I don't agree. I think it depends on what you test. I have a small channel (less than 2K subscribers, not yet monetized), and I am trying the thumbnail test for the first time with a video I just posted. I have brief, bold text in my thumbnails and I did three variations only changing the text. (After crafting a single thumbnail like I always do, it took very little time to spin up two more variations - doing the first one is the hard thing still for me.) After only 3 days, I see very interesting results with differences of click-thru at 25%, 30%, and 44% (test is still running). The first one, which would have been my only thumbnail before using a/b testing, is doing the worst at only 25%, so if these results hold up, the testing will be valuable and result in a thumbnail getting better results.