So a lot of it is documented on the README, but the simplified process is this:
Training process:
The training process takes into account successful trades - failed trades and the overall portfolio value. There is also a time_delta so it gives bias to current trends. This is so that the bot is more reactive and this makes sense because we shouldn't give an equal ranking to a strategy that worked 4 years ago but isn't performing now vs a strategy that worked terrible 4 years ago but is working wonderful now.
Trading process:
It only buys & sells from the NDAQ-100 tickers - this is so that the securities are vetted. Each ticker is run through every strategies, then those decisions are given weights based on their ranks on the training data. It runs the trading bot and buys on basis of which has the highest buy weight - sell weight since funds are limited. If the sell coefficient is higher than hold and buy, it will automatically sell.
In terms of back period, optimization isn't one of the primary concerns. In fact, there's actually a finding that proves that optimizing back period is not as good strategy as working with different random periods because if tends to overfit and make the strategy rather static. Anyways, thank you for you interest, and please let me know if you need any help with set ups.
Training process:
The training process takes into account successful trades - failed trades and the overall portfolio value. There is also a time_delta so it gives bias to current trends. This is so that the bot is more reactive and this makes sense because we shouldn't give an equal ranking to a strategy that worked 4 years ago but isn't performing now vs a strategy that worked terrible 4 years ago but is working wonderful now.
Trading process:
It only buys & sells from the NDAQ-100 tickers - this is so that the securities are vetted. Each ticker is run through every strategies, then those decisions are given weights based on their ranks on the training data. It runs the trading bot and buys on basis of which has the highest buy weight - sell weight since funds are limited. If the sell coefficient is higher than hold and buy, it will automatically sell.
In terms of back period, optimization isn't one of the primary concerns. In fact, there's actually a finding that proves that optimizing back period is not as good strategy as working with different random periods because if tends to overfit and make the strategy rather static. Anyways, thank you for you interest, and please let me know if you need any help with set ups.