Can the RMS running count be preserved by some shuffles?

[Automatic Monkey]

If, at the end of the shuffle, there is a lot of variation between density of high and low cards in different parts of the shoe, you're going to get either a high RC or a low RC, and which one and where you get it are going to depend on the cut which is pretty much random. But if the high and low cards are evenly distributed the balanced RC isn't going to deviate far from 0 and you're going to have a "dead" shoe no matter where it is cut. No opportunity to either raise your bet or Wong out and this is a negative EV situation.

Now for certain shuffles that are actually used, could a low RMS RC have a higher than random probability of being carried over to the next shoe? I think it's possible that if the cards are handled the exact same way every shoe, that the count could go into a repeating pattern that it will take a while to get out of. So if (and only if) this is the case, it would make sense to switch tables if you get a shoe where the count never strays far from 0 because of a greater likelihood of the next shoe also being negative EV. This is a little different from regular shuffle tracking theory because I'm not thinking about trying to determine where the good/bad counts are, just if they are more or less likely to occur than at random. This will not require any extra work on the part of a counter other than just remembering to change tables.

Hope this all was clear. Does anyone know if this has been researched in the literature yet? Thanks.

 

[alienated]

Harvey Cohen has discussed this question *LINK*

What you are contemplating is not inconceivable. Harvey Cohen has spent some time on this question and posted some of his findings at reg.gambling.blackjack. You might like to do a google search. Some key words might be 'peak', 'parent', 'child', and of course 'rec.gambling.blackjack', 'harvey' and 'cohen'. The link below is to a short post by Cohen that summarizes the basic idea. Apparently Doug Grant had previously suggested that the running count peak of child (new) shoes tends to be about 75% of parent (previous) shoes. Cohen disagreed with the 75% figure, but appeared to agree that the effect may be present in commonly used shuffles. He suggests that the running count peak of the child shoe will tend to be equal to the average running count peak plus one fifth of the difference between the parent shoe's peak and the average running count peak; i.e.

E(C) = x + 0.2(P - x)

where E(C) is the expected peak of the child shoe, x is the average peak of all shoes, and P is the peak of the parent shoe. Cohen offers an example where the average running count peak of all shoes is 15 and the parent shoe had a peak of 35, implying an expected peak for the child shoe of 15 + 0.2(35 - 15) = 19. Cohen notes that the strength of the relationship between the running count peaks of parent and child shoes will depend on the precise nature of the shuffle. Simulation would probably be required to determine if knowledge of the parent shoe's running count peak offered any additional information for the particular shuffle you happened to be facing.