If you make millions of dollars with this trading algorithm don't blame me! This was merely meant as an exercise, to feel like I've dabbled in the exciting world of trading. It was fun to think about strategies and price graphs. This runs, though I cannot vouch for its correctness. Please contact me if you want to discuss :)
What I've done actually doesn't totally make sense. I've trained an algorithm and then tried the algorithm on the training data. And it didn't work well. I hope to get back to this once I finish the prototype of my game, which is currently my number 1 side project priority (as of 12/26/21).
The program reads in a set of prices (a price history) that was manually retrieved from Binance and creates 'stretches' - a stretch is meant ot represent a period of time (a stretch of time) that a price is moving in a single direction (up or down, I can't remember how I handle staying constant). Average stretches are calculated: the average time of a price momvement. For instance, the average up stretch is 45 minutes, the average down is 29 minutes. The idea is that at 'the current time' I look at the current direction and how long the stretch has been. If the average stretch is longer than the current stretch, then hold the asset. Once the current stretch is wthin x time of the average stretch, sell.
Currently buy max will spend all of the starting funds on a single coin. Also, sell max sells all of the coins available at once.
If I started off this trading strategy at the beginning of the coin, I would've been better off just holding.
The tail of the program output:
------------------------
profit: $-998.2803310037507, 0.0x of initial savings
price: $1915.07
price increasing
Bought 0.0 coins at $1915.07 at time 936 for $1
savings: 0 num_coins: 0.0004266411871086535
------------------
Profits if held from first day:
14938.14
total_buys: 240
total_sells: 239
This is a very unsophisicated algorithm! get_average_stretch only considers how long a stretch is, not the slope of the stretch* - it does not consider how much a stretch is increasing or decreasing, just that it's doing one or the other. I
Another problem is that I train on the data I'm evaluating!
I think I'm also realizing I didn't implement a short.
I'm also realizing there are probably a million more of these algorithms on quantopian like sites :)
https://omscs.gatech.edu/cs-7646-machine-learning-trading https://find.minlib.net/iii/encore/record/C__Rb3814960__Spython%20for%20finance__Orightresult__U__X7?lang=eng&suite=cobalt https://www.amazon.com/Machine-Learning-Tom-M-Mitchell/dp/0070428077 https://www.amazon.com/What-Hedge-Funds-Really-Introduction/dp/1631570897/ref=pd_lpo_2?pd_rd_i=1631570897&psc=1