sleeping dozens compared to sleeping numbers.

[John Gold]

It is my humble opinion that if anybody is ever going to find a way to win at roulette without actually having an edge, then that edge will come from trying to take advantage of the variance within the game.
I was reading some old posts today and in particular looking at some of the statistics that members have provided.
One of them was a sample of a million spins. In the sample, the longest absence of a dozen was 35 spins.
However, if you take any sample of 37 numbers, you will very often find that 12 numbers are missing. This is known by some as the rule of the third. Now if it was simply a case of going by the 1 million spin sample, you could say now one of these 12 numbers is pretty much a good bet to appear on the next few spins and you could design a progression to give you 10 steps or so to catch it without the progression going over table limits.
The thing here is that I have tested it and it is not a sure thing. What this means is that after the 37 spins, there is still the possibility that those 12 missing numbers will still sleep for a further 10 spins and take the number of missing spins to over 47. This was not a rare event either.
So either I am reading it completely wrong or there seems to be some kind of discrepency here. Surely over the million spin sample, a dozen would have slept for a lot more than 35 spins.
It seems that by focusing on single numbers as opposed to streets, lines, dozens, colums etc... the variance can be much larger.
Obviously there could be a multitude of reasons for this. Wear and tear on the wheel creating some kind of bias could be at work. Maybe there is such a thing as a temporary bias which operates on wheels at any given time.
Assuming any of this is correct, then surely it makes sense to attack the wheel from the angle of betting on single numbers as opposed to any other combination or combinations of numbers.

[Bayes]

Hi John,

So either I am reading it completely wrong or there seems to be some kind of discrepency here.

There's no discrepancy, because in the 1 million spins sample only 3 sets of 12 numbers were tracked - dozen 1, dozen 2, and dozen 3. The dozens (as in the outside bets) are only 3 of a possible huge number of ways of picking 12 numbers from 37. If you pick a number (or set of numbers) in advance, then the probability that this particular group will sleep is lower than the probability that any (undefined) set of numbers will sleep.

If this seems confusing, there's a classic problem in probability which might make it clearer. Imagine there are a number of people together in a room; how many would there have to be for the probability to be 50% that some pair of them share a birthday?. The answer is 23, which is a very counter-intuitive result. In other words, if there are 23 people gathered together, there is a 50% chance that 2 have a birthday in common.

But, supposing you were more specific and wanted to know what the probability was that 2 people shared a birthday on, say, April 4th, how many people would you need  for a 50% chance in that case? The answer is 253 - a lot more.

To understand why, you have to see that in the first problem, we want to know whether any of the people in the group has a birthday matching any of the others. So to solve this problem you need to generate a lot of pairs; you need to check whether person 1 shares a birthday with person 2, person 3 etc, then whether person 2 shares a birthday with person 3, person 4 and so on. You are looking for a match from any of the possible birthday (365 days). But in the 2nd problem, there is a restriction in that a pair sharing a birthday other than April 4th is not included, and this removes a lot of potential matches.

If you apply the same reasoning to dozens, the selection of the outside dozens is analogous to the 2nd case of the birthday problem, ie; you are looking for a specific cases (only 3), but if you don't specify them (as you don't in the case of the "law of the third"), then there are many more cases (sets of 12 numbers), and the chance that at least one of them will sleep longer than 35 spins is much higher.

[I have cookies]

I only read the titel of this post and the game has 37 degree of freedom - but here is some-thing you can test if you want.
If you skip the idea of fource your self that a slepping event should show then you can play does events wish has a show when the sleeping events sleep to a certan degree of succes.
Is like reverse the method to deal with favorites wish is much more effective.

During 37 trails is very rare that every individual number will have a show once itch.
There is imbalance and fluctations at rare moments true bias.

Many attempts has been done to capture costum variants or some kind of pattern recognizon to capture the imbalance of certan numbers or sectors wish would be some kind of HG if you succed with out having a true bias behind certan numbers.

So some numbers sleep and some does not and it has its own life with variations.
I assume the best way would find some way to use a certan degree of frequense of attack when certan events show and others sleep.

Steve claims to have CVs wish I dont belive in and he offers to send him trails or a sofware wish you could buy and I would not recommend it if you dont know some one who you trust wish could confirm it works to a certan degree of succes.

Good Luck

PS it would be much better if you implement certan degree of visual ballistics to what i mention above regarding CVs ...

[John Gold]

Thank you Bayes for a detailed answer. I understood it and it makes good sense. 
Thanks for the suggestion IHC, I will look into it more for sure.

[John Gold]

-