Our members are dedicated to PASSION and PURPOSE without drama!

Regression toward the mean and SD

Started by GreatGrampa, September 05, 2013, 03:56:09 PM

Previous topic - Next topic

0 Members and 3 Guests are viewing this topic.

GreatGrampa

Well, am not a great fan of these techniques, but I can safely say that I have been fortunate enough to know people who plays these techniques with greater amount of success and have learnt a lot from them. As some of you have pointed out that this is a subject which is like "I will tell you, but have to kill you", I have decided to post the knowledge I have gathered from some of my close friends who practice this technique. I am not worried about them killing me, as I am anyway counting my days :)


Standard deviation
Now anyone who wants to read something about standard deviation, I would suggest reading this wonderful article from our mathematician friend Bayes
http://www.rouletteforum.cc/index.php?topic=1093.0

Things can't get simpler than that and I am not going to explain here what 2SD means or what 3SD means. That will essentially defeat the purpose of this thread. And moreover, I can't put that simpler than what Bayes has put in his thread. What I will explain is the practical applicability of this in roulette in simple English with some examples and explain the common fallacies and pitfalls that you should avoid. As in all the threads that I have initiated, you need to develop your own strategy around it, eventhough I will touch upon a couple of methods of using it. In fact one who would like to understand statistics in a fun and easy way, I would recommend the book Cartoon guide to statistics by Larry Gonick. Very nice read :) , but don't expect to become a statistician after that :)

Regression to mean
The explanation goes it is the phenomenon in which if a variable is extreme on its first measurement, it will tend to be closer to the average on its second measurement—and, paradoxically, if it is extreme on its second measurement, it will tend to have been closer to the average on its first. Using Regression, as with any statistic concepts, you cannot be certain of something to happen, but you can certainly say that it will happen within a degree of predictability.

Simple applicability in roulette could be (ofcourse some people might disagree as statistics is common sense and what is common sense is always questionable), you see that there are 10 Reds in a row, then in the next 10 spins, there is a higher degree of probability for a balanced mix of reds and blacks to be present. Often people misunderstand that the next 10 spins will have more blacks than reds or completely black so that it will all even out. NO! That's not the case. Regression to the mean will just imply that the next set of 10 spins will have a higher chance of a mix of blacks and reds. Again note, nothing is certain.

Now in my next post, I shall explain some concepts that one should take into consideration while practicing regression to the mean.

Greatgrampa - Your friend and mine

GreatGrampa

Sample size
Sample size is very critical if you have to adopt any statistics techniques. Standard deviation will vary between a sample of 10 spins, 100 spins and 300 spins. What we are looking here is practical applicability in roulette and hence any sample size that we take need to be playable. Waiting for a higher sample size will take the fun away from the game and make it really insane. Typically (or should I say commonly) people consider 3 out 10 of an EC as a deviation that has crossed SD levels and start betting from that position onwards.

To me anything between 8 to 10 is a decent sample size to consider to be playable.  For all practical purposes in this thread, I will consider a sample size of 8 spins and look for deviations from mean within this sample size.

Chasing the deviation
Often people commit the mistake of chasing endlessly after a deviation. That is equivalent of betting black to appear endlessly after a series of reds and with a progression that might dig their own graves. To make the entire subject playable and affordable and not write your will to the casino, the best is not to chase it endlessly. Then how much should one chase?

Well! Different friends of mine has different opinion. For one just placing 2 bets is sufficient chasing. Most common is placing 3 bets as it fits nicely within a 1,1,2 or a 124 or 123 progression. My advice – chase until half of your sample size. That is one of the reasons why we have chosen a sample size of 8. We will chase until half of this sample size which is four spins. Chasing until half the sample size will give a good indication of whether your chosen approach on betting is a winner or loser for the next sample size. Again, I stress on good indication because nothing is certain.

Next post, I will explore on how our bet selection should look like when doing these bets.
Greatgrampa - Your friend and mine

Blood Angel

Luck happens when Preparation meets Opportunity.

Kav


Sputnik


Nice topic GG ...

This is how regression towards the mean looks like using a modified march by Marigny.
After at least 7 reds or blacks.

+1 WWLWLWWLL       
+1 WLWWWWWWLL       
+1 LWWLWWLL       
+1 LLWWWWLL       
+1 WLWWWWWLWLWWWWLL   
+1 WWWL           
-3 LLWLL           
+1 WWLWLL         

Here is other results.

920108 Ecart 3.44 +1
920124 Ecart 3.27 +1
920127 Ecart 3.00 +1
920131 Ecart 3.00 +1
920205 Ecart 3.40 +1
920211 Ecart 3.12 +1
920213 Ecart 3.41 .X
920214 Ecart 3.12 +1
920218 Ecart 3.12 .X
920219 Ecart 3.27 +1
920302 Ecart 3.64 +1
920325 Ecart 3.40 +1
920406 Ecart 3.00 +1
920428 Ecart 3.00 +1
920519 Ecart 3.54 -3
920520 Ecart 3.00 +1
920527 Ecart 3.15 +1
920609 Ecart 3.00 -3
920612 Ecart 3.15 .X
920613 Ecart 3.00 +1
920617 Ecart 3.00 +1
920622 Ecart 3.00 +1
920716 Ecart 3.00 +1
920720 Ecart 3.00 +1
920723 Ecart 3.00 +1
920810 Ecart 3.02 .X
920818 Ecart 3.15 +1
920824 Ecart 3.00 +1
920826 Ecart 3.15 -3
920827 Ecart 3.15 +1
920903 Ecart 3.15 +1
920904 Ecart 3.00 +1
920909 Ecart 3.00 +1
920915 Ecart 3.18 +1
920923 Ecart 3.02 +1
920924 Ecart 3.00 +1
920930 Ecart 3.29 +1
921002 Ecart 3.12 +1
921005 Ecart 3.88 +1
921005 Ecart 3.15 +1
921014 Ecart 3.70 -3
921022 Ecart 3.50 .X
921024 Ecart 3.65 +1
921027 Ecart 3.00 +1
921030 Ecart 3.00 .X


TOTALS 35/12


Turner

GG...you point to Bayes post...which is great...but for a good further read look no further than the help file for analysis in Roulette Xtreme.
Pull up the stat graphs and click help. It tells you how its calculated and what each ellement is. This is where i got my understanding of SD.. etc

GreatGrampa

Thanks Turner, very interesting pointer. And Sputnik, nice little LW registry there.

Now, lets get into the bet selection process.

Bet selection
Hmm! There are a number of ways you can play SD and regression. My preferred one is on even chances. When I say even chances, it could be colours, odd/even, high/lows, 9/9 splits, 3/3 lines, 6/6 streets, LW registries, single/series.. i could go on and on. I will touch upon various selection methods as and when i go through examples, so that it can be seen better. But I will restrict them only to even chances to start with.

On a side note, I remember once Lanky using John solitude's method of standard deviation on straight ups to greater success.

Dominance/following the wheel vs regression
I think this is one place where most of us stutter and we tend to take a fallacy driven approach. This is where, i believe the approach from some of the professional players are distinct. Eventhough, there might be an adoption amongst some of my freinds to a more complicated method that works for them, I will try to simplify this for the purpose of understanding and then you can devise your own playable method around it. Most of us (including myself) used to think that when I bet for something to regress after an extreme deviation, I will go opposing the deviation. For example, if i have seen 7 Evens and 1 Odd in my sample set of 8 spins, I used to and many of us still suggest betting on Odd for a defined set of spins.

Think about it! What does regression to mean imply? It says if one of the sample set of measurements is extreme, the other set tends to be closer to mean. It doesn't say that the other set will drive towards opposite extreme. So in our example, we expect the second set of 8 spins to be a more balanced mix of evens and odds. Now take these two together! What is more clever? Betting on odds to appear atleast once in the next 4 spins or betting evens to appear atleast once in the next 4 spins. I think the answer is no brainer - bet on even. Why? Lets explore a bit further.

You have three possibilities that you can expect in the next set of 8 spins.

Possibility 1 - If you have observed deviation towards one side, it is possible that the deviation is getting even extreme on that side (Good chance of winning betting evens)

Possibility 2 - If you have observed deviation, you could experience regression to mean, which will make a more balanced outcome in the next set (good chance of winning betting evens)

Possibility 3 - If you have observed deviation towards one side, it is possible that you could experience deviation on the other side. (good chance of winning betting odds)

Now 2 out of 3 possibilities, ask you to bet on evens or the dominant side in the deviation and 1 out of 3 asks you to chose the dormant side. what's your choice! My choice is to go with the dominant side. Ofcourse as I said, it is questionable and we may agree to disagree with those believing in going for the dormant side, but apply a bit of common sense and see for yourselves what works for you. Following the wheel works in most cases and it is an art to follow the wheel. It is not as straightforward as going with the dominant one, but for simplicity, lets assume it just means that.

I have deliberately not used terms in statistics to make it plain and practical. Hope you are finding this useful. Next post, I will explore using the concepts that we have discussed so far in a sample set of spins using a simple bet selection.
Greatgrampa - Your friend and mine

GreatGrampa

Greatgrampa - Your friend and mine

GreatGrampa

Singles and series
It is an interesting concept. We all know about the law of series and how every increment of one series is a 50% probability on a lower series.  Now I will treat this slightly different from pure red/black and high/low or odd/even scenarios. One might argue that it is the same, what difference do you figure between a simple EC and a single/series combination.

The reasoning I will try explaining in simple english without getting into the complex terminologies.

Red is a red, black is a black. Red can be one of 18 numbers, black can be one of 18 numbers. Single is a single, but a series constitutes of the following. A single, a series of 2, a series of 3, a series of 4 and so on and so forth. Again, am not denying the fact that a single to become a series is 50% probability ignoring zeroes and a series to continue as a series is a similar probability. But what am trying to bring out is the fact that it has a slight differentiation with respect to a plain even chance.

The differentiation is while red and black are independent events, series and singles are dependent events. You can just drop one spin to get a number and say that it is a red or black. But can you drop just one spin and say that whether it is a single or series? No you can't. That essentially is what am saying as the difference. It is a dependent event. You need one spin on both the sides to definitely say that whether it is a single or a series. You following me? And the permutations are endless and almost unplayable when you take all the kinds of streaks that will form.

With that in mind, lets see in the next post whether we should treat it differently from other ECs while adopting a regression based play. Enjoy your weekend and if I see you on Monday I will continue the topic.
Greatgrampa - Your friend and mine

Kav

Hello GreatGrampa,

Where are you?
Anyone know?

Badger

Sadly, he passed away.
His niece, Priyanka, used to post on his behalf, but she has also left the forum.