I basically use primarily small and medium bets and hope that sporadic big bets will make up for losses and propel me forward. And the big bets are not so big that you can't recover.
I forgot to mention the probability. My current thoughts are that a probability above 50 is the way to go, because, well, you're putting the odds in your favor. Of course the ratio of money lost to money won increases, but again, by mixing in big bets, you can hypothetically always make up for losses.
Just tested this out again and I have some things to add:
It seems one should also vary the probability, because this enables one to increase your profits without increasing your bet size. So it seems like you can gain a definite advantage over the machine, it's just that you need keep changing things, and the software doesn't make that a quick, streamlined process, so it would be pretty laborious.
That's going to work every time until the time it doesn't. You can make money on bets with a negative expected value by being lucky. there is no way to consistently be lucky. Even if the probability is above 50% the odds aren't in your favor because they aren't paying you the correct odds.