カスタム Projects ワークフロー

I have made the exact configurations that were provided in the video course and would like to share it with members, please review each tab before start building, also you need to configure the first tab “Build Strategies” as you want.

I didn’t add the last OOS tests and the filtering of similar trades test because they need to be done manually.

Let me know your thoughts and if we can enhance this.

Thanks, I will check it with the workflow generator and make an analysis

トーマス

Thanks for sharing! What is your long term experience with this workflow?

Hi Richard,

I am trying to test many workflows, but didn’t see that much difference on the outcome of the strategies, maybe I am missing something on the road, but regarding this workflow I think it is better to define one direction for the market “Buy or Sell” and not both, also in the data tab it is better to have in-sample and out-of-sample two or three times during the period specified for generating strategies, in addition to increasing number of trades up to 4 or 5 per month and not 2.

Thank you but when I open the zip file I get a file named config.xml.

How do you make it into a workflow in SQX ?

After opening SQX, go to custom projects and you will see a button named “open existing projects”, press on it and search for the file.

There will be more free projects coming soon, you can check my website quant-bot.com

Thank you Hani. Sorry for this late response! I had not received any notification and I was busy with other problems.

I downloaded the Custom-Project-Workflow.cfx file but it is in fact still a Zip files containing only a config.xml file which does not change anything to the Custom Projects. Contraty to the name of the link, there is no .cfx file.

Hi there. I downloaded the config file and discovered some improvements as could be include the other markets test. I attach the config here if it’s usefull for anyone. Thanks Hani

Thank you for the workflow. I test this workflow at the moment. I will repeat this test 10 times. I shift the time periods 60 days in the past in every test. But I think this workflow is not so good at the moment. The frist try generate 1500 Strategies and at the end one loosing/(sideways) strategy left. I attached screenshot of this strategy.

The problem of this workflow is the curvefitting. We have many curvefitting tests in this workflow. But the curvefitting filtering is not enough to completely filter out all the bad strategies.

I will increase the amount of generated strategies to 5000. If only one strategy is left at the end, it has no statistical significance.

This is a general problem. We need more curvefittingtests for the strategyquant.

My experience tells me. If, if more bad strategies are generated during the generation, you can no longer filter out the bad strategies with the curvefitting tests available in the SQ.

トーマス

Hello tnickel,

Thank you for sharing your findings, I am no longer using such workflows, although I have now around 150 strategies that are performing “not bad” which are generated from such workflows.

I would like to add my new insights, “Simplicity” is the key.

When adding more filtration factors to the strategy bank, which means additional testing, then we are biased towards curve fitting, so the more factors and tests, the more curve fitting strategies.

I recommend the following:

Start with 1Hr timeframe, add only 2 factors while generating strategies, keep number of indicators between 2 and 4, use small IS and OOS, later add 1 factor to filter strategies.

You need to define the time of the IS and OOS based on your type of trading.

I think, the bigger the IS, the more less performance the strategies will be in a short period of time, but if you are a vampire and will live long enough, then increase your IS as you can 🙂

Yes I think so to.

But with this way of generating strategies, however, you need to know exactly what you’re doing. You choose appropriate indicators, and appropriate trading hours.

Best if If you have trading know how.

トーマス

I made a workflow analysis for this workflow. I checked 5 different time periods.

The result ist clear.

This workflow is only suitable for learning purposes. In order to create profitable strategies with it, you have to modify it.

Hani Hamdan has already given us tips on what we need to do to convert this workflow into a profitable workflow.

I attached my analysis.

トーマス