It seems some of your ideas basically amount to "no techniques really work, and people only think they work because of random variation and measurement error and stuff". Not sure how plausible this is.

Currently, it's pretty darn plausible.