I seem to be getting that question a lot.
Just an actual situation I'm dealing with. The scenario I describe is not hypothetical — I have to create a test, and understanding these various probabilities will help me know what format of test to create. The idea is to create a test that will minimize a student's odds of succeeding purely by guessing; and then, as a secondary priority and to the extent possible, to also minimize a student's odds of succeeding by eliminating incorrect questions and then
guessing. The reason for these goals is to minimize, as much as possible, the extent to which test-taking skill and pure (good) luck are reflected in the results, such that the results will more accurately reflect what a student actually knows
That's my context, and I need to make a data-driven decision.