Evaluating factors influencing Trade Entry and Exit...
Author: rmikkili
Creation Date: 12/5/2011 1:23 AM
profile picture

rmikkili

#1
Hello All,

I am new to wealth-lab pro and based on what I saw in the last two days on what could be done and how easily, I am simply amazed. The support from the community is just incredible.

So before I reinvented anything, I wanted to seek out advice on what is already available for what I would like to try out.

I have a trend following a system that I am reasonably comfortable with, but sometimes, just like any other trend following system, it generates whipsaws. So, I would like to identify and evaluate the factors that have significant influence on my entry/exits and possibly eliminate some of the false positives or wait until the trend is confirmed.

So far nothing new here.

But, what I would like to do is, run a blind search on say 30-50 different factors like 1) RSI, 2) Price Above 20MA, 3) Stoch 4) Pivots or 5) ROC against its own sector or a set of Fundamental factors.... and so on.

I understand that the optimizer allows me to optimize a given indicator/factor for certain values. But I would like to go one step beyond and be able to *identify* the indicator/factor itself first.

Note that I am not trying to curve fit my system by making it look at 20 indicators for every entry exit. On the contrary, I would like to keep the system very simple and see if I can find at most one *MOST* significant positive contributor, if any.

So as an example, If I have a 20-50 MA crossover system that generates entry/exit signals, I would like to be able to conclude, for example, that the entry is significantly influenced by "ROC of the equity against its own sector". With this knowledge, I can then choose to reduce my initial position size or wait for some more time before entry.

I know that brute force will take ages to "Find & Optimize" a large number of conditions. So a Genetic Optimizer would probably be a good idea.

With my limited knowledge of wealth-lab, One way I thought I could implement this is:

1. Write a function that outputs a true/false indicating if an Entry can be taken after evaluating all entry conditions. Something like:

CanEnterTrade = EvaluateAllConditions();

if(CanEnterTrade) {
"Take Entry Signal"
else
"Wait for some more time"
}

2. Inside EvaluateAllConditions() function, I would define two values for each indicator:

- A Flag, for the optimizer to decide if the factor/indicator should be considered in this round of evaluation or not (Test for Inclusion)
- Two, if condition is included from one above, what step values for the indicator should be evaluated. (Optimize for the value, if included)

So, for RSI indicator, first flag says whether to consider RSI as part of entry signal or not and if RSI is included what value of RSI has the most significant impact.

3. Run the optimizer on all the 30-50 conditions which means the optimizer will have one boolean flag for each indicator to decide if the indicator should be included and a whole bunch of step values for indicator optimization if included...

If I havn't lost you already, Here are my questions:

1. Has this already been done before with wealth-lab that I could look at?
2. Am I trying to make this over complicated than necessary? Anyway to simplify what I want to do?
3. Is there a Genetic Optimizer that I could use with wealth-lab? I found something but need to spend more time to understand how it could be used.

Thanks for your kind help.

-RM

PS: Just to be clear, I am familiar with a whole bunch of programming languages and technologies including genetic algorithms so the language itself is not a problem.
profile picture

Eugene

#2
You might want to take a look at Analysis Series tab from the Extra Visualizers library:

Indicators | Testing Trading System Filters with Analysis Series View

More details available in the User Guide > Preferences > Performance Visualizers > Analysis Series.
profile picture

rmikkili

#3
Thank you Eugene for a lightening quick response. The link you provided looks very interesting. I will read further. After posting this message I also started experimenting with the Genetic Optimizer as well. I will post my findings here.

-RM