Quant trading & QuantLib

Hello,

I am researching neural network and quant trading. I am curious if there are any people that have used QuantLib and what their opinions are.

I am hoping to be able to release something to the public in few months. Right now everything is very experimental.

Best Regards!

What inputs do you intend to use? Strictly price?

Quant trading is new to me but I would like to target profitable price action setups.

I think it's all about probability in this case.

In other words what I expect to get out of this is identify patterns such as the PIN bars in James16's thread. But it doesn't need to be that obvious to the human eye as long as it's identified by the software - you will have a lot more high probability setups.

I think even if you take the last 2 1H or 4H candle sticks you can run genetic algorithms and find good setups with over 65% probability.

So... to answer your input - ABSOLUTELY NOT!

Strictly price is definitely not what I will be using.

I am creating a tool where you can visually link the inputs and outputs.

That way you can train your own network using for example the rainbow MA (as in linuxtroll), or inputs from custom news/stock/futures feeds. (e.g. take into acct DOW or gold prices)

Ahhhh very interesting.... I guess I should have primed you that I've been investigating NNs pretty thoroughly lately....

And the linuxtroll rainbow was my first thought.... in fact, I used 100 MAs to train a NN, but then I realized something....

MAs are really just recalculations of price itself, so price is the purest form of its movement. In other words, patterns found in price should be better than any NN rainbow patterns (in theory.) I don't swear to it, but it seems true to me for now.

NOW: As for other inputs.... we really need to find reports that are predictive in nature. For example, although interest rates seem logical to help determine a pairs direction, I have read studies showing they ARE NOT helpful in determining anything. So COT, CPI, etc, may or may not be helpful, but if there is no strong linear correlation to price, they would only hurt (create false positives)

QuantLib wont help at all in the spot fx market, it's geared more towards building models for the fixed income / interest rate derivatives markets

Interest rates have plenty to do with Forex, i.e: Australian/NZ bonds

I am currently employed at a Quant Prop Tradiing firm working on NN/SVM focused primarily on the Sterling Pound & GILT rate correlation; implementing path-integrals; Quantlib is a useful foundation -but, by no means, the end-all-be-all of quant-trading libraries. Most of the stuff is written in C++ (Which is bland, yet easy to understand); fortunately, everything can be rewritten on Ruby or Python via BOOST.

Best of Luck, Happy trading.