It just occurred to me (I am a bit slow) that while the data analysis is interesting, it is correlated from tests that are NOT testing the test taker - rather testing for a specific difference in beer. I may misunderstand how triangles really work, but the foci is on the test. We might be able to correlate tester confidence in their answers, but not measure their specific skills by reverse engineering this data.
In other words, when people say, I am good at taking triangle tests, that is a meaningless statement. Specifically that being right or wrong is null in the analysis and a triangle test levels the playing field between unskilled and skilled when applied in this manner. It is a form of implied confirmation bias that there SHOULD be a difference between the beers, when statistically there may NOT be a difference, based solely on the tasting skills.
I feel like we always get caught up in our hypothesis that we should taste the difference, and that somehow our skills are lesser if we do not.
Then again, I might just suck at triangle taste testing