When searching for successful parameters for systems, each indicator/building block added increases the search area by many multiples. This creates HUGE search areas very quickly, which even the fastest machines would struggle to completely map.
Adding and removing indicators/building blocks is a vital step in the successful search outcome.
However at the moment we cannot tell with any surety if an indicator/block being searched through is offering any results.
How many of the x thousand of selected best fitness results have used y block? Is y block never used so should be discarded? Is y block often used and a successful indicator?
Having a count on how many times each indicator is selected in the databank would be extremely useful in refining the search area to only those indicators which are of value.
Do you see value in adding this to SQX?
Say that there exists 10 Billion people on every planet, 1 Billion planets in every solar system, 200 Billion solar systems in every galaxy and 500 Billion galaxies in the universe. If every single person on every planet has been shuffling decks of cards completely at random at 1 Million shuffles per second since the BEGINNING OF TIME, every possible deck combination would still yet to have been "shuffled"."
That's only 52 variables with 13 options. We have MANY more multiples of that. Giving us an aid to get that number down is very valuable.
and to come with those number of variations we have genetics algorithms, which will find the ways what works and what not
Attachment unknown.png added
I posted a great idea about this subject in our Discord channel,
Already found a cool name for this "Automated Internal Weight Adjustment" - AIWA
This method will have the same concept of Blocks weightings but it will be automatically adjusting those weights based on block's performance after multiple runs.
More detailed explanation:
Lets say we want to start our Builder,
We selected all our Blocks and even gave them some weightings here and there based on our assumptions.., like usual.
Now here when this thing comes to play its part and save us huge time of searching:
So the builder will start searching for strategies with those rules above in-mind,
in addition to that, we want now to drop-off and remove those blocks that performed poorly,
and we want to uplift those blocks who did performed profitably,
EXAMPLE:
Builder ran 1000 times already,
We set AIWA to evaluate our "Internal Weights" each 100 runs..
So each 100 times AIWA will check evaluate what blocks ware profitable the most and which ware un-profitable and re-adjust the weightings each 100 runs.
Benefits:
We will use use the most relevent blocks to the spesific Instrument/TimeFrame we want to search for,
This will reduce the time of searching by many!
Few Points:
The AIWA method needs to be flexible so the user can have the option to use it or not.
All the parameters of the method should be opened to the user so he can change for example after how many times the AIWA will kick in to adjust the weightings (100 as the example above).
We focused only on the block it self so It does not matter what parameters the block used in the evaluations of AIWA.
ALL the AIWA evaluations will be based on our Fitness Criteria.
Thats all, Original Idea was by (Fatso Wombat on Discord which i believe it is Stormin_Norman here),
I just thought about an awesome way to implement it :)
Status changed from New to Refused
You can also export whole databank content to Excel, so you can count it there with a little coding or macros.
Not everything needs to be in SQ.
"the true answer is to reduce the number of tunable parameters to an absolute minimum to make a robust strategy" is a quote from John Ehlers' recent video entitled "How to optimize strategies for robustness (properly)"