I also work on a GA futures trading algo, the thing is : 1- how do you combine conditions? is it C1 And C2 and C3 ... or do you give room for more complex strategies such (C1 or C2) And C3..... 2- Regarding crossover, do you just take a split point of two randomly selected fit dna's or do you account for the type of indicators being combined and the dna's being combined too? because you can have two great parents but their child is really bad. 3 - do you gurantee that overall generation fitness is increasing to some extent each generation or is it somewhat random?
hello, the specifics of the SQ genetic algo are not entirely known to me since it's proprietary. I can only play with the options available to me. To answer your questions: 1) By default, AND conditions are used. I think you can code a custom template with a combination of OR and AND, but I haven't tried it. SQ also has a fuzzy logic option, whereby an entry is triggered when a certain % of conditions are met, but I don't use it. 2) I don't know the exact details, but I think the parents are selected probabilistically based on fitness, and the split point is randomly determined. 3) Since fitter parents have a higher probability of being selected, the fitness SHOULD increase over the generations. But for times where it stagnates, there is an option to restart generation after a certain number of generations have stagnated.
@@TradingTact like there maybe many variation of genetic algorithms that might come to play so yes you are configuring the setting under genetic options tab but which algorithm is used exactly..btw there is mot only one genetic algo
@@fromzerotospyro i'm not sure if the SQ GA has a specific category. As a customer I don't know the exact ins-and-outs of it. I just explain it in terms of whatever info the developers have made available, and what's available in the GA literature
hello again, SQ contains a few preprogrammed candlestick patterns like hammers and bullish engulfings. You can add those patterns into a builder template, and then run the genetic builder on that template. This video explains the template process: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-6ONJoWj0tBM.html
Thanks. It is not entirely clear to me when I should choose Random Generation or Geneitic Evolution in SQX. I thought GO was quicker than RG for a large set of building blocks but a recent try eroded this belief. What do you think ?
Genetic evolution will probably need more time to generate strategies, because of all the processes running in the background, but you will need much less time to arrive at profitable combinations. The flip side is that you will need more robustness testing to counter all the curve fitting.
That is correct. I've made many changes, hopefully better consistency will follow. Bad months/years are a reality of trading. Fortunately, am still up 70% since Feb 2021.