They're using the specific example of GF(2) as the field. Any given assignment of values to variables then gives a value of 0 or 1, so there's a 50:50 chance of getting it right.
If you're working over F_p then about 1/p of the randomly chosen vectors will satisfy the equation.
If you're working over F_p then about 1/p of the randomly chosen vectors will satisfy the equation.