>If a discretionary trader wants to see if his return is NOT due to
>randomness, what does he do?
>Just check his past results? How large his sample of historical
>trades it should be?
One can write a book on the subject
Maybe I will do so but in french not in english or would be too difficult for me
.
As short recipe nevertheless, if you want to know if your P/L has some chance to be over 50% without making any assumption about normal law - since you are a discretionary trader but even if you are not - you can use Tchebycheff to get an estimation. A numerical example is given - for a poll
:
http://www.elitetrader.com/vb/showthread.php?s=&postid=396656
Since the size of a sample is taken into account it doesn't matter how size you have.
After that draw an histogram, if it ressembles a bell curve then you can refine Tchebicheff (there are some stat method to test the bell curve also). It's better to aggregate the data by day if you take many trades in the same day so as to erase independancy problem (If you are a trend follower, since trend to persist from day to day, it would be better to refine even further due to this problem but I won't enter into details it's too long).
>Check other statistical facts like drawdown?
Drawdown is more difficult to estimate because even if a distribution is normal the extreme doesn't follow a normal distribution. It is worse if the distribution is not normal. Some use Monte Carlo Simulation but it is not rigorous as they generally assume implictly some random law so you musn't trust Monte Carlo Software Vendors to do the good assumptions for you. It can give you an estimation but you must be very conservative in your input so as to avoid false confidence given by their simulations - it is that kind of false confidence that may explain the Chicago legend reported by Nassim Taleb, that a trader "made 8 million in eight years and lost 80 million in eight minutes"
Quote from Hamb-ltrd:
Thank you Harry.
I look forward for the Martingale Thread.
Very interesting thread about randomness.
Some of you comments
"Now although there is way to detect NON Randomness, there is no way to detect the opposite Randomness."
If a discretionary trader wants to see if his return is NOT due to randomness, what does he do?
Just check his past results? How large his sample of historical trades it should be? Check other statistical facts like drawdown?
Also " If you find indice of NON Randomness you can say that the process is not random "
How do you do this?
I might lack the education to understand you here, ( I just had one course in statistics in college)
Thanks again
Hamb
