Sorry to pile on, but this doesn't look great. Most people with any quantitative finance experience will tell you this is the wrong approach / the wrong application of these types of techniques. ML in trading is primarily used for portfolio optimization and relative value-type analysis. It (and especially Reinforcement Learning) are not useful if you a just trying to build signals to trade with. And Fibonacci Retracements are only kinda useful as a visual tool for assets that are mean-reverting in some capacity (i.e. forex). It is applicable to Crypto to an extent, but the volatility of Crypto nullifies most of the usefulness as this as an actual trading signal.

If you really want to pursue this, it would be helpful if you provided a bit of info on how the strategies worked, what type of hedging / risk management is going on, etc.

Best of luck