I think that the problem is that we have the option of choosing "Other." You can tell who my favorite Zelda is judging by my profile pic.
The way you construct these polls will affect the type of answers you get. I do not mean how you phrase the poll question and the answers. I mean what the poll is looking for in the first place.
N choose 1 type of poll such as this one will tell us how many people have a character as their favorite.
N choose n type of poll will tell us how many people like the character.
Upvote/downvote type of poll will tell us how many people like or dislike the character.
Rate the character type of poll will tell us how well liked the character is.
Lets look at an example. Suppose we have Zelda, Samus, and Lara Croft and a sample of 10 people.
6 people love Samus (rating of 10), 1 person likes her (rating of 7), and 3 people dislike her (rating of 1).
1 person loves Zelda, and 9 people like Zelda.
3 people love Lara Croft, 1 person likes her, and 6 people dislike her.
N choose 1: This poll will ask the users to pick their favorite. You will have Samus 6, Lara 3, and Zelda 1.
N choose n: This poll will ask the user to choose all the options they like. You will have Zelda 10, Samus 7, Lara 4.
Upvote/downvote: This is a Youtube style like/dislike voting system. You will have Zelda 10, Samus 4, Lara -2.
Rate the character: This will be a survey that asks the user to rate each character individually. This will give Zelda 7.3, Samus 7, Lara 4.3.
If you want to know how well the fandom (Smash or Zelda) likes each Zelda as a whole, the third or fourth options would give you the best answer.