SF Signal has been tracking the number of books released monthly in our genre for the past few months. They are running just shy of 300 per month. Now, some of those aren’t eligible for the Hugo novel, so let’s cut that down to 200 books per month. That means 2400 books a year are eligible to short-list, meaning each book has a 0.2% chance of making the short-list. (5 / 2400)
Any attempt to get a valid statistical sample of a population of 2400 by looking at 5 individuals will not result (except by random chance) in an accurate representation of the total population. It will produce wild swings.
Secondly, your analysis fails to consider the fact that the entire population changes each year. In other words, after we've looked at this year’s slate of potential candidates, we not only shuffle the deck, we go out and get a new one. Again based on statistics, there should be no correlation between this year and any other year.
[Freer writes] Elvis (very popular) on the head of a coin, and a cat-litterbox on the tail, it will always land heads? – irrelevant. We’re not randomly tossing coins, we’re actively looking for ones with Elvis on the head and carefully displaying them heads-up.
[Freer writes] sample of teenagers with $200 who like that sort of music you do realize you don’t actually have to go to Worldcon to vote – supporting (AKA voting) memberships cost about $40.
[Freer writes] 10 years of novel data. 50 events – that’s 10 events, not 50, looking at 10 entirely different population sets.