Top Answer asks you a survey question and shows you the ranked list of answers people gave, with every one hidden. Each answer carries the share of people who said it, and finding it adds that share to your score. You get five strikes, and the point of the game is guessing what everybody else thought rather than what you think.
The question is "name a topping people put on pizza" and the board has five slots. You type mushroom and it lands in slot two at 18%. You type cheese and it lands in slot three at 14%. Slot one is still blank, which tells you something more popular than mushroom is still out there — and for pizza that can only really be pepperoni. Type peperoni, one letter short, and it still lands, because an answer that long tolerates two characters of misspelling. Ham, meanwhile, is a perfectly sensible topping and a strike, because nobody in the survey said it.
The list is what people actually said, not what is correct or interesting. For "name a fruit" the crowd says apple and banana, not persimmon. Your first three guesses should be the three most boring answers you can think of, because the boring answers are the ones carrying the big percentages.
The slots are ordered by popularity and stay in place as you fill them, so an empty slot above one you have found is a live clue. If you have filled ranks two and four, you know there is a more popular answer than the one at two and something sitting between them, which narrows the kind of answer you should be reaching for.
The category line above the question sets the register. A geography question wants countries and capitals, not regions; a food question usually wants the plain everyday name of a thing. Matching the level of specificity the category implies saves strikes on answers that are right in spirit and absent from the list.
Loose spelling matching forgives typos but it does not forgive a different word. If chips has landed, fries will not land again as a second answer, and trying a near-synonym of something already on the board is one of the cheapest ways to lose a strike. Move to a genuinely different answer instead.
Skipping costs exactly the same strike a wrong guess costs and gives you nothing back — no reveal, no clue. A guess you only half-believe in is strictly better, because it at least has a chance of landing.