Yesterday I was watching a bunch of my friends play a word association party game. The basic idea is that you have two teams, and on each team one person is the caller. There are twenty five cards with random words on them laid out on the table, and the caller has to try to give clues to get their teammates to guess their own cards without guessing the opponent's. Teams alternate giving clues until one team discovers all of their own cards. Each clue has to be a word and a number, and the obvious thing the numbers are supposed to mean is 'how many cards match this clue'.
So if the cards were
I might say "Animal, 2" to get my partner to guess Dog and Cat.
I started immediately thinking about how to optimize the clue giving to improve my odds. The people playing were all just playing like the rulebook suggests, getting 1-2 cards each turn. Fairly quickly I realized the optimal solution to this problem is to simply use an 18 digit number after the word to describe exactly which cards were the ones to pick. Each set of 2 digits describes a particular card in the group, so the word itself doesn't even matter, the 18 digit number tells your partner all 9 cards they need to pick to win the game.
The rulebook does not suggest that this strategy would be against the rules. They apparently hadn't counted on me trying to ruin a perfectly fine party game by applying thinking to the problem.
However, my solution doesn't simply provide a really good strategy - it outright ruins the game. Given this, I decided to try to think of some more cool strategies that were powerful but which didn't render the game a coinflip nor completely remove the whole word association aspect. One thing I decided is that the number you give definitely has to be a single digit integer. Allowing people to go up to 9 to establish signals should definitely give me some cool design space, but hopefully will allow the players to still play the game.
What I was thinking with this is just leading with the assumption that any number you give above 0 is a signal as to card placement. So if I said "Animal, 5" my team should assume that they should find precisely 2 cards associated with Animal and then pick the fifth unknown card in order. This loses out if you can actually clue 4 cards at once with a single category, but lets you get 3 cards on the great majority of your clues, and I suspect that is good enough. You could use 0 as a signal to just pick the single card associated with the word you state, which seems like a thing you want to have in your arsenal but which is pretty bad.
By my way of thinking, the power to nearly always get 3 is the ticket to winning. If you screw it up, fine, but aggressively going after good turns is going to be the way to win the majority of the games.
In my head this kind of thinking means that people who make party games should totally hire me to review their rule sections to write things to prevent shenanigans like this. I suspect what it really means is that I don't get how people play party games at all.