with us. On QC, you can download your trade logs and run them internally for further analysis, you can install LEAN (zipline equivalent) and use it as your trade execution engine internally.

The primary value in QC is the ability to do local development and you own your data for offline analysis. QC's trading engine is available as a Git fork like zipline but as far as I know, zipline can't be installed internally as an execution engine.

I like to conduct R&D in Python because of what Quantopian has created, which is priceless. I look forward to having access to raw TICK data on Q one day.

The natural progression is towards improved execution speeds, regardless of "frequency". Even if I re-balance like a dinosaur fund manager, speed is an area where I can harvest Alpha, period. Python will never achieve the latency of C#, and C# will never match the latency of Linux/C++. Python tuned for daily resolution or even minute resolution = low frequency trading.

It should be very simple to port models to any API for trading. If Quantopian allowed this, it would be a game changer.

Sorry to say...

QuantConnect is absolutely aweful for beginners. Despite all their effort in their marketing strategy.

Here is a good review of both : https://robusttechhouse.com/quantopian-vs-quantconnect-review/

The biggest problem with Quantopian is speed. Backtests are increasingly slow and you can not purchase a dedicate machine to run your code on unlike with QuantConnect.

There are some other players like Quantiacs and AlphaLabs

Ty 4 link.

Margin and accounting for it (or not) are key in my opinion and I'd appreciate anyone who can shed more light on this ...

On Thu, Sep 3, 2015 at 6:43 PM, JB from QuantConnect wrote:

[snip],  

Yes we do -- sort of,  

With Quantopian -- you can spend more than you own (infinitely, beyond leverage even). They are PnL/StartingCash but the starting cash is meaningless.  

With QuantConnect -- we're 99% realistic with our modelling -- so if you have 2x leverage, you can maximally get 2xCash margin. We are PnL / Starting Cash (but can't spend more than cash).  

We even model margin calls from the brokerage -- so if you're leveraged, and suddenly the market collapses, our paper trading brokerage will liquidate your position :)  

We still use Starting Cash since this is the total amount of risk in your brokerage account. Maximum spent is strictly speaking the actual return --- but that assumes the rest of your cash balance wasnt earning interest. Maybe your cash came from an investor who expects a return on his total cash.  

JB

One of the best things Quantopian could do is provide set_nonmargin() for nonmargin account testing. Quantopian members cannot simulate what will happen on Robinhood for example or those without margin accounts on Interactive Brokers. The lack of accounting for margin results in floods of irrational exuberance, we see people optimistic, going live then falling away, not a good sign.

I've used QuantConnect a fair bit and it's ok. I definitely prefer C# to python. I've been drifting towards Quantopian however. It seems stabler and there seems to be more opportunity to access capital. I think Quantopian should offer C# coding and maybe other languages if they want to stay competitive though.