Let's say you created a strategy. And by changing its parameters you could achieve the following results:
a) APR = 52% Maximum Drawdown = 27% (Highest APR possible)
b) APR = 24% Maximum Drawdown = 8% (Lowest Maximum Drawdown possible)
c) APR = 50% Maximum Drawdown = 12%
You want to use the optimization function to achieve the optimal parameters. You cannot go for an exhaustive optimization, because that would take much too long, so you choose "Shrinking Window".
1. When you optimize for maximum APR, you will find the parameters for result a).
2. When you optimize for a minimum maximum drawdown, you will find the parameters for result b).
However, there is now way to find the parameters for result c). You will gladly lose 2 percent APR, if your maximum drawdown could go down to 12%. But there are no optimization settings to go for this.
However, if you could say: "If I have a maximum drawdown 5% lower, I will gladly give up 1% APR", you would be able to find it. So that would mean optimizing:
APR - maximum drawdown / 5
The scores for the 3 results a) to c) would be as follows:
a) 52 - 27 / 5 = 46.6
b) 24 - 8 / 5 = 22.4
c) 50 - 12 / 5 = 47.6
This would enable you to find a result like c) through optimization, which your strategy possibly is capable to achieve.
a) APR = 52% Maximum Drawdown = 27% (Highest APR possible)
b) APR = 24% Maximum Drawdown = 8% (Lowest Maximum Drawdown possible)
c) APR = 50% Maximum Drawdown = 12%
You want to use the optimization function to achieve the optimal parameters. You cannot go for an exhaustive optimization, because that would take much too long, so you choose "Shrinking Window".
1. When you optimize for maximum APR, you will find the parameters for result a).
2. When you optimize for a minimum maximum drawdown, you will find the parameters for result b).
However, there is now way to find the parameters for result c). You will gladly lose 2 percent APR, if your maximum drawdown could go down to 12%. But there are no optimization settings to go for this.
However, if you could say: "If I have a maximum drawdown 5% lower, I will gladly give up 1% APR", you would be able to find it. So that would mean optimizing:
APR - maximum drawdown / 5
The scores for the 3 results a) to c) would be as follows:
a) 52 - 27 / 5 = 46.6
b) 24 - 8 / 5 = 22.4
c) 50 - 12 / 5 = 47.6
This would enable you to find a result like c) through optimization, which your strategy possibly is capable to achieve.
Rename
You can create a new ScoreCard with the desired Metric formula and target that new Metric in the Optimizer.
I think the suggestion in Reply# 1 is the very best way.
But if you don't want to write a ScoreCard merit metric, then a substandard approach would be to apply an APR/Drawdown ratio with the meta-ScoreCards feature to approximate a compromising result. You can vote up that feature request if you like. https://www.wealth-lab.com/Discussion/Meta-Scorecards-and-for-Multi-Objective-Optimization-5947
But be aware, empirical "interaction terms" are typically product terms between two directly-proportional orthogonal metrics (like APR and WinRate). Dividing these terms can have undesirable consequences.
But if you don't want to write a ScoreCard merit metric, then a substandard approach would be to apply an APR/Drawdown ratio with the meta-ScoreCards feature to approximate a compromising result. You can vote up that feature request if you like. https://www.wealth-lab.com/Discussion/Meta-Scorecards-and-for-Multi-Objective-Optimization-5947
But be aware, empirical "interaction terms" are typically product terms between two directly-proportional orthogonal metrics (like APR and WinRate). Dividing these terms can have undesirable consequences.
Thanks for your answers!
Hello Glitch,
thanks for your answer. Could you advice me how to go about creating a new scorecard: What steps would I have to take in WL7?
Thanks!
Werner
thanks for your answer. Could you advice me how to go about creating a new scorecard: What steps would I have to take in WL7?
Thanks!
Werner
Werner,
Scorecard development is an advanced topic which requires that you master C# and Visual Studio. It's covered in this API doc:
https://www.wealth-lab.com/Support/ExtensionApi/ScoreCard
Scorecard development is an advanced topic which requires that you master C# and Visual Studio. It's covered in this API doc:
https://www.wealth-lab.com/Support/ExtensionApi/ScoreCard
Here's a half-baked feature request:
A drag and drop formula builder for an optimization target based on existing score card metrics.
A drag and drop formula builder for an optimization target based on existing score card metrics.
QUOTE:
A drag and drop formula builder for an optimization target based on existing score card metrics.
Please explain how that differs from https://www.wealth-lab.com/Discussion/Meta-Scorecards-and-for-Multi-Objective-Optimization-5947 They kind of sound the same to me. (Obviously, there's something I don't understand.)
Also, it would be desirable to have any "derived" ScoreCard metric "somewhat" compiled for speed. I'm not sure how you do that.
First, meta scorecards is complete.
Second, Dr. Koch's request is not as concise or explicit as "drag and drop formula". I can't really even tell what's really requested there.
I'm suggesting just creating a formula by dropping objects to use as an optimization target. Voila! I did it again in one sentence.
Second, Dr. Koch's request is not as concise or explicit as "drag and drop formula". I can't really even tell what's really requested there.
I'm suggesting just creating a formula by dropping objects to use as an optimization target. Voila! I did it again in one sentence.
Your Response
Post
Edit Post
Login is required