Sys. Param Permutation(SSP) test recomendations

1 replies

Martin Villafañe

Subscriber, bbp_participant, customer, community, sq-ultimate, 1 replies.

Visit profile

1 month ago #285102

Good morning, I am trying to carry out the robustness tests of the strategies that have passed the previous filters. The SSP/sys. param permutation) since it is a complete test, the computing time for only 100 strategies can be many days. with the data in TICK format.
My question is, could it be a testing strategy to carry it out in the same timeframe, and the strategies that are surpassed in this TF, subject them to 1M, and then subject them to TICK?
My intention is to discard the strategies that do not pass the test in the original TF, nor the M1. So that those who remain can be tested at TICK.

Is this possible, is it recommended?

Another question is to consult you about the parameters you use for these tests.

I use the following:
Max Test: 500
Max Step: 10

% of suitable optimization > 30
average profit > 0
uniform distribution – less than 5
best optimization – disable

Net profit <= 150% net profit (median)
Drawdown >= 50% od drawdown (median)
ret/DD ratio <= 150% ret/dd ratio (median)

Your recommendations on how to optimize this test would be very useful to me.

thank you

You must be logged in to view attached files.



Customer, bbp_participant, community, sq-ultimate, 481 replies.

Visit profile

1 month ago #285201

I can write down one way how to find the optimal settings for this robustness test.

1) You generate e.g. 1000 strategies with one workflow
2) Then you do a final test for unknown data using these strategies.
3) Then you create a portfolio from the final test and write down the winning amount.
4) Then you incorporate the robustness test into your workflow before the final test.
5) You go through steps 2)-3) again and compare.

If the result is better, the robustness test has shown an effect.
If not, then change the parameters of the robustness test and run it again. This way you can gradually find optimal parameters for the robustness test.

If you have found the optimal parameters for the robustness test, then you are not finished yet. Then the biggest work begins.
1) You take the existing workflow with the built-in robustnestest and clone the workflow.
2) You move the times 1-2 months into the past.
Now you have 2 workflows.

You repeat the whole thing until you have 10-20 workflows with different start times.
Now you let the 10-20 workflows run until each workflow has generated around 100 strategies.

Then you look at the profits of the individual portfolios.
If most portfolios are profitable, then your workflow is good and the robustness test settings are good.

If not, then unfortunately you have to go back to Step1. And try again.
I’m not sure if there are optimal parameters for the robustness test. It can depend on the workflow and what strategies you have generated.
But here’s what I found out.
1) You have to have a good idea. An optimal setting to generate strategies. This means you choose good building blocks. And you have to have an idea of what you’re looking for.

2) If the first portfolio generated is too much in the red, then you can’t improve much with the robustness test. The robustness test only brings a maximum of 20% improvement.

I know it’s all a long road. But everything else is just gambling.
The idea you have is the most important. It’s best to already have the strategy you want to search in mind and set the building blocks appropriately. But to be able to do that you have to have some knowledge of trading.
I will attach a screenshot from my actual parameter permutation test.

You should also keep the following thing in mind:
If there are no good strategies in the generated set, then the robustness test cannot make anything positive out of this.

You must be logged in to view attached files.


Viewing 1 replies (of 1 total)