The statistics and probabilistic methods of science are there specifically for controlling and quantifying the effect of noise and uncertainty and making sure experiments are reproducible a high percentage of the time.

If you don't get a quantifiable amount of reproducibility, there is no point to using statistics at all and what you are doing is not science.