right, but 50% of the effect of opiates is placebo, so we shouldn't be too surprised. https://www.ncbi.nlm.nih.gov/pubmed/11207392

Placebo's effect on pain is one of if not the largest placebo effect, but the effect size on sleep is much smaller.

Well the reasoning was pointing towards high chance of placebo though. He said there was lots of grime and also something he could measure was different, which means psychologically a lot greater chance and effect for placebo.

I don't think that's how it works. In the hypothetical world where he reports no grime on filters and no difference in readings yet his sleep is better, would you predict lower chance of placebo?

That is actually a great thought experiment.

A treatment being a placebo implies nothing about the effect size!

Most placebo effects are modest and the ones that aren't are usually mixed in with regression to the mean effects(which are usually pretty modest). So OP is implying that if an effect is especially large it's less likely to be placebo.

Placebo effects on depression are enormous.