Regression towards the mean.

I write this topic for my self and if you are interested, then you are more then welcome to contribute.
I been thinking about the STD and how you measuring things.

So one red is one event and one black is one event.
This mean i can measuring red and black using math.
I can measuring balance and imbalance.

This means i can also measuring loses and winnings in the same way.
Where one loss is one event and one win is one event.

And if we take the law of series, then one singles is one event and one series is one event.
50/50 situation as there is as many singles as there is series.

Different benchmark and different variance.

The bell curve has no limit as nothing is due to happen, but the bell curve tend to change after reaching 3.0 STD.
There is three main states to talk about and observe.
One is that the imbalance can continue to grow stronger or it can hovering around zero state or getting weaker.
The two last states is what you hope to catch after a strong imbalance, hovering state and opposite draw-downs.

But you can create your own state and benchmark with regression towards the mean.
A wave of overrepresented events can be 1.5 STD and you look for tendency of change and playing that the waves will change before reaching 3.0 STD, but only if there is present change that indicate hovering state or draw-downs.
They come in tiny, medium and large states.

You lower you set the imbalance of overrepresented events you more action you will get and in the same time higher variance.
With other words a more bumpy ride.

Mapping or clustering patterns.

Bayes have made software that match and miss-match patterns to get winning and losing sequences.
Where each winning and losing mark is one event and you can act upon that information.

Drazen once mention if i remember it correct 28 loses and 2 wins.
That is 4.74 STD.

But the beauty of the animal is that you would never place any bets during your observations as there is no tendency for present change.
Hovering state or draw-downs.

This made me thinking, so i come up with my own clustering march of the random flow where i set my own benchmark for regression towards the mean.
I pick a more bumpy ride.

The benefit of this methodology.

That is when imbalance grow stronger, you do nothing, so in that way you are never chasing for the opposite to show.
You just observe the flow grow until you get your indications of correction.
They come in tiny, medium and large states.

This is the true meaning with tendency play or you could also name it trending, but not based upon guessing.
This methodology is based upon probability and math.

Ego....IMO....that's observing SD obviously.


And I agree....BUT


Like series and single are equal over time, or Rvs B and they grow to high SD where you can play for regression,  they do something else much much more.


Its none of the above.....


What they do most is ebb and flow in small waves. You can bet this simply and all the time and win more than "not" playing these and waiting for high SD. High SD is rare. Small waves are not. They ...well...are always there apart from when one carries on to a high SD....which can take a while.


If you wait for say, 6 against 1, then wait for 6 against 2 to predict the SD has turned back, on average, a series of 6 is seen once in 74 spins RRRRRRB.....quicker is RRRBRRR with series of 3 expected in 7 spins...but it doesn't play nice like that, and the other 4 spins in 7 will probably be singles...so your RRRBRRR with an SD of 1.9.. is more likely to be  RRRBRBBRRRBRB with an SD of +1
In other words, The higher the series, the higher the SD of R vs B and the longer you wait to see it.


SD3.0 is RRRRRRRRR  and the probability tells us you will see this once in 654 spins


Or RRRRRRBRRRRRR with series of 6 which should appear once in 74 spins....so its highly improbable to have 2 x 6 in 13 spins. And that's the best you can do for SD3..... 2 x 6 R and 1 B




You can see visibly that SD turns back more before it got too high by the fact there are more RBR, than RRBR, than RRRBR, than RRRRBR etc.


Am I talking out of my Ar.se Baxksie?....perhaps I am.

QuoteWhat they do most is ebb and flow in small waves. You can bet this simply and all the time and win more than "not" playing these and waiting for high SD. High SD is rare. Small waves are not. They ...well...are always there apart from when one carries on to a high SD....which can take a while.

I have also been thinking about that ...
Current development with my march show it can be done to certain extant ...

At least when STD grow you do nothing, that is a good sign.
But when the STD fall back to back, then you get caught.
The question is if there is any clever MM that can handle does situations.

One other thing i been thinking about is how it works when you match and miss-match patterns to get LW-Registry to act upon.
Lets say i have eight events and then match and miss-match opposite and same with eight future events.
Then i have 16 events.
Now the question is ,,, if my last eight results/events comes as zig zag OSOSOSOS do i then have 16 singles?
Then it would be above 3.0 STD ...

As you can see so can you twist how to measuring the random flow in different ways.
Bayes match and miss-matching pattern software comes with two sequences.
Hovering state or zig zag state.

There is work to explore regarding this topic.
Personally so do i think this is the best way attacking even money bets.
Because we do at least base our bets upon probability and math and not guessing with taste 

QuoteThe question is if there is any clever MM that can handle does situations

Nothing that you could print on paper, its up to YOU the player to make a decision as to when to increase your unit size to recover some of the losses at that point. In my opinion you cannot use the same MM every time, I am sure Drazen will agree with me on that, no 2 sessions are the same, if they are it will be very rare to use the same staking sizes.

QuoteThis means i can also measuring loses and winnings in the same way.
Where one loss is one event and one win is one event.

Quite right, as you may know I play red black follow the last, I don't bet every spin but bet most of them, from my decisions over time I know what MY LLWWLLLLWWWWWWL looks like, or ends up looking like during a session, so I can work with that to decide if I can safely, or as near as damn it, increase to recoup.

As Drazen said, practice, practice, practice at YOUR way of playing so you know how it should perform.

Good luck

QuoteQuite right, as you may know I play red black follow the last, I don't bet every spin but bet most of them, from my decisions over time I know what MY LLWWLLLLWWWWWWL looks like, or ends up looking like during a session, so I can work with that to decide if I can safely, or as near as damn it, increase to recoup.

I have similar conclusion that you can use any kind of march that create a LW-Registry and act upon that.
I done some testing and some times i flat betting 300 trails with out raising any bets, pretty cool.

It also feel pretty cool being the father behind this method, as i first write about the subject in 2009
http://www.vlsroulette.com/index.php?topic=11336.msg73476#msg73476

Albalaha have this signature: You can be a winner in gambling only if you know how to survive through the best, the average and the worst sessions.

I think i am very close to solve that issue with even money bets based upon the methodology above.
I feel confident as i have identify the worst that can happen.

EXPERIMENT - Predetermined march

As the random flow is independent, so can you clustering and slice it into any algorithm you like, as long you are consistent.
Lets assume i would use "Follow The Last" then it would look like this.

I see red i would follow red.
If i win then stop and start from the beginning again, wait for a new outcome and follow it once.

So if you would get wins chopping then it would look like this.

R
R W

B
B W

B
B W

You slice the random flow into blocks or groups and it's a reason for that.
Lets look what happens when you get one single and one series where 1+1=0 as the cancel each other out, the hovering state at zero ground.
This is because series has the value of 1 and singles has the value of 1 and you have equally many of both 50/50 ...

R
B L
B W

B
R L
R W

Now watch happen when you get three singles events in a row as clustering pattern with two loses.

R
B L
R L

B
R L
B L

This is a good example how to use LW-Registry as part of a march as Follow The Last where you identify each state.
Series to chop or singles to chop or hovering state.

So if you use ECART play using FTL march you can have pretty clever MM to handle worst case scenario.
Then you would observe singles as your main imbalance and get FTL to pin point out hovering state or direct draw-down with series chopping.

The math and probability is simple.
14 singles and 2 series is 3.0 STD
7 singles and 1 series is 1.5 STD
Or you could get even down more, but then variance increase.