Whether or not a trading algorithm generates yields is irrelevant, if they are significantly worse than a buy-and-hold strategy. Given these charts (which are all bull markets, btw), it seems like the algorithm doesn’t really do better than the hodl approach at all. In fact, it looks like it underperforms a lot of the time. This is a worry, since you said that you needed to spend a lot of time optimizing your parameters to fit these particular data sets. In other words, you are over-fitting your model to the backtesting data, and you are not showing blind rund results, nor results of bear market trading.

In my mind, the only measure of a trading algorithm is whether it consistently beats the market, and whether it can be profitable in a bear market.