But 250 SEK(Swedish Crowns, equivalent of ~25 EUR/USD) is expected to be consumed by commission. But if I rebase the curve on percentage increase in stock price and a fixed commission, the top curve shows approximately what the result would have looked like if I'd injected 8k SEK.
My actual increase in value over four days between A and B would have been 4.71%. Which is probably, but not certainly, due to luck. 4.71% over four days amounts to over 6000% on a yearly basis. That is highly unrealistic irl.
But I've just injected a bit more cache. Learning money, let's call it. This type of hands-on is a lot better than a theoretical course on the stock market, and probably a lot cheaper too. In a few more days I should know if the robot was just lucky, or if I actually could use it as it is.
My thinking is still that I need to transition to using normal stock (ETFs) to be able to make a profit. And possibly also implement an AI instead of some dumb if-statements. Currently I only use the stock price as input, but I may need to use the volume too to get a good enough estimator. As of yet I hardly understand the workings of a convolutional ANN, and there seems to be a little bit of a theoretical threshold before one can get started on such a venture. For the first time in years I'd have to start digging into documentation. Peuh!
