@lysergic said:

"I'm obviously just speculating, but my guess is that Rater C only rates shows they think are under-rated, and D only rates shows they think are over-rated. If I'm right, their votes could actually help assess show quality.

You may be right. Then again, you may be wrong: Rater C may sincerely believe that every show they've rated is, indeed a '5', or Rater C may be strategically rating the shows high for some other reason ("I've been to these 49 shows and I'm gonna give them the highest rating possible.") Short of figuring out who Rater C is, and asking them about each show, we'll never know.

Another possibility is that Rater C is so selective about the shows they rate, that they only rate those that we all agree are the "best." This is not so much a response bias as a sampling bias (see tomorrow's blogpost.) Again, how much does it really help if Rater C looks for highly rated shows, listens to those shows, and then rates them highly? Or if they choose to rate only those shows they view as worthy of a '5'?

Again, the beauty of the entropy measure is that one does not need to speculate why someone's ratings follow a given pattern.

But let's come at it from a different perspective. Let's assume that Rater C will submit a show rating for the first night of Dick's 2024. Let's further assume that we agree to bet $100 today, almost a full week before the show happens, on what Rater C will submit. My $100 says that C will submit a '5', and your $100 says it will not be a '5'. Would you take that bet?

I suspect that most folks would not take that bet because, regardless of what happens Dick's 2024 N1, Rater C is gonna come in with a '5'. The measure of entropy, the expected average value of information, is zero because they are perfectly predictable.