Effects of not keeping a Stop Loss

[oxusmorouz]

Hi CV, thanks for your reply.

Oxy, I am a bit confused with the detrending system returns, benchmark part.One problem with the mean you describe is the compounding issue, correct?
Also isnt sample size the no. of trades you generate and not the testing data?

Yes CV, but the process looks a lot more complex taking into consideration the number of trades. I thought a better idea would be to plot the "n day ROC of the equity curve" and use this data for further statistics, rather than taking "per trade profitability of the system" which is generally used.
I have jotted down a blue print as to how I should proceed in the following steps. Please say if the following method looks acceptable, CV:
1) Determine "n" period %return of system equity curve for "y" samples.
2) Determine "n" period %return of market index buy-hold (or buy-hold of stock, which ever is used as the standard for determining out performance) for "y" samples.
3) Subtract 1 - 2 and log its value (Since log is only for positive values, sign adjustment can be done after finding absolute log value)
4) Find mean and standard error of 3.
5) Subtract mean from each value of 3.
6) Sort 5 ascending order so that it is mean adjusted, and represents a log normal distribution.
7) Determine the probability of the system being greater than "- mean" using z test.
8) Repeat this process for "x" number of data sets.
9) Find average probability of system returns out performing standard.
10) Accept if probability is greater than expected figure.

 

[oxusmorouz]

Ajay,

Plz correct me if i have any wrong understanding.
To Abhijeet's & my query of 'Period for back testing' Jesse has given an input & how he felt it is to be Moduled.
You have suggested a method ;when 't' (trade duration) of Benchmark performance vis-a-vis our system is inequal how to bring down the anamoly in the 'Scale of Comparision' & also expressed a doubt that as Out of sample data may also contain 'n' no of biases hence you wanted to know from Jesse how to tackle this issue; as Live test of EoD systems will be Time consuming.
Here Jesse has suggested QC of Input for the actual trades taken by system.

Asish

Well yes sir. Since you were already on the topic, I thought it would be a good idea to post my question here...

 

[CreditViolet]

Hi CV, thanks for your reply.



Yes CV, but the process looks a lot more complex taking into consideration the number of trades. I thought a better idea would be to plot the "n day ROC of the equity curve" and use this data for further statistics, rather than taking "per trade profitability of the system" which is generally used.
I have jotted down a blue print as to how I should proceed in the following steps. Please say if the following method looks acceptable, CV:
1) Determine "n" period %return of system equity curve for "y" samples.
2) Determine "n" period %return of market index buy-hold (or buy-hold of stock, which ever is used as the standard for determining out performance) for "y" samples.
3) Subtract 1 - 2 and log its value (Since log is only for positive values, sign adjustment can be done after finding absolute log value)
4) Find mean and standard error of 3.
5) Subtract mean from each value of 3.
6) Sort 5 ascending order so that it is mean adjusted, and represents a log normal distribution.
7) Determine the probability of the system being greater than "- mean" using z test.
8) Repeat this process for "x" number of data sets.
9) Find average probability of system returns out performing standard.
10) Accept if probability is greater than expected figure.

Yes, I see your point.Using Geometric Mean and Resampling methods is a good idea.

 

[uasish]

Ajay,

Regarding your point No 6.
I have seen traders using other methods than 5 ascending orders,but finding a Log normal
distribution & Z test ranking in comparision is logical & more appropiate.
From point No 8. it is difficult in Metastock .

Asish

 

[oxusmorouz]

Ajay,

Regarding your point No 6.
I have seen traders using other methods than 5 ascending orders,but finding a Log normal
distribution & Z test ranking in comparision is logical & more appropiate.
From point No 8. it is difficult in Metastock .

Asish

I guess the whole process can be done in metastock, except the step where the data is plotted in ascending order. Arranging the data in ascending order again would serve of little purpose except in graphical viewing, since all that is necessary for calculating a z score is mean and standard error, having defined the population as a normal distribution.

 

[newbee]

I guess the whole process can be done in metastock, except the step where the data is plotted in ascending order. Arranging the data in ascending order again would serve of little purpose except in graphical viewing, since all that is necessary for calculating a z score is mean and standard error, having defined the population as a normal distribution.

Hello oxusmorouz

I wanted to apologize for my behavior, on our first interaction. I am sorry pal.

Now that the tough part is over, would like to confess, don't understand one bit about what u ppl are discussing here.

Maybe few years from now, will need your help, hope you will be avialable then.

Thanks
nb

 

[newbee]

Yes, I see your point.Using Geometric Mean and Resampling methods is a good idea.

Hello CreditViolet

Do you take Apprentice?

If yes what would be the qualifications required?

I am just starting out, and planning to take up trading as a profession, if you get some time please leave a message on the Journal I have started here at traderji.

Thanks
nb

Link

 

[oxusmorouz]

Hello oxusmorouz

I wanted to apologize for my behavior, on our first interaction. I am sorry pal.

Hello newbee,
No apology necessary

 

[newbee]

Hello newbee,
No apology necessary

Well you might change your mind, if you know what i was thinking to write, but decided against posting it.



Thanks n Peace

nb

 

[beginner_av]

Hi CV, thanks for your reply.



Yes CV, but the process looks a lot more complex taking into consideration the number of trades. I thought a better idea would be to plot the "n day ROC of the equity curve" and use this data for further statistics, rather than taking "per trade profitability of the system" which is generally used.
I have jotted down a blue print as to how I should proceed in the following steps. Please say if the following method looks acceptable, CV:
1) Determine "n" period %return of system equity curve for "y" samples.
2) Determine "n" period %return of market index buy-hold (or buy-hold of stock, which ever is used as the standard for determining out performance) for "y" samples.
3) Subtract 1 - 2 and log its value (Since log is only for positive values, sign adjustment can be done after finding absolute log value)
4) Find mean and standard error of 3.
5) Subtract mean from each value of 3.
6) Sort 5 ascending order so that it is mean adjusted, and represents a log normal distribution.
7) Determine the probability of the system being greater than "- mean" using z test.
8) Repeat this process for "x" number of data sets.
9) Find average probability of system returns out performing standard.
10) Accept if probability is greater than expected figure.

without taking away anything from your brilliant work, sometimes the simple things may be sufficient. just like someone using garch to look out for volatility in volatility and someone just using ATR to do the same for the instrument.