Thanks for doing all this research for us. I hope you had some good rest over Christmas and a nice time with family and friends. Looking forward to Jan 15th :) best regards, Rod
Hello Ali, Thank you, as always scientific high quality information. Very few channels provide such quality. I have got a question please, I have searched a lot, but no convincing answer. It would be great if you answer me. I have created an algo and tested it in MT5, I noticed that the Z-score is -75.73(99.74%). What that can tell, how could the Z-score be more than 3?
Thanks for the video. I find it interesting but we should also look into the negative volatility of the equity curve for each strategy because it is kind a feel good about how your trading bot is losing. You want to consider low drawdowns too
You are right, but Risk is different for each person, some are looking for absolute returns. John Henry made a fortune but took many 50% DD along the way
Amazing job done here! Really gives an understanding of fitness function... Im trying to implement ULCER INDEX in Multichart as a custom fitness function but look like i won't be able to do this ever. Im not a programmer and they saying something about JavaScript???? Crazy .... Is the anybody here , Maybe author of the vide, who actually has or can create a code of ulcer index as fitness function for multichart or trade station? It would be so helpful for me and im shure for many others Thanks
Yes, they are not very fast to answer on emails. Maybe because of smaller community in comparison with tradestation... Who knows! Does any body manage to create custom fitness function based on ulcer index in tradestation or multicharts?
1. Net profit is the one that get you moving, but max DD can make you dead :) So I think of combining these two. What is your take on this? 2. “Done and Done”? What is the golden book you have there?
In these cases you kept position size fixed, but if it varied, then wouldnt CALMAR be more important? Point being that even if net profit was low for a strat with high CALMAR, you can leverage to a point with same DD as other strategies while having higher net profit.
IMHO this is just 50% of the story. I've done a similar experiment, searching for correlation between the in sample performance metrics and (true) out of sample Net Profit. NONE of the SQX performance metrics on In Sample testing data show ANY correlation in regards to OOS performance. There is no way to predict future performance based on the in sample performance metrics. The only real value that I see using In Sample performance metrics is setting boundaries on when you should consider a strategy "expired". I have yet to test if there is any correlation between MC/SPP and OOS performance.
@@StatOasis I thinks so. Building according to CAGR, then retesting filter on UPI or Sharp ratio or Stability or Max DD %. I am surprised about average trade, amazing.
There is something unclear. The test should have been done on unseen oos data. For example generation 2 years and test over the following one. Can you please clarify exactly what has been done here ?
Unfortunately because of crashing I had to export, delete all and start the next batch. The idea that I wanted to verify is not what is the best metric to yield robust strategies, I just wanted to see if a simple NP will yield to something close to other complicated metrics yield and it is. Simple works, but that doesn’t mean you can’t use other metrics
Many fall in this trap, but Van Tharp himself says it is an indicator of how your system is suited for position sizing, but I will see if I have the data and post the chart
Thanks for doing all this research for us. I hope you had some good rest over Christmas and a nice time with family and friends. Looking forward to Jan 15th :) best regards, Rod
Happy holidays for everyone, I wish everyone health, happiness and success in 2022
Hello Ali, Thank you, as always scientific high quality information. Very few channels provide such quality.
I have got a question please, I have searched a lot, but no convincing answer. It would be great if you answer me.
I have created an algo and tested it in MT5, I noticed that the Z-score is -75.73(99.74%). What that can tell, how could the Z-score be more than 3?
I don’t use MT5, but you can export the trades and calculate z score to compare
Thanks for the video. I find it interesting but we should also look into the negative volatility of the equity curve for each strategy because it is kind a feel good about how your trading bot is losing. You want to consider low drawdowns too
of course you can add your own functions
Awesome video. Thx.
Other than Net Profit, Drawdown, and the shape of the curve, what will be other important factors?
UPI can replace DD and Shape.
Avg Trade, consecutive losers,exposure, etc
Really you can go as deep as you want but you are on the right track
Really appreciate this valuable content!
Thank being a part of the channel
That was truly amazing !!
Glad you found it helpful
“So, you know, Java”
😂 I felt that one to my core!
We’re currently developing a platform in Python to replace the role of SQX in our work flow.
Let me know when you have a stable working version
Only risk adjusted returns matter. Net profit doesn't mean much if the risk of the strategy is significantly higher
You are right, but Risk is different for each person, some are looking for absolute returns. John Henry made a fortune but took many 50% DD along the way
@@StatOasis i don't know Johny but i would say he is a case of survivor bias
All-time strategies generated: +165 k Still working on it :)
😊
Amazing job done here! Really gives an understanding of fitness function...
Im trying to implement ULCER INDEX in Multichart as a custom fitness function but look like i won't be able to do this ever. Im not a programmer and they saying something about JavaScript???? Crazy ....
Is the anybody here , Maybe author of the vide, who actually has or can create a code of ulcer index as fitness function for multichart or trade station? It would be so helpful for me and im shure for many others
Thanks
Multichart team is very strange.I have been there, but they don’t co-operate and help their users
Yes, they are not very fast to answer on emails. Maybe because of smaller community in comparison with tradestation... Who knows!
Does any body manage to create custom fitness function based on ulcer index in tradestation or multicharts?
1. Net profit is the one that get you moving, but max DD can make you dead :)
So I think of combining these two. What is your take on this?
2. “Done and Done”? What is the golden book you have there?
1 of course you can combine them in a fitness function or develop on NP and then filter on DD
2 that’s my motivation sign “not a book”
In these cases you kept position size fixed, but if it varied, then wouldnt CALMAR be more important? Point being that even if net profit was low for a strat with high CALMAR, you can leverage to a point with same DD as other strategies while having higher net profit.
You can use Calmar ratio or other metric, just overall it looks like simple net profit is just the same in prediction value
IMHO this is just 50% of the story. I've done a similar experiment, searching for correlation between the in sample performance metrics and (true) out of sample Net Profit. NONE of the SQX performance metrics on In Sample testing data show ANY correlation in regards to OOS performance. There is no way to predict future performance based on the in sample performance metrics. The only real value that I see using In Sample performance metrics is setting boundaries on when you should consider a strategy "expired". I have yet to test if there is any correlation between MC/SPP and OOS performance.
You are on the right path with testing
Do you mean that fitness functions Net profit(Return) in SQX refers to annual % return?
Annual returns is a function of net profit that’s why ther are highly correlated
What then is your ideal set of custom filters in SQX?
Develop on net profit then filter on DD
@@StatOasis I thinks so. Building according to CAGR, then retesting filter on UPI or Sharp ratio or Stability or Max DD %. I am surprised about average trade, amazing.
There is something unclear. The test should have been done on unseen oos data. For example generation 2 years and test over the following one. Can you please clarify exactly what has been done here ?
I build all strategies with IS/OOS but the metrics are for the whole data
@@StatOasis thanks. Do you still have the strategies to run a test on unseen data ? That would be way more beneficial to the research
Unfortunately because of crashing I had to export, delete all and start the next batch.
The idea that I wanted to verify is not what is the best metric to yield robust strategies, I just wanted to see if a simple NP will yield to something close to other complicated metrics yield and it is.
Simple works, but that doesn’t mean you can’t use other metrics
@@StatOasis understood. Thank you for all the work you do. it is truly inspirational.
Thank you for watching and wish you a great 2022
I miss the SQN :D, it might have a reason you didnt showed it ;-)
Many fall in this trap, but Van Tharp himself says it is an indicator of how your system is suited for position sizing, but I will see if I have the data and post the chart
Do you prefer Genetic or random generation?
Genetic only make sense if you have many variables which leads to 100 of thousands of possibilities. Please check my video on genetic algos