If you find something that works let me know. When working with random data it's always possible to come
to a solution that looks more probable than the others. The most probable will hit more often overall but this
does not tell us when. The problem is further compounded when working with more than one string.
In the above example " (0 45 2%) (1 299 11%) (2 792 29%) (3 1035 38%) (4 530 20%)." If we add the totals
of 2 and 4 we get 49%, so the two combined are greater than 3=38%. This is how random works. What we
need to know is, "When?" If we play the most probable then we will hit more often but don't expect it to hit on
que. The three top values would do very well game to game but then we are faced with too many lines to play.
My solution to this problem was to create a gauntlet of filters setting each one so that it covered the top 2 or 3
best values. Again this suffers the same effect if too many filters are required to reach a play budget.
Keep in mind that 0,1,2 and 4 will show in 62% of the draws and 3 is only expected to show 3.8 out of 10 on average.
0 and 1 can be eliminated as both added together equal 13% or around 1.3 out of 10 games on average. These 2 are
small enough to exclude without any real loss but lets say that you have to make a selection from 5 strings. The 13%*5
then becomes relevant in at least one of the 5 strings, which one, I don't have a clue. Sufficiently random is not random
but as it's name implies it's close enough to derail our best efforts much of the time.
Momentum seems to do well in prediction and one thing I have found is that a value, be it a filter, number, digit etc..
The #1 best time to play a value is right after it's self. A skip of zero shows more that any other skip.