Australian (ASX) Stock Market Forum

Amibroker Monte Carlo Question

I feel like I am a victim of analysis paralyses at the moment.....
Curve fitting victims keep altering parameters until the results are acceptable. Also being aware of back test gotchas is helpful otherwise you will never get a system to work going forward. Fact.
 
Is there any way to import .csv files into the Amibroker backtester? You can export csv and html but you can only import html for some reason. I may have seen a script here on ASF, probably by Howard Bandy, but can't remember where. It would be useful to get all the nice AB statistics done automatically, including MC.
 
Is there any way to import .csv files into the Amibroker backtester? You can export csv and html but you can only import html for some reason. I may have seen a script here on ASF, probably by Howard Bandy, but can't remember where. It would be useful to get all the nice AB statistics done automatically, including MC.

What data do you want to process? Do you mean trade data from a CSV file?
If so then at AB knowledge base there is an example
http://www.amibroker.com/kb/2014/09/30/gen-backtest-from-a-file/
 
Is there any way to import .csv files into the Amibroker backtester? You can export csv and html but you can only import html for some reason. I may have seen a script here on ASF, probably by Howard Bandy, but can't remember where. It would be useful to get all the nice AB statistics done automatically, including MC.

Greetings --

What format are the trades in in the csv file? After the trades have been imported, how will they be further processed?

Anticipating your answer -- If you are planning to do some Monte Carlo analysis of the trades, consider instead using the Excel Monte Carlo add-in that goes along with my "Modeling" book. Or the Python program listed in my "Quantitative Technical Analysis" book. Both of these are much more flexible and much more powerful than AmiBroker for that purpose.

Best,
Howard
 
In what way is an Excel addon or some phyton script more powerful?

Monte Carlo is a simple process.

What does an Excel addon do differently?
 
What data do you want to process? Do you mean trade data from a CSV file?
If so then at AB knowledge base there is an example
http://www.amibroker.com/kb/2014/09/30/gen-backtest-from-a-file/

That's the actual post I was looking for. I couldn't remember where I had seen it; I've been searching in the wrong place.
Thanks, trash.


What format are the trades in in the csv file? After the trades have been imported, how will they be further processed?


I wanted to import the trades from two different system backtests to get the combined statistics. There would be no problem with the data format in that case. Also interesting would be to import 'real' trades from Excel. That would be more of a challenge as my real trades don't match the format in AB backtest.
If I'm serious about this, changes will have to be made. So far the only statistic I have been concerned with is net profit ...

Thanks for your help.
 
In what way is an Excel addon or some phyton script more powerful?

Monte Carlo is a simple process.

What does an Excel addon do differently?

Greetings --

Monte Carlo Analysis is a very broad set of tools and techniques. I would argue that it is not, in general, a simple process. As related to trading, MC can be used in several different ways, including:
Testing robustness of a model to changes in data.
Testing robustness of a model to changes in logic, rule, indicators, and parameters.
Testing risk and profit potential of a set of trades.

The trading system development platforms we use -- AmiBroker, TradeStation, Ninja, etc -- are designed for use developing decision tree trading models. And they are very good for that.

The first two ways listed above can be done within the trading system development platform. Data can be perturbed. Parameters can be perturbed. These could be considered simple processes.

The third can "sort of" be done in the platform, but not simply, easily, or completely. It would be done with great difficulty and impracticality -- nearing impossibility -- when working with the validation and trading management phases of trading system development. Analyzing risk and profit potential uses an iterative procedure that draws many equally likely trade sequences from the the set of trades being analyzed to establish the distribution of risk, with the metaparameters for the analysis guiding the fit which is based on a complex objective function. That is followed by use of the MC model to draw another set of equally likely trade sequences to establish the distribution of profit potential. These can be done easily in a single Python program; medium easily including some operator manipulation in Excel; and with considerable difficulty -- probably not at all if done correctly -- in a trading system development platform. This is definitely not a simple process. If some implementation of it appears to simple, that implementation is incorrect and the results will be misleading.

This is not a criticism of any of the trading system development platforms. They do what they were designed to do very well. It is a statement that some of the analysis a trading system developer might want to perform using Monte Carlo techniques are better done with tools designed for that purpose.

Best regards,
Howard
 
AB is not Ninja or TS. It is a completely different (universal) platform. On the Yahoo list there have been several times where you said this and that could not be done still it has been negated by providing actual examples within same day. But I understand you need to promote your books.
 
AB is not Ninja or TS. It is a completely different (universal) platform. On the Yahoo list there have been several times where you said this and that could not be done still it has been negated by providing actual examples within same day. But I understand you need to promote your books.

Greetings --

I assume this message is directed toward me.

I have not yet seen an example of use of Monte Carlo tools within a trading system development platform that implements rigorous modeling, simulation, validation, and analysis techniques.

Best,
Howard
 
Top