Forget the content of these challenges. It is impossible to collect a reliable data from the comments, process and get statistically meaningful result. Totally futile attempt. I am not an experienced data analyst but there is no need for expertise to see the outcome... I guess who cares as long and much as you can manipulate your audience towards your way...
The only goal there is obfuscation, and it's safe to assume that the 1st test didn't reach the conclusion he was looking for, hence attempt #2. There will be no useful conclusion.
Since the 1st test appears to be the original recording and the other two were recordings of his systems, and the 2nd test appears to be again original recording vs recording of his system, my guess is that he was expecting that people would prefer the latter two. And/or that there is some sort of gotcha, but I can't really see what it could be in context of GR-R vs ASR.
Also, recordings are just that; recordings. Some recordings are good, some are bad, and in his case, we don't really know the original is even supposed to sound like, if it's not the 1st track. It's even possible that some bad recordings might sound better
after being re-recorded with a 2nd system; which will effectively add crossfeed, equalization, maybe 'spaciousness' or whatever attribute. In audio, people often prefer objectively worse sound reproduction, as sadly high fidelity isn't about high fidelity for most audiofools, rather a set of subjective preferences. So such exercise as
commenting on the playback of a recording of a playback of an audio master (mix/manipulation of multiple recordings) will indeed result in nothing of value.
But knowing what I know about Danny, just like having any discussion with him, the 'challenges' will turn out to be completely useless. His test doesn't exist to provide anything of value, in his snake-oil vs ASR, it likely only exist as a smoke screen, for whatever ludicrous point he wants to make.