I think you misunderstand - you don't always play black in chess. So on average white wins 50% of the time (modulo first move advantage and draws as the article states).
But they can observe that Alice always wins when playing against Bob. (If they somehow can determine that one player is black and the other is white, they probably can also distinguish the players as well?)
They can't see a chess piece but they can see humans well enough to tell them apart?
Analogies don't work if you deliberately make assumptions that are inconsistent with what they are trying to convey.
We could rephrase the analogy such that the aliens are listening to a specific broadcast from earth from 1000 light years away. Some percentage of the time the broadcast is "White won", the rest of the time the broadcast is "White lost." Is it reasonable to assume that it's a game of pure chance just because you don't have a good method of predicting the outcome?
Yeah yeah, I agree. But, I think, the thing is that once you start making so many carefully made assumptions what is and isn't happening, then the nice parable format falls apart quickly.