I'll reiterate a point that he made while presenting these two charts: there is no claim to 100% accuracy here. Even a large sample size could not provide that. The only people with 100% accuracy on this matter are Blizzard Developers, and they aren't telling.
In fact, I'll soon be arguing that developers ought to be ashamed of this, so it's no wonder they don't want to talk about it.
So, check these charts: (and please let me apologize for the weird blur. This always happens when I post charts, and I just don't know what to do about it).
This first shows BG activity prior to 2011. Note that the percentage range is only between 40 and 60%, so the wide distance between bars isn't really as sharp as it might first appear.
The second seems to include data only since the start of 2011. It also encompasses new battlegrounds.
Quite simply, this is unacceptable.