It's a hosted algo trading platform. You write your algos in Python (which is just an awesome language, imho), and you can backtest it against up to 10 years of minute data. Right now, it's limited to US equities, and like I said minute data is the best you'll get for now, but the product is still in beta. It's also FREE. Ultimately, they plan on extending it so that you can have it execute trades for you, which will be the paid portion -- sorta like the NinjaTrader business model.
The in-browser IDE isn't too bad either. It's also nice to not have backtesting taking up resources on your own machine (esp for those of us who have to sit in VirtualBox/Fusion to run our software).
The following user says Thank You to vbt101 for this post:
I work at Quantopian. Thanks for checking us out, and thanks for posting about us.
We're trying to make it very easy to use algorithms for trading - basically, easy to be a quant. Backtesting, data, trading, etc. are all a real pain when you get started, and we're trying to move all those obstacles out of the way.
I'm happy to answer any questions you have about us, either here or in email. Thanks,
I haven't had a ton of time to play with it in the past month, but definitely plan on ramping that up soon. It's a very cool product. I LOVE that it is in Python. I actually spent most of January building my algo trading system/backtester in Python, and it was just too much work to do correctly (too much work on the stuff that isn't the algorithm itself), so I dropped the project and went back to Ninjatrader...what a bummer that was.
Love your concept and very excited about start-ups like Quantopian.
The thing I didnít understand is why you went down the proprietary language route. Why didnít you guys run with a C#, Java or other implementation? The proprietary language route will really reduce your addressable market.
Take another look - it's all Python, not a proprietary language. Traderwerks correctly notes the power of Python plus the pandas library. And he has the details quite correct, too, with the open-sourced backtester. Check it out.
Yep, pretty cool indeed. Kudos to these guys for kicking off a long-needed web capability.
My concerns for them are that, if it catches on, the stream of feature/function requests will swamp their "free" model or open them up to a clone-competitor who skims the market with a quasi-free alternative (Aside: I'm a technology marketing guy by career, and this sounds very much like a case where the "pioneer" ends up face down in the dust with his back full of arrows -- hope not though).
The site is a bit limited for me at the moment because it lacks tick-level data (maybe a paid feature?) and I'm not clear on how to implement multi-time-frame algos (probably my ignorance though). All that said, I really hope these guys are wildly successful