I dislike the general concept of hosting code challenges on SO, so my main feedback would be to get rid of that experiment. With that out of the way, here's some specific criticism of the current challenge:

• I would expect a "code challenge" to focus on, well, writing code. The challenge here is about steganography, hiding a message in a board game. At first glance, the primary goal of this challenge is to invent a clever way to map the state of a board game to a character set, so basically creating a mapping or algorithm for building a state from a message and reading a message from a state. All of that could be done in plain english and it's unclear why any code would be required at all, or what role the code should fulfill specifically.
• As people can choose any board game and any encoding, the solution space seems pretty big, which means comparing submissions to others is close to impossible. You acknowledge that your current way of simply counting upvotes to determine the best solutions is not good, but for a challenge this broad, you will never be able to find an objective way of judging.
• As you don't exclude scrabble, I forsee 275897823 entries with some variety of "simply write the message on a scrabble board", maybe in rot13 if someone is clever. There are also various other games which allow for trivial solutions (e.g. a 6 bit encoding in connect four). I don't see anything for weeding out trivial or troll submissions, which means you might get flooded in them and interesting submissions might get lost in the noise.