The probability of biased is b = p – .5 (In a non-biased condition, the probability of biased would be 0 because p would equal .5) Using the XOR does not completely get rid of the bias but it helps a lot. The bias after the XOR, the bias is b = 2 * b^2
EXAMPLE: If p = .75, the original bias is b = .75 - .5 = .25 The bias AFTER the XOR becomes b = 2 * .25^2 = .125
The derivation of this equation can be seen at: http://crypto.stanford.edu/~dabo/courses/cs355_fall07/yao.pdf
