It doesn't have to be. As long as you can specify what your system actually does, you can test it, and you can falsify it as a hypothesis about how the market w

It doesn't have to be. As long as you can specify what your system actually does, you can test it, and you can falsify it as a hypothesis about how the market works. However it has to be done in a statistically sound way. Just showing a backtest that makes money or beats the market is not enough evidence.