GARCH-model

Are you sure it was this forum? It doesn't ring a bell.

I would be surprised if it wasn't this forum, I didn't read another forums this week.

The team didn't delete (nor reply in) a topic about GARCH.

Thanks a lot.

A question: I think that with GARCH(1,1) model can be a numerical series created from the daily close prices eg. S&P500 index (or may be ES Emini futures ?), which series is the volatility of S&P 500. Now with the volatility we have one number for each day, so we could lay a Stochastic or MACD on it. In your approach I should create the series of numbers (volatility) in R, and that should I upload into wealth-lab to lay the indicators on it and be able to examine it ? Am I right ?

My problem with R is that I have no knowledge to it, opposite to Python, in which I have some programming practice. Can not Python be connected to WL in any similar way ?

QUOTE:
My problem with R is that I have no knowledge to it, opposite to Python, in which I have some programming practice. Cannot Python be connected to WL in any similar way?

There's two separate issues here. Let take the easy one first.

1) Any compiler that's compatible with the .NET framework and libraries (C++, C#, Visual Basic, F) can work with Wealth-Lab. But I'm not familiar with Python and I can't say it will work with the .NET framework. That might be a question to take up in a Python user group. Because the .NET framework (and Wealth-Lab) are compiled (actually JIT'ted) and Python is interpreted, it would be **tricky** to do. What compiler do you use with Wealth-Lab now? C# ?

2) Our goal is to do as little coding as possible. That gets us where we need to go with the least amount of debugging. That means using the highest level language possible for the problem at hand. Python is a low level language and what you want is a high level functional language that supports GARCH modeling and solutions. That's R, not Python! So install R and "pick" an R package especially made for modeling GARCH problems and use that instead. That will save you lots of time and debugging. And this approach will get you where you want to be the fastest with minimal coding.

Sure you could write a trading simulator (backtester) in Python instead of using Wealth-Lab as your trading simulator. But it would take you 2 or 3 years to reinvent the wheel, and we don't want to do that. So use a functional level language (such as R or even MatLab) to do your GARCH modeling with. Think high-level, not low-level. It's not how much code you write. Rather it's about writing less code and doing more with it. Software Engineering 101.

I have a question for you. If time series analysis is your thrust (and we know it is because you're modeling with GARCH), then how does event-driven simulation (i.e. Wealth-Lab backtesting) fit in? There are stock traders that do everything with continuous simulation on R rather than event-driven simulation (Wealth-Lab). Is your goal a "hybrid" simulator (both event-driven and continuous married together)?

I can't answer your Reply# 6. But a stock trading forum using R may be able to answer it. They would be doing more "continuous" modeling with GARCH and ARMA.

Thank you for your answers, in fact I started learning C# programming at first to be able to work with WL, but than I saw in every job advertisement, that financial companies work with Python. That's why I stopped C# and started Python, because when I don't find what I am looking for, at least I will be able to fulfill a job requiremen (Python) with a financial company. But I am still at the beginning of my journey, and realized that if I want to be proficient in Python that will take years and years. So in fact I don't know what I want exactly, and I am starting to realize nowadays what the difference means between a commercial backtesting software and Python. Now I am turning again to WL and started learning C# again, because I am planning to subscribe WL8, when I am at that programming level in C#.

So What compiler I use ? Now I don't use WL yet, when I was using WL6, I used the built in editor in WL6.

My only goal is to finally find something in financial markets, I can make some money from, after the many years I spent observing them.

So after many years, and still now it's starting to develop which way I take. Now I think I will try the R.NET software you recommended, because a friend of mine is proficient in R and when we join forces we may find out something.

Thanks for your answers and your advices.

QUOTE:
... I started learning C# programming at first to be able to work with WL, but than I saw in every job advertisement, that financial companies work with Python.

Python is for throwing something together quick and dirty. Or for transforming data from one ASCII format (schema) to another so you can input it into a different application. Perhaps they do a great deal of that at financial companies. However, Python is an interpreted language and no one would employ an interpreted language (which is 20 times slower than a compiled language) for a simulation problem. That just isn't done.

Now R is also an interpreted language, which is why you try to avoid R FOR loops. However, R calls R packages (like a GARCH package) to do the numerical analysis stuff. And R packages are written with "R Studio" in C++, which is fast.

Should you learn C++ or C#? Well, either will work with Wealth-Lab. And C++ is recognized more in industry. If your only purpose is to use Wealth-Lab, C# may be a better choice because it avoids pointer arithmetic and uses indexes instead. I just hate throwing a pointer in C++ and crashing the whole application. And that can take a half a day to find that bug.

In contrast, C++ is the better choice if you're coding low-level numerical analysis because it uses pointer arithmetic, which makes it fast. But it sounds to me that you're into the application level. That is, you're an R user and not an R developer writing low-level R packages (in C++) with R Studio. In that case, I would avoid C++.

And for developing your GARCH model, I would do that development in R with a GARCH package. But once R finds (fits) the best GARCH coefficients, you could pass that solution (those coefficients) into Wealth-Lab for an event-driven trading simulation (backtest).

QUOTE:
I will try the R.NET software you recommended,...

I would get everything running right on R first before installing R.NET. Just be aware R.NET is a couple releases behind R, so it will only work on an older version of R.