Author Topic: Proving existence of lower variance  (Read 7260 times)

becker

  • New
  • **
  • Posts: 29
Proving existence of lower variance
« on: May 14, 2015, 04:36:01 PM »
Some people are claiming that regression toward the mean is a good way how variance can be reduced in this game. And I see there are some people here, who seems to have good understanding of this phenomena in roulette.

But I didn't saw anyone shown this over some bigger number of bets. Doing this manually would take too long and it is totally unpractical in my opinion.

So my question is: What is the simplest way this can be proven and shown in a representative sample for roulette?

So maybe this should be more of a programming question, as some computer simulation would be perfect for this.

Anyone who can maybe do this?
« Last Edit: May 14, 2015, 04:51:04 PM by becker »


 

kav

  • www.Roulette30.com
  • Administrator
  • Hero Member
  • *****
  • Posts: 1709
  • Thanked: 721 times
  • Gender: Male
Re: Proving existence of lower variance
« Reply #1 on: May 14, 2015, 09:35:16 PM »
Hmmm

Some time ago, Bayes created (for me) a little program that calculates how even chances behave after one of them has heavily dominated the previous spins. It is actually a program that you can run many timesand each time it produces different results, like a monte carlo simulation.

Would that be of interest to you? Is it relevant to your question?
If yes, I could upload it.
 

dobbelsteen

  • Hero Member
  • ******
  • Posts: 1293
  • Thanked: 310 times
Re: Proving existence of lower variance
« Reply #2 on: May 14, 2015, 10:01:57 PM »
My analyse exel  for HIGH and LOW is very suitable to answer your question. It is free available.
 

Reyth

  • Global Moderator
  • Hero Member
  • *****
  • Posts: 3335
  • Thanked: 988 times
Re: Proving existence of lower variance
« Reply #3 on: May 15, 2015, 01:24:28 AM »

Code: [Select]
' Stations/Straight Losses
' 7/58, 10/35?,11/40, 12/31, 13/28, 14/20, 15/20, 16/19, 17/27, 18/11
5 DIM so(500)
7 INPUT "# of stations"; s
10 RANDOMIZE TIMER
20 rs = INT(RND * 37): sp = sp + 1
30 IF rs < s THEN so(sl) = so(sl) + 1: sl = 0: GOTO 70 'output
40 'loss
50 sl = sl + 1: IF sl > ml THEN ml = sl
60 ' output
70 'CLS: PRINT "Spins:"; sp
80 'PRINT "Losses:"; ml
90 IF sp = 16000000 THEN 120
101 a$ = INKEY$: IF a$ = "q" OR a$ = "Q" THEN END
110 GOTO 20

120 cr = cr + 1
130 CLS: PRINT "Completed Results:"; cr
135 IF ml >= oml THEN oml = ml
140 PRINT "Max Loss:"; oml
142 PRINT: GOSUB 200
150 sp = 0: GOTO 20

200 FOR i = 1 TO 58
210 PRINT so(i);: so(i) = 0: NEXT i
220 RETURN

This will output the maximum number of straight losses per inside betting station you specify.  I think its a great demonstration about the "mean" and how roulette is an extremely balanced & reliable game.
 

dobbelsteen

  • Hero Member
  • ******
  • Posts: 1293
  • Thanked: 310 times
Re: Proving existence of lower variance
« Reply #4 on: May 15, 2015, 08:45:24 AM »
For an outsider is a VB-program unsuitable.
An exel sheet is more direct and give you results with one touch of the F9 button
Knowledge, experience and skill is the basic to become a succesful roulette player.
 

becker

  • New
  • **
  • Posts: 29
Re: Proving existence of lower variance
« Reply #5 on: May 15, 2015, 09:07:59 AM »
Hmmm

Some time ago, Bayes created (for me) a little program that calculates how even chances behave after one of them has heavily dominated the previous spins. It is actually a program that you can run many timesand each time it produces different results, like a monte carlo simulation.

Would that be of interest to you? Is it relevant to your question?
If yes, I could upload it.

Yes, please. Thanks for the offer. I hope Bayes also agrees.
 

Reyth

  • Global Moderator
  • Hero Member
  • *****
  • Posts: 3335
  • Thanked: 988 times
Re: Proving existence of lower variance
« Reply #6 on: May 15, 2015, 09:13:12 AM »
For an outsider is a VB-program unsuitable.
An exel sheet is more direct and give you results with one touch of the F9 button
Knowledge, experience and skill is the basic to become a succesful roulette player.

LOL.  Anyone ever tell you that you have the ability and habit of being quite direct?

LOL and don't force me to quote your posts in a huge wall!
 

dobbelsteen

  • Hero Member
  • ******
  • Posts: 1293
  • Thanked: 310 times
Re: Proving existence of lower variance
« Reply #7 on: May 15, 2015, 11:21:18 AM »
Reyth i am of another generation. I try Always to be polite, fair and honest
 I learned my first program language, I suppose long before you were born.
 

becker

  • New
  • **
  • Posts: 29
Re: Proving existence of lower variance
« Reply #8 on: May 15, 2015, 12:05:32 PM »
@ Reyth

Many thanks for your efforts but to be honest I dont know what and how to do with that code...

I was thinking if you could for example show that test (simulation) through the two comparing graphs for example?

To compare raw variance in only betting one EC all the time and variance with the bet you suggested.

 

Bayes

  • Veteran Member
  • ****
  • Posts: 688
  • Thanked: 547 times
  • roulettician.com
Re: Proving existence of lower variance
« Reply #9 on: May 15, 2015, 12:33:43 PM »
becker,

You really need something more accurate than a graph to compare the "control group" with the other data (your hypothesis). I have some software which will analyze the results and report whether there is a statistically significant difference between the two groups in terms of variances, but first I need the data. One way of comparing the bet selections would be to generate "gap lengths" between wins for each bet selection, then the lengths could be compared and any significant differences tested for, both in the mean gap length and the variance.

The control group will be betting continually on red (or any ec). You could do it for any other bet on the layout but the EC's will be easier to calculate.

What did you have in have in mind for the other group?
 

becker

  • New
  • **
  • Posts: 29
Re: Proving existence of lower variance
« Reply #10 on: May 15, 2015, 01:40:39 PM »
Thanks Slacker

Well I didnt had anything complicated in mind. Just something to see what all it takes to prove lower variance in RTM.

Lets say next 10 spins after 10 in a row of the same EC as the bet which should bring lower variance.

(Of course for easier testing it is the same as betting against sequence of last 10 EC-s in a row)

 

kav

  • www.Roulette30.com
  • Administrator
  • Hero Member
  • *****
  • Posts: 1709
  • Thanked: 721 times
  • Gender: Male
Re: Proving existence of lower variance
« Reply #11 on: May 15, 2015, 01:52:52 PM »
Yes, please. Thanks for the offer. I hope Bayes also agrees.
He made it specifically for a related subject and made it public.

My original post was
Quote
According to current probabilities theory, roulette is a game of independent trials and independent probabilities. No mainstream professor on probabilities would accept that past results can influence future results.

One of the reasons for this thesis, is that academics care about "single spin" probabilities than groups of spins.
Anyway, if someone can prove that previous spins can give us information for next spins, he would turn mainstream probability theory upside down, and most probably would win a Nobel prize.
Here's a simple test to prove that previous spins offer info for following spins.

Research
1. Take a database of past roulette spins - like that of Wiesbaden
2. Search for series of 50 consecutive spins with less than 12 appearances of one color. (that is around three standard deviations from normal distribution)
3. Record the appearances of that color in the following 50 consecutive spins.

Result
Theoretically in step 3, both colors have the same probabilities of appearing. For example, in 20 tests, 10 times should be in favor of the one color (the prevailing color in the previous 50 spins) and 10 times in favor of the other (the less appearing color in the previous 50 spins).
However, real tests show that the next 50 spins are in favor of the color that did not appear in the previous 50 spins.
Test for yourself.

Bayes said in his reply
Quote
I knocked up this little tool for experimenting -

But don't get too excited, for one thing, I coded a win to count as "not a loss" (ie; win or break even).
Also, it's one thing to get > average in the next sequence MOST of the time, but that's not the same as making a profit overall, you have to take into account how much you LOSE in those sessions where you get less than average. Another thing to be aware of are those cases where you get a rush of wins early on and then losses later;  just betting the next X spins mechanically through to the end may not be the best way to test the hypothesis. I tend to agree with Ego in that the best way is to wait for the correction to manifest before you start your attack.

Note that the 2nd sample (the one you would actually be betting in) doesn't have to be the same length as the first sample (but obviously, the trigger has to be less than the first sample!). Play around with different figures and see what you get.

I haven't actually implemented the file option yet, so you're stuck with RNG for now. Also, it can take a long time to do the analysis (250,000 spins are taken at a time), so be patient. It takes 2-3 minutes on my computer.

I'm pretty sure I coded this idea years ago and the end result in terms of profits was... you guessed it - right on the mathematical expectation.

Here it is, in the attachment.
« Last Edit: May 15, 2015, 02:00:00 PM by kav »
 

becker

  • New
  • **
  • Posts: 29
Re: Proving existence of lower variance
« Reply #12 on: May 15, 2015, 02:46:13 PM »
Thanks Kav

Well nice tool indeed, but this doesn't tell nor show anything about the sequences.

With this we can only prove that no matter how strong deviation we take, results after it will be on mathematical expectation in the long run. So only a proof of no possibility in getting an edge this way. But I don't see how using this tool we can know that we are facing or not facing lower variance?
« Last Edit: May 15, 2015, 03:55:44 PM by kav »
 

Bayes

  • Veteran Member
  • ****
  • Posts: 688
  • Thanked: 547 times
  • roulettician.com
Re: Proving existence of lower variance
« Reply #13 on: May 15, 2015, 03:21:28 PM »
I'd forgotten about that software, but It's not very useful for doing a proper statistical test. I'll write some code over the weekend and do the test. By the way, if anyone's interested, the software I'm using is free and easy to use, a bit like a spreadsheet. No knowledge of programming required - just point and click. You do need to know a bit about how to interpret results and what the various tests are for though. There are some videos on youtube, just google PSPP Statistical analysis software, or you can buy  a manual which is also an introduction to basic statistical analysis.

The software is here: https://www.gnu.org/software/pspp/

And the manual here:
[url]http://www.amazon.com/The-PSPP-Guide-Expanded-Edition-ebook/dp/B00POKE3RW[/url]
« Last Edit: May 15, 2015, 03:55:04 PM by kav »
 

kav

  • www.Roulette30.com
  • Administrator
  • Hero Member
  • *****
  • Posts: 1709
  • Thanked: 721 times
  • Gender: Male
Re: Proving existence of lower variance
« Reply #14 on: May 15, 2015, 04:01:53 PM »
Thanks Kav

Well nice tool indeed, but this doesn't tell nor show anything about the sequences.

With this we can only prove that no matter how strong deviation we take, results after it will be on mathematical expectation in the long run. So only a proof of no possibility in getting an edge this way. But I don't see how using this tool we can know that we are facing or not facing lower variance?
The thing is that none ever used that tool seriously doing thousands of tests and recording the results. Not even me. If we did that we might be surprised. So we don't even have the "no possibility" seriously tested and validated.