we are going to have to agree to disagree on my #6. a true probability method does not allow you to delete balls of any color, no matter what happens. otherwise you would also have to delete a successful ball for every success. by the same token, an 80% success rate applies to each iteration. the complement, a 20% failure rate, when serially ANDed becomes a truly rare event very quickly. but it still has the same 20% probability each time. run a series of 80% prob 10,000 times and see how many long strings of bad results show up. there is a way to calculate the chances of say, 5 bads in a row. surprisingly high when you are playing with real money.
My example was for a defined period in the future that one is predicting (from the historic probability of 80%), a future 80% result. For 24 months in the future our model predicts 80%. Therefore if a hypothetical 80% result is "Given", then each loss(20%) will, in fact make your next cycle have a probability of less than 20%.
Yes, if "Given" an infinite future time period to get a 20% loss showing up, you could have years of consecutive losses and still retain your 20% loss rate. The difference between that hypothetical and mine is that we know from history that the 20% loss rate "occurs" historically over 24 cycles, not infinite cycles.
Therefore if we are using history as a probability statistic for the future(same time period...I.e. 24 months) then the ball analogy has to be statistically true. So with our "Given" (20% loss over 24 cycles) each trade after a loss, is that much closer to a win over that defined time period.