[DarkMagus]

Well, he's right, that model DOES assume win-rate is constant.

Unless you're a robot, it probably isn't constant, since you have to deal with things like tilt that can compound a losing streak.

If you have played tons and tons of hands, then you can average out your win-rate and pretend it's a constant. You'd get a fairly good approximation. But you'd probably require over a hundred thousand hands, and would also have to assume that your true win-rate remains roughly the same between the start and end of your sample. And of course it also assumes you never go on tilt.