Because we actually have data and some scientific understand about gender differences, that question is not very representative of bias. Change gender with race like: "an unseeded white player beat the top-seeded African American tennis player" - Should you rate it less likely to be true if the race was reversed?

Trump vs Obama becomes less a bias test the more factual information a participant has. The test I can only guess assume that participants do not actually have factual information about either president, thus all being equal they should not have anything concrete to base a logic conclusion.