How Frequently Should I re-run SQ or change settings?

Currently, I am running SQ for days, harvesting 3-5k strategies and repeating the process. My intention is to collect a large pool of strategies before I run any testing or before changing to another set of settings or building blocks.

So my question is:
Are there any advantages or disadvantages in running SQ repeatedly without changing any settings in generation or ranking?
The question behind the question is what is SQ ability to generate UNIQUE strategies after long periods of strategy generation compared to fresh starts every day or 2?

Always randomized searching,

it does not matter when you start or restart your generations,

you can have a very unique strategy only when just pressed the Start button, or you can even not get any strategy for days to come,

its random, thats the beauty of it 🙂

just dont set a very strict criteria at the Rankings tab and you will be fine, after some time you will learn what Ranking criteria you should set to avoid bad results,

using strict criteria will make your searching hard and you will doubtfully find many strategies that pass your strict criteria, start by setting simple rules like DD% + Trading Number > 100 and then make a research of that fits you best and then add more criteria like Stagnation In Days and etc…

Always randomized searching,

it does not matter when you start or restart your generations,

 

you can have a very unique strategy only when just pressed the Start button, or you can even not get any strategy for days to come,

its random, thats the beauty of it 🙂

 

just dont set a very strict criteria at the Rankings tab and you will be fine, after some time you will learn what Ranking criteria you should set to avoid bad results,

 

using strict criteria will make your searching hard and you will doubtfully find many strategies that pass your strict criteria, start by setting simple rules like DD% + Trading Number > 100 and then make a research of that fits you best and then add more criteria like Stagnation In Days and etc…

Thank you Karish. I like your contributions. The next question is has this assumption been test or is it based on the knowledge and belief that the generation random is random and remains random?

Hello,

the random generation remains random while genetic is random on a first/initial population only

what tomas said :), good luck

Hello,

 

the random generation remains random while genetic is random on a first/initial population only

Thank you Tomas262. Could you please explain what the statement means? What is first population? And on what are subsequent generations based on? Does the system learn what it has done in the past (first batch/generation) so avoids repeat in subsequent generations? I have noted that for example PF seems to be dropping – hence my interest in asking these questions.

what tomas meant is that when you use Genetic Generation then your first initial population is already set a head and will not be random,

and from those initial population the Genetic process will be random until some point that the genetic will begin to build it self up into its roots and progress from there,

i would recommend you to stick with Random Generation rather then Genetic Generation if you are a newbie,

i was also a newbie ones and started with Random Generation and using it since then and never looking back, the theory of Random Gen is that it can generate even what Genetic Gen can, but Genetic Gen cannot generate what Random Gen can,

so i would say use only Random Gen, you also welcome to join into our mind group on Slack, just PM me your email and i will send you an invite if you not already with us there,

there are a lot of helpful info e-books articles tips and tricks that we share there, few very important stuff already pinned and highlighted for new joiners and over all,

good luck hope i helped.

Hi Karish, am already a Slack member and do read the posts there as well. Thank you for your advice above and yes, am very new to SQ and programmatic ea creation and development. So am reading extensively. Sometimes, concepts can be difficult at the beginning but then am learn quickly.

Thank you again for your support.

I have gone back and read and watched further training again! As said above, if Genetic evolution is based on the first random generation, then theoretically, there is a benefit is restarting genetic evolution periodically so you have a chance of genetically evolving from more varied initial samples. Because if the initial random result is not that great, the resulting evolution outcomes will be limited. Am I right in my thinking?

If what I have said makes sense, how often should one restart the process? How can one tell if the evolution process has reached its limit? Stagnation?

Yes, with the genetic evolution process SQ creates the initial population and tries to combine the strongest members to create a new version based on them. Once this process stops showing signs of improvement everything starts over with a new initial population. This is why there is the evolution management included on the genetic options tab so you can define when the restart happens

Thanks Tomas262.

Following on then, would it be acceptable (or beneficial) by FIRST collecting Random Generated strats and then feeding the strats to SQ to Genetically Evolve? I have a collection of 100 random strats, which I would like to feed to SQ to evolve. How do I import a random strat into SQ to genetically evolve?

Hello,

now you can let evolve only a single strategy using the improve function. It can be a strategy produced by SQ or manually defined in editor. Not sure now about exact form of this feature in SQ4. It needs to be checked with Mark.

The ability to define the entire initial population of strategies for evolution process manually might be useful though. Now only missing function is in databank to load STR files. It is already possible to use databank content for initial population

Thank you tomas262, that is really useful information. I wanted to use one randomly generated strategy at a time to evolve genetically. Of course it would wonderful to have a way to automatically by run these experiments batch of strats from random generation into evolution. I cannot see how currently I can import databank content into generate strategies tab for initial population. Could you guide me?

Hello,

I realized that SQ offers a load tool for the initial population already. Check the screenshot attached

File: databank_init_pop.jpg

Thank you, sir.

Hi Tomas262,

It has been fascinating loading up randomly generated strats as initial population for Genetic evolution.

First observation – random generation does not create strategies as fast s Genetic evolution. Can anyone explain?

Second, running Genetic evolution based on pre-prepared Randomly generated strats population is generating strategies faster than Genetic evolution. Can anyone explain this?

Has anyone tested these two approaches to see which one is more efficient?