Newcomb's paradox (summary) to me sort of falls apart when you quit playing with the parameters and think about it from a pro poker player's point of view.

The payout has already been set before the player appears. The player has no ability to change anything, the player has only the ability to choose a box, and the predictor cannot change the payout in between asking the player to sit down at the table and when handing the player their choice.

Therefor, there's only one choice from the gambler's point of view: pick A+B, i.e. maximize the payout *regardless of what the predictor's accuracy and precision are* because there's no possible way the player can know or influence the predictor's ability to guess what the player is going to do once the player knows she's involved in the game.

The only way choosing B is a possibility is (and here's the sci-fi writer's chisel into the game) *if and only if* the player knows the game is on, and that the predictor's accuracy and precision are not predetermined, and there is sufficient time for the player to use this knowledge before being faced with the choice.