I just posted a message on bjmath,
http://www.bjmath.com/bin-cgi/bjmath.pl?read=4309,
in which I propose a simple model for which progressive betting is useful. The only (bold) assumption is that the deck composition change can be described by Gaussain (rather than Poisson) random process, which I believe is reasonable. Otherwise the model is pretty realistic.
If you don't understand the "scientifice method", please ignore the message.
~Jack