If you get a losing backtest with a certain asset selection, you should of course not trade it nevertheless. In such a case you have two valid options. Either you go back to the original selection and trade it. Or you do not trade the system at all, not even with a different asset set, since your test has shown that it is apparently too sensitive to asset selection.

We have not optimized the asset selections, but just collected the assets from the top 10 or top 5 lists of cryptocurrency websites. However we had not experienced any negative 2017 backtest in that way. I think it is in fact quite difficult to get a negative result with cryptos in 2017.