Opinions on risk models

what are your opinions on a robust risk model that fulfills these two things

1- positive probability expectancy

2- reasonable risk parameter




historical VaR
Monte Carlo brownian motion VaR
parametric VaR
log normal VaR

which from the above in your opinion is better?

or perhaps the following which is my running hypothesis at the moment.

would you think it more wise to take the 95% confidence level (of lognormal VaR T=5)for your first anchoring position using one week as parameter to catch the monthly movement which in reality it would be 15.68% confidence level for the month. This will be +22.39% positive expectancy.
I suppose this would require to have the position in the green the first week as the theta decays after that.

profit Target would be at one standard deviation.

If what I am posting sounds unclear I am happy to clarify ..... I suppose the only way to really know is to test and run statistics both live and perhaps simulation. Ideas ARE WELCOME

Here is the risk model I am using ''The risk of losing one standart deviation depends on the ratio of the average return to standart deviation ....'' http://www.futuresmag.com/2009/08/01...ignated&page=2