blockdamofo

### Author Topic: The original - Regression twoards the mean - Marigny de Grilleau  (Read 20719 times)

0 Members and 1 Guest are viewing this topic.

#### Sputnik

##### The original - Regression twoards the mean - Marigny de Grilleau
« on: May 15, 2015, 09:32:57 PM »

The french word for SD is Ecart

First you have to get the Absolute Ecart when you calculate.
So lets assume you have an sequence with 14 series alternating with two singles present.

Then you take 14 - 2 = 12

Now we want to get the statistical ecart so we continue with...

14 + 2 = 16

Now we take the sqr of 16 = 4

And finally we divide the absolute ecart whit the sqr

12 sqr 4 = 3,00

The Statistical Ecart 3,00

#### Sputnik

##### Re: The original - Regression twoards the mean - Marigny de Grilleau
« Reply #1 on: May 15, 2015, 09:34:54 PM »

In the old days i did many hundred thousands simulations and will end this topic about algorithms or march with money management.
Here i will go deep into how you can measuring the random flow or distribution.
Its based upon taking advantage of the law of series in all existing aspects.

This is the values and existing playing models based upon pure math and probability.
This is how you find you windows of bias or overrepresented events.

Series contra Singles.
Series has the value of 1 and Singles has the value of 1.
There is as many singles as existing series no matter length.

Singles contra series.
Singles has the value of 1 and Series has the value of 1.
there's is as many singles as existing series no matter length.

Singles contra larger series.
Singles has the value of 1
Series of two has the value of 0
Series of three has the value of 1
Series of four has the value of 2
Series of five has the value of 3
Series of six has the value of 4
And so it continues

Series of two contra larger series.
Singles has the value of 0 (you just skip them as none existing)
Series of two has the value of 1
Series of three has the value of 0
Series of four has the value of 1
Series of five has the value of 2
Series of six has the value of 3
Series of seven has the value of 4
And so it continues ...

Series of three contra larger series
Singles has the value of 0 (you just skip them as none existing)
Series of two has the value of 0 (you just skip them as none existing)
Series of three has the value of 1
Series of four has the value of 0
Series of five has the value of 1
Series of six has the value of 2
Series of seven has the value of 3
Series of eight has the value of 4
And so it continues.

Now to the underlying dimension.
You can divide singles only into singles of singles and series of singles.
And you can divide singles series versus series of series.
With the same math and probability measuring the random flow.

#### Sputnik

##### Re: The original - Regression twoards the mean - Marigny de Grilleau
« Reply #2 on: May 15, 2015, 09:46:27 PM »
I have one simulaton software for singles contra series and series contra singles.
I load my txt files from random org with 10.000 trails each.

Just click on next and the sofware will pin Point out 3.0 SD
After that you can click on spin button and see for your self how regression towards the mean behave.

You can also click on win/lose button to see how strong your betting march is.
You can also save output file with results.

Feel free to develop your own march or algorithm.
I call it tendency play as i recommend to bet after regression is present.
So you need indications or triggers.

Here is one example:

singles contra series - March 1

This is flat betting with a total of four attempts.

1) Rule number one is to wait for two series to chop, the formation of two series look like this RRRBBB and after that formation appears you attack twice and gain +1 unit.

If you don´t get a straight hit with three series in a row like this RRRBBRR you will have the following formation RRRBBR and aim for it to hovering with your secound bet RRRBBRBBBB if at zero you just follow the betting behavior, the march, until +1 or -2 being your first attack.

2) If you don't get two series to chop as I mention above you make your first attack when a state of hovering appears and it looks like this RRRBRRR and play that next formation will be followed with a serie like this RRRBRRRBB if not you will have the following formation RRRBRRRBR and would play it will continue to hovering and ride it out until +1 or -2 being your first attack.

3) If first attack fail you would repeat the betting signals above for one more attack with the difference that you don't aim to win +1 and now aim to break even because that would give you one less formation of correction to appers more (variance)  regulary then an longer correction to end up with +1 and this is a way to make the play more stabel with good results.

Result:

+1 +1 +0 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +0 +1 +1 +0 +1 -4 +1 +1 +1 +0 +1 +1 +1 +1 +0 +1 +1 +1 +1 +1 +1 +1

31 won
4 loss
27 total gain

Illustration with my example march:
You wait until two series chop after each other, then you play that the third will show.

R
R

B
B
B
R
R
W

Now if you lose this first bet, then you bet you will win the second bet.

R
R

B
B
B
R
B L
B W

Now lets assume none of the bet wins then the STD start to grow again with out drop point after stop growing stronger.
So now you have to wait for this two betting signals again.
Two series to chop or the hovering state.

R
R

B
B
B

or

B
B
R
B
B

After 3.0 or strong imbalance and this two state start to show as drop point or indication that the bias stop growing stronger - then this two states start to chop or clustering in waves where you end up +1 unit or break even at +0.

If you follow the march above you should get similar results.
+1 +1 +0 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +1 +0 +1 +1 +0 +1 -4 +1 +1 +1 +0 +1 +1 +1 +1 +0 +1 +1 +1 +1 +1 +1 +1
If i remember it correct i once reach 3.0 STD flat betting this way or was it another march, not sure ...

What you will see after 3.0 STD if it stop growing and go into the other direction is the two following states.

R
R

B
B
R
R
R

There will be many events that will become at least three series that chop after each other.
Tiny drop point and larger drop points.

The other common one is the hovering state follow by hovering state or series that chop.

R
R

B
R
R

B
B

Or

R
R

B
R
R
R

B
R
R
R
R

This creates two kinds of Loses and Winnings ...
Direct wins or break even W or LW chops.
When there is no correction you only end up with two loses LL.

I will show that how the regression towards the mean look like.
You can get very long strings of winnings before reaching a loss limit ...

LWWLWLWLWWLWWWWLWLWLWLL LWWLWLWLWWLWLWWLWWLWWLWLL LWLWWLL

All does are series that chop or hovering with out the STD to start growing again.
So if not flat betting, then i would suggest D'alambert as staking plan.

This is the kind or similiar LW-Registry Bayes gets ( i assume ) and others get with there methods when the use tendency play.

For software see attachment:

« Last Edit: May 15, 2015, 10:02:30 PM by Sputnik »

#### Sputnik

##### Re: The original - Regression twoards the mean - Marigny de Grilleau
« Reply #3 on: May 15, 2015, 09:51:52 PM »
Sometimes it does not have to have a drop point or even out.
You win if it just stop growing stronger and stay at same level without getting stronger or weaker.
This is what i call or name the hovering state and i define it being part of the drop point or correction as it is present change.

I have simulation software where you can observe the values.
You can run 10.000 trails and look how the imbalance behave ( The ECART )

By this observations you can build a working march.
You can see what happens after 3 4 5 6 STD windows.

The software show the values on a rolling basis.
This way you can make statistical relevant observation of hundred thousand simulations.

It check R/B H/L O/E and measuring the ECART.
I attach the simulation software ecart and result file of one simulation.

You run the simulation software and load a spin file.
Then the software will convert the file into Ecart values.

This is also a good simulation software for deeper understanding of waves or Ecart play.
The simulation software indicate at right side when it reach 3.0 STD ...

Code: [Select]
`B-0.94    E 1.57    H 1.00B-0.94    E 1.57    H 1.50B-0.94    E 1.57    L 1.50B-0.94    O 1.57    H 1.21B-0.94    E 1.37    L 1.00R-0.94    O 1.18    H-0.85R 1.00    O 1.33    H 0.94R 1.00    O 1.33    H 0.94B 1.00    O 1.33    H 0.94B 1.15    E 1.33    H 0.94R 1.15    E 1.48    L 0.94B 0.97    E 1.48    H 0.69B 1.12    O 1.48    L-0.73B 1.12    O 1.62    L 0.65B 1.12    O 1.62    H 0.65B 1.12    E 1.62    L-1.00R 1.12    E 1.76    L-1.00R 1.26    E 1.76    L-1.00B 1.26    E 1.76    L-1.00B 1.41    O 1.76    L-1.00B 1.41    E 1.58    H-1.00B 1.41    E 1.72    L-1.21R 1.41    O 1.72    H-1.50B 1.23    O 1.85    H-1.21R 1.07    O 1.85    L-1.21B 0.90    E 1.85    H-1.50B 1.04    E 1.98    L-1.70B 1.04    O 1.98    L-1.41R 1.04    O 2.11    H-1.41B 1.00    E 2.11    L-1.61R-0.73    E 2.24    L-1.34B-0.90    O 2.24    L-1.34R-1.06    O 2.50    H-1.34B-1.22    O 2.50    L-1.53R-1.50    E 2.50    H-1.71B-1.70    E 2.67    H-1.46B-1.50    O 2.67    H-1.46R-1.50    E 2.36    H-1.46B-2.00    O 2.14    L-1.46R-2.18    O 2.26    H-1.63B-3.00    E 2.26    H-1.40  */25R-3.15    O 2.00    L-1.40  */25R-2.83    O 2.18    H-1.57B-2.83    O 2.18    H-1.35R-2.98    E 2.18    H-1.35B-3.13    E 2.36    L-1.35  */27R-3.27    E 2.36    H-1.51  */28B-3.41    E 2.36    L-1.67  */29R-3.54    O 2.36    L-1.46  */30B-3.67    E 2.06    L-1.46  */30B-3.40    E 2.24    L-1.46  */30R-3.40    O 2.24    L-1.46  */30R-3.14    E 1.98    H-1.46  */30R-3.14    O 1.73    L-1.62  */31R-3.14    E 1.51    H-1.77  */32B-3.14    E 1.67    H-1.57  */33R-3.27    O 1.67    L-1.57  */33B-3.40    E 1.46    L-1.37  */34B-3.16    O 1.26    H-1.37  */34B-3.16    E-1.21    L-1.52  */35R-3.16    O-1.50    H-1.67  */36B-3.29    E-1.70    H-1.48  */37B-3.05    E-1.41    H-1.48  */37R-3.05    O-1.41    H-1.48  */37R-2.83    E-1.61    H-1.48R-2.83    E-1.50    H-1.48B-2.83    E-1.50    H-1.48R-2.96    E-1.50    L-1.48R-2.74    E-1.50    H-1.62R-2.74    O-1.50    H-1.44R-2.74    O-1.50    L-1.44R-2.74    O-1.50    L-1.26R-2.74    O-1.50    L-1.26R-2.74    O-1.50    H-1.26B-2.74    E-1.50    H-1.09B-2.54    O-1.70    L-1.09B-2.54    E-1.89    H-1.23R-2.54    E-1.61    L-1.37B-2.67    O-1.61    L-1.21B-2.47    O-1.34    L-1.21B-2.47    O-1.34    L-1.21R-2.47    O-1.34    L-1.21B-2.60    O-1.34    L-1.21R-2.72    O-1.34    H-1.21B-2.85    E-1.34    H-1.04B-2.65    E-1.09    L-1.04B-2.65    E-1.09    H-1.18B-2.65    O-1.09    H-1.02B-2.65    E-1.28    H-1.02B-2.65    O-1.46    L-1.02R-2.65    O-1.22    H-1.15R-2.47    E-1.22    L-1.29R-2.47    O-1.40    L-1.13R-2.47    O-1.18    H-1.13B-2.47    E-1.18    H-0.87B-2.29    E 1.00    L-0.87`
See software with attachment
I use txt files from random org with 10.000 trails each

#### Sputnik

##### Re: The original - Regression twoards the mean - Marigny de Grilleau
« Reply #4 on: May 15, 2015, 10:09:57 PM »
The Ecart windows strength.

We deal with independent random flow where singles and series come in any combination.
As you can see with the simulation software above so does the Ecart grow and getting weaker in waves.
The nature of the game is chaos with tiny, medium and large waves of imbalance or some times even out.

Some rules that effect the strength behind the bias window of events.
Say that you search for 3.0 Ecart windows before you are ready to attack.
Then the minim window is 16 events or trails.

Then you have to set a limit that the 3.0 STD should appear during 16 to 30 events and not more or less.
Maybe you set your limit to 16 to 25 events / trails.

This is because if you have a stretch window with 50 events and a Ecart of 3.0.
Then parts of the events during this window is underrepresented events / trails.
That makes the weaker stretch of the window.

Then the likelihood with equilibrium to have some events / trails already been showed - can effect the future drop point to become weaker or with same weak stretch over length as the window with 50 events.
That makes it harder to catch.

So with other words so does the probability window being small give a more rapid drop point overall.
As none of the underrepresented events / trails had a show during the Ecart window.

This is about the length and how much you stretch the bias and limit with your Ecart window - that in the end effect what kind or regression towards the mean you can expect.

#### Sputnik

##### Re: The original - Regression twoards the mean - Marigny de Grilleau
« Reply #5 on: May 15, 2015, 10:16:08 PM »

This is how you use the random flow with known playing models.

Now each red and black event are independent.
And have the value of 1 each.

So the values apply to red and black.
But then how are you going to indentify indications and triggers?

I end this topic with a qoute from Wiki:

"Regression toward the mean simply says that, following an extreme random event, the next random event is likely to be less extreme. In no sense does the future event "compensate for" or "even out" the previous event, though this is assumed in the gambler's fallacy (and variant law of averages). Similarly, the law of large numbers states that in the long term, the average will tend towards the expected value, but makes no statement about individual trials. For example, following a run of 10 heads on a flip of a fair coin (a rare, extreme event), regression to the mean states that the next run of heads will likely be less than 10, while the law of large numbers states that in the long term, this event will likely average out, and the average fraction of heads will tend to 1/2. By contrast, the gambler's fallacy incorrectly assumes that the coin is now "due" for a run of tails, to balance out."

#### random

##### Re: The original - Regression twoards the mean - Marigny de Grilleau
« Reply #6 on: May 25, 2015, 06:35:45 PM »

Quote

singles contra series
- March 1
This is flat betting with a total of four attempts.
1) Rule number one is to wait for two series to chop, the formation of two series look like this RRRBBB and after that formation appears you attack twice and gain +1 unit.If you don´t get a straight hit with three series in a row like this RRRBBRR you will have the following formation RRRBBR and aim for it to hovering with your secound bet RRRBBRBBBB if at zero you just follow the betting behavior, the march, until +1 or -2 being your first attack.
2) If you don't get two series to chop as I mention above you make your first attack when a state of hovering appears and it looks like this RRRBRRR and play that next formation will be followed with a serie like this RRRBRRRBB if not you will have the following formation RRRBRRRBR and would play it will continue to hovering and ride it out until +1 or -2 being your first attack.
3) If first attack fail you would repeat the betting signals above for one more attack with the difference that you don't aim to win +1 and now aim to break even because that would give you one less formation of correction to appers more (variance)  regulary then an longer correction to end up with +1 and this is a way to make the play more stabel with good results.

Trying to experiment with the march you propose. Am i getting it correctly?

Thank you for every enlightenement.
« Last Edit: May 25, 2015, 07:02:30 PM by random »

#### kav

• Hero Member
• Posts: 2397
• Thanked: 1346 times
• Gender:
##### Re: The original - Regression twoards the mean - Marigny de Grilleau
« Reply #7 on: May 25, 2015, 07:20:18 PM »
Great topic! Thank you Sputnik. Welcome to the forum random

#### random

##### Re: The original - Regression twoards the mean - Marigny de Grilleau
« Reply #8 on: May 26, 2015, 03:24:07 PM »
Thank you very much and congrats for the forum and website. Very nice look.

Just realized that my application of the march1 was wrong.
Will try and test it with today's actuals from random.org.
So trigger can be two underrepresented events in a row or
hovering state (underrepresented-overrepresented-underrepresented)
win +1 or -2 for first attack
even out or -4 total for second attack.

I attach the results to check.
« Last Edit: May 26, 2015, 11:22:05 PM by random »

#### Real

• Fighting the war on absurdity one foolish idea at a time.
• Hero Member
• Posts: 1684
• Thanked: 283 times
• Gender:
##### Re: The original - Regression twoards the mean - Marigny de Grilleau
« Reply #9 on: May 27, 2015, 03:02:20 AM »
I don't know which is worse.  The fact that you guys have to test this, or the fact that you don't readily understand the flaw in your line of thinking.

In order for equilibrium to "appear" to happen, all that must happen is the number of trials must grow. I don't know why you guys are wasting your time on such a test.  The results are already widely known.  Such tests have been done, countless times of the century.

If you believe that you already have test results that show it works, then your test results are flawed, you're misinterpreting the data, or there's a bug in the programming.

#### becker

• New
• Posts: 29
##### Re: The original - Regression twoards the mean - Marigny de Grilleau
« Reply #10 on: May 27, 2015, 07:04:53 AM »
I don't know which is worse.  The fact that you guys have to test this, or the fact that you don't readily understand the flaw in your line of thinking.

In order for equilibrium to "appear" to happen, all that must happen is the number of trials must grow. I don't know why you guys are wasting your time on such a test.  The results are already widely known.  Such tests have been done, countless times of the century.

If you believe that you already have test results that show it works, then your test results are flawed, you're misinterpreting the data, or there's a bug in the programming.

I am not sure about what kind of equilibrium you are talking about but point of waiting for a strong deviation isn't that it predicts winners, but it makes runs of losses shorter after that. For example you face 25 reds in 100 spins. Now after that point we can be sure that such deviation wont repeat again having  only 50 reds in 200 spins. Outcomes will start going more balanced from that point (nobody says that red will start hitting more to make up deficiency and that is big difference here) So after you face a strong deviation you can only expect that deviation wont go so severe again. That is what is the thing here. And it is safer to apply some progression and profit from it. Although you still wont have an edge and you still may hit slightly under expectation (in some cases), but you wont face some catastrophic losses which you would if you bet all the time without deviation. It is only important  that deviation wont be so severe in the next length sequence.  Now prove me wrong.

All this is provable and been scientifically tested. And also it happens in many areas of life in different aspects wherever you have some random elements. It is natural phenomenon.
« Last Edit: May 27, 2015, 07:22:34 AM by becker »

#### Mike

##### Re: The original - Regression twoards the mean - Marigny de Grilleau
« Reply #11 on: May 27, 2015, 08:15:28 AM »
If you're going to use past spins as a guide, the rational thing to do is bet on the numbers which are hitting the most. It really makes no sense to a) bet on the even chances, and b) bet on a return to some kind of equilibrium.

Quote
It is only important  that deviation wont be so severe in the next length sequence.

But the trigger of a strong deviation won't give you any advantage. Eventually things will even out, but you cannot rely on it in the short term.

I don't know why  more people don't play systems which are based on the wheel, instead 99% of them are based on the layout. Try to incorporate some strategy which tries to exploit the fact that that you're dealing with a ball and a wheel. Targeting hot numbers and sectors, tracking pocket distances and so on, is at least a step in the right direction, although short of full advantage play.
« Last Edit: May 27, 2015, 08:24:30 AM by Mike »

#### Mike

##### Re: The original - Regression twoards the mean - Marigny de Grilleau
« Reply #12 on: May 27, 2015, 08:22:53 AM »
All this is provable and been scientifically tested.

Can you cite any scientific papers or evidence that this has been proven? I think the concept is used by some traders who use technical analysis (Bollinger Bands), but this is hardly scientific, since technical trading is a bit of a black art and regarded by many as pseudo-science.

#### becker

• New
• Posts: 29
##### Re: The original - Regression twoards the mean - Marigny de Grilleau
« Reply #13 on: May 27, 2015, 08:40:45 AM »
But the trigger of a strong deviation won't give you any advantage. Eventually things will even out, but you cannot rely on it in the short term.

You guys simply cant grasp this... Who says about evening out? And what you mean by this? Nobody expects that red will catch black in next X spins meaning after there were 25 blacks in first 100 spins, now red will start hitting more to even out in the next 100 spins meaning it will be even like 100 vs 100 on 200th spin.

If all this you are saying is true, then whole subject of statistics would be meaningless. And it wouldn't matter about length of a sequence at all. You would see same often length of 1 EC or 20 EC-s in a row, but we know that is absurd of course.

Well if some things don't arrange themselves in the short run to some degree, how they can in the long run then? So we cant say that the long term isn't made of a more short terms?

#### becker

• New
• Posts: 29
##### Re: The original - Regression twoards the mean - Marigny de Grilleau
« Reply #14 on: May 27, 2015, 08:48:59 AM »
All this is provable and been scientifically tested.

Can you cite any scientific papers or evidence that this has been proven? I think the concept is used by some traders who use technical analysis (Bollinger Bands), but this is hardly scientific, since technical trading is a bit of a black art and regarded by many as pseudo-science.