I think that is the next big thing - teaching the algo program recognize graphical patterns instead of purely mathematical/computed ones.
I'm pretty new to the algo process, but so far so good. I'm using MT5/cqg data. I didnt try to make it too complicated, just a simple vol expansion setup. I didn't even backtest it that much 2020-ytd looks good so i went with it. But compared to the high flyers, this is like a retarded turtle performance. 3-4% a month maybe 40-50 per annum based on volatility. But it fits my goals as being passive.
Don't be so mystified by "algo". Look you can run an Algo buy or sell on IB TWS with a simple click. Sure, they are just price arrival algos, but people seem to lump them together with complete systems which are 100x more of about everything. Learn the basics and then write some simple ones yourself. Then start adding in all the more difficult parts, mostly exception handling in both signals and order execution. Then add in risk management, then .... With scripting systems that are built in a lot of platforms, it is a simple next step after a custom TA indicator.
Of course it is. Every indicator is mathmatically written. Every algo is mathmatically written. The algo basically wants every instrument you follow to recognize the written instructions. An algo is just instructions for following your many ideas down into fine detail consistently for every instrument you follow. Writing algos is an art, you have to enjoy maths and being very fussy, my current favorite algo at this point in time took 21 attempts and months of writing. It tells me what not to trade, what to trade, when to trade and uses a grading method, best to worst. Its basically an assistant to keep you on track on your trading plan. AI will not work imo unless you yourself custom build it. The internet version of AI is as per usual, following the crowd.
Algo writing is feasible, you just have to be a bit nerdy, like sitting hunched over a keyboard for hours plotting lines of code, its not for everyone that lifestyle. What I've found though, writing algos generates new ideas. I have several algos, all doing different jobs, another favorite is one I use when US mkt closes, it is like a stock screener, tells me which stocks in which sectors are performing, so it tracks hundreds of US stocks and they are placed in their sectors. At a glance can see where the money is moving.
This video is nonsense. The provided bullet points describe a holy grail strategy. The assumption should be - it’s very hard to make money in the market’s consistently, so whatever you come up with is most likely crap. Now prove to yourself otherwise.
IF your idea or trading mechanism is consistent, but your trading fails at human factor, then algorithmic trading is the way to go. However most people prefer to lose their money by hand, so to speak. Algorithm will always be biased to market conditions in which it performs best, just the nature of the beast. Your problem becomes managing an algorithm like you would an employee or a team. YES, you can't rely on one algorithm, you should have distinctly behaving units to rotate between sandbox and playground.
I and many others have said, "trade what you SEE". The chart formations reveal themselves as they develop. The only complex formations are "complex bottom" and "complex top"... and you have time to see them develop. And then there is the "match low"... you see a price low and the market is approaching that price again. If it bounces there, you have a matched low to play a long.... vice versa for a matched high sell and/or short.