Matlab: Most powerful/flexible data analysis/algorithm development tool?
In your opinion, is Matlab the most powerful/flexible data analysis and algorithm development tool?
I'm just starting out in algorithm development, I want to start w/ a platform that is powerful/flexible enough to meet any needs that I'm likely to have in the future, even if I get very advanced. This way I don't waste time by learning one system, then having to start from scratch learning an entirely different system when I outgrow the first system.
One key feature that I'm hoping to find is the ability to stream realtime data into the analysis algorithm, to automatically modify the algorithm in real time. An algorithm that learns, like a neural network.
Does it sound like I'm on the right track?
Any info is much appreciated
Last edited by maxplanck; November 12th, 2011 at 12:02 PM.
MATLAB is far from the most powerful solution, but it is good for prototyping. It is not built for HFT stuff. Combining it with CQG or RTS is excellent, though. RTS RTD Tango should be enough for all your needs, but, if it's not, why don't you just start off with Apama instead!
I have used Matlab professionally in a science and enginering context for about 15 years. Matlab is powerful, and is especially good a rapid prototyping code and ideas. Using Ninjatrader now, I miss the flexible plotting mechanism.
However, compared to almost anything else, Matlab is slow. It is an interpreted, and not compiled language. Matlab is designed for ease of programming large vectors and matricies of data, not computational speed. Matlab also is not particuarily stable.
Matlab was my orignal platform. Speed, stability, and complex broker/Matlab APIs convinced me to look elsewhere. This is my stance: if you are doing hard core math to justify needing Matlab, then you most likely know enough to write something equivalent in C.
Now, once I see a Mathematica-Ninjatrader API, or figure out how to compile a Mathematica dll for NT, I may change my tune...
The following user says Thank You to serac for this post:
-Prototype and backtest in Matlab
-then, translate the algorithm for implementation through a broker's API
Who's the best/fastest broker/platform for running non-HFT algorithms on a retail budget? Server colocation rates (I assume this is necessary for HFT) are probably out of my budget (they require a nondisclosure agreement to even quote a price), though it's something that I'd like to explore and get into if I can find a way to learn and get my foot in the door.
Advantage Futures offers co-lo. That would be faster, of course. You have to pick your broker in accordance with your requirements, I simply linked to AF due to their information about algo trading.
It is always wise to check the excess capital of the broker(s) one is considering, most seem to overlook that.
The API is not the same. IB has their own, and other FCMs offer products like CQG, RTS, Stellar, TickIt and ORC (to name a few) that have their own API's. IB's global account is nice and their API is easily accessible, but their data is not reliable. There is some relevant information on the site I posted, I would advise you to read it...
Last edited by Lornz; November 14th, 2011 at 05:07 AM.
Reason: unintentionally came across as hostile
The following user says Thank You to Lornz for this post: