回複：IBM 一月

(1)
Consider the strategy: pick a cut off value b, if the last ticket value a>b, stop, otherwise continune.

Let y (as a function of b) is the future expected return if the last ticket value a y = \int_b^1 x dx + y b -c
so y = 1/2(1+b)-c/(1-b)
y is maximized when b = 1-sqrt(2c), and y_max = 1-sqrt(2c)

Hence the strategy is: don't play the game when c>1/2; if c
(2)
Still working on this, my first thought is that we should use the same strategy as in (1)...