I was asking OP. I am at this point pretty convinced that he actually used deep learning methodologies not ML in the strict sense. Hardly any ML methodologies produce indecipherable algorithms and models.
Buy and hold would achieve the same thing - no? Buy and hold covers everything including saving money on commissions and saving time on generating 450 strategies.
I was just reading and watching more videos on this subject, many sources describe Deep Learning as a subset of Machine Learning. I guess we should ask @frostengine if he could clarify for us if he is using Deep Learning?
Buy and hold is like working for a living. You only have so many hours a week in which to work. If you have strategies it's like trading on margin that you can scale up. You can't scale up your hard manual labor because there is only one of you and only 24 hours in a day. Of course there is no guarantee of success and it might take you 10 years to be profitable. But if in your 11th year you make just as much income as the previous 10 and then you have another 20 years to go then it clearly makes sense to keep trying.
The way I was thinking, with 450 strategies, prolly a very large number are working against each other, cancelling out the winners. So in which case, it's better to work just a couple of strategies and go longer term. There's no free lunch with trading, every strategy has advantages and disadvantages. Winners work a while, then they don't. Losers lose a while, then they don't. Attempting at mkt timing is full of traps. Logic sometimes works, often doesn't. The more you trade, higher the costs.
Jim Simons refers to it as machine learning but then what he actually describes sounded more like regular backtesting: Video starts at relevant bit, 29:50 minutes in
I use the term Machine Learning as a catch all. It's easily recognizable by most without getting overly semantic. Deep learning is simply a subset of machine learning. Yes, what I do would be more considered deep learning.