Yeah, I went through the same thing when writing this last year and I decided to leave it the way it is even though mypy complained. Here's my thinking—we know these methods are supposed to return a tuple of two things. So, in this case I think our hints are correct even though mypy says they are not. My understanding is that the "fixed" hint is saying we will return a tuple of multiple things of the same type. Now, it probably wants that because technically methods like choices() and nlargest() could return more than 2 things, so that's what their type signatures say. But the parameters we are using ensure they always return 2. So, even though our hints are technically wrong for the purposes of type checking, they are right for the purposes of readability.

Does that make sense? I'm going to leave this open so we can discuss it further. I am open to changing it though if the consensus is that it's more important the type hints be type checker correct than human reader correct.

Your reasoning definitely makes sense. It's that trade-off of doing what you know is correct or writing to the limitations of the type checker. I'm using Visual Studio Code with my linter set to mypy so I tend to write to the limitations of the type checker. Otherwise I have PROBLEMS ❷ yelling at me all the time. But I'd never argue that this is the one and only way to do it.