The "polls" are people who are willing to set the betting lines for other people and Kalshi is deriving a probability on it. Let's say the best offer right now for Franklin NOT to be the next head coach is 3:1 odds. Somebody feels really confident about Franklin being the coach, so they go in and they stake $100 to raise the offer to 4:1. Kalshi moves the probability up when that happens. If somebody takes that offer and bets to win that $100 at 4:1 odds that Franklin won't be the coach, it moves the best odds being offered back down to the 3:1 odds. Kalshi then moves the probability back down. But there's not enough action on the non-Franklin candidates for their probabilities to be a reliable gauge. Those can be easily moved by somebody betting on those candidates or offering different lines that make Kalshi calculate weird probabilities, which is what happened with Chesney briefly.