I would like to know what and how are the impact of bad religions of historical data set in back-testing an algorithm or a strategy.
as an example, suppose a historical data set has a sudden movements due to news impact. when we use this data set for back-testing an algorithm, definitely algorithm cannot be properly tested because of these kind of bad religion.
I would like to know your ideas on this, how to overcome this kind of issue when back-testing a strategy developed.
if I create a full set of data, annotation these kind of all bad religions, which can easily skip or thoroughly consider depending on the situation when back-testing, I would like to know the effect I am going to put for this project will valuable or not.
To overcome this is simple.
Or if you insist you could add a filter for abnormal moves, but i don't recommend it.
No it's to manually test strategy's so you can practice your idea's and develop your trading skills.